Question about Gigabit Switches - slight problem at home network

[g0dM@n]

I'll try to keep it brief, but I have a tough time doing that. :)

Cable Modem is in basement off MAIN split (for best connection), then goes to router, then goes to Gigabit 8port switch #1.

Since 2 of my brothers and I have each a bedroom on 2nd floor (2 flights up from basement), and I didn't wanna run all new lines, I figured I'd buy a 2nd Gigabit switch (I think this is my problem).

Server is in basement, hooked up to Switch #1. Switch #1 then has two wires going up two floors - one to my bedroom and to one brother (he ran his own line thank God). Since I now have two computers, and my other brother doesn't have a line, and since I didn't wanna run a bunch more lines from basment, I installed switch #2 in my bedroom. Switch #2 is connected directly to Switch #1 in basement. Switch #2 goes to each of my 2 computers, AND to my brothers computer in the next bedroom (we share a wall so wire goes through it and looks fairly neat).

Here's the problem: When I transfer files between my two computers (both connected to switch #2), transfer rates with gigabit (1.0gbps) are 2-3x faster than the regular 100mbps. When I try to transfer files between one of my computers and the server in the basment (through switch #2 to switch #1 to SERVER), the transfer rates seem the same speed as 100mbps (what it was before).

My questions...

1. IS MY PROBLEM THE FACT THAT I HAVE TWO GIGABIT SWITCHES??
2. DO I NEED TO RUN ALL LINES FROM ONE GIGABIT SWITCH AND COMPLETELY REMOVE THE 2ND SWITCH??

I hope this is understandable. Remember, both of my computers on switch #2 can transfer with each other fairly quicker than a normal connection, but not when I transfer with server (going through two gigabit switches), it seems normal speed.

Both of my brothers computers don't have a gigabit NIC installed yet (still gotta do that), so it's useless to test with them.

Yes, both of my computers AND the server have onboard gigabit, and when I open the network connection properties, it shows 1.0gpbs.

Thanks in advance.

[hkgonra]

My first guess would be the wiring running to the basement.
I have several switches hooked up here at work and do not run into problems using multiple switches.
Try hooking the wire currently connecting switch #1 to switch #2 directly to your pc , check it that way.
I would also try hooking that same wire back up to switch #2 upstairs and directly to the server downstairs.

[cliffboss]

the seed will go as fast as the slowest link

you could be getting interference, are any of the cable running by power cables or you cable is over 100m which again would give u interference

[Uqdroma]

Sounds like your router is providing Windows Domain regulation for the machines on Switch#1. Take the router out of the situation since this is the only component I see in your config with 100mb. That way the router is no longer part of your Windows Domain. If the problem still exists, you may want to check Ethernet cables to make sure they are at least Cat5e. Don't quote me but I don't think Cat5 (non-e) is gigabit. Further question to the uber-gurus...Wouldn't connecting to switches without a crossover cable cause the duplex to be cut to half? Or even Switch#1 to the router w/o crossover because the router has switched ports also? Just a side question from a networking intermediary.

[g0dM@n]

the seed will go as fast as the slowest link

you could be getting interference, are any of the cable running by power cables or you cable is over 100m which again would give u interference
100m = ~330ft I believe... no WAY it's that long. :)

Wire goes from switch #1 to server (~10-15ft long), then from switch#1 through attic then down my inside my wall to the rj45 wall-jack in my bedroom (this cable was originally 100ft, but i've shortened it a bit)... then i used a ~6ft wire to switch#2... this totals at MAX 120ft. From switch#2, one computer is on a 30ft wire in my bedroom, and the other is on a 9-10ft wire.

Overestimating, it's 160ft (50m).

My first guess would be the wiring running to the basement.
I have several switches hooked up here at work and do not run into problems using multiple switches.
Try hooking the wire currently connecting switch #1 to switch #2 directly to your pc , check it that way.
I would also try hooking that same wire back up to switch #2 upstairs and directly to the server downstairs.

HAHA... this is exactly what I wanted to do during lunchbreak, but decided to nap instead.

I'm going to try that sometime this week... now that you said it, I'm going to put it into my PDA as a reminder before I forget to put it into my calendar.


I'm thinking... worst case if I can't do multiple switches, I'll go into the attic and drop a line into my brother's wall and then my wall, and hook everyone upstairs to my giga switch... then, since i have a line going from my bedroom to basement, i can just hook the server up to the switch in my room...

i bought two switches so i wouldn't have to do the extra work. :(
Are you guys are sure you can use multiple gigabit switches and still maintain gigabit bandwidth through a connection that goes through BOTH switches???

[g0dM@n]

1. Sounds like your router is providing Windows Domain regulation for the machines on Switch#1. Take the router out of the situation since this is the only component I see in your config with 100mb. That way the router is no longer part of your Windows Domain.

2. If the problem still exists, you may want to check Ethernet cables to make sure they are at least Cat5e. Don't quote me but I don't think Cat5 (non-e) is gigabit. Further question to the uber-gurus...

3. Wouldn't connecting to switches without a crossover cable cause the duplex to be cut to half? Or even Switch#1 to the router w/o crossover because the router has switched ports also? Just a side question from a networking intermediary.

1. Buy how come my two computers in my bedroom that are hooked up to switch#2 can do faster connection with one another? The router is what is still giving them IP addresses.

I'm using the router b/c I want a firewall. =/ If I remove it, should I just turn on windows firewall... not sure if I trust doing that...

2. I'm almost positive I'm cat5e all around...

3. I don't think a crossover would even work between switches. I think it only works when you hook up two computers via ethernet... I could be wrong, though.


I'll have to check my cables and do further testing... I just wanted to see if anyone could tell right off the bat on if I did something wrong... I don't even know how my PCs would get IPs without a router... or how I would maintain all of my port forwarding and whatnot without a router.

[Uqdroma]

1. Buy how come my two computers in my bedroom that are hooked up to switch#2 can do faster connection with one another? The router is what is still giving them IP addresses. And might only be providing Windows domain control to Switch#1, which could be why Switch#2 is full speed.

3. I don't think a crossover would even work between switches. I think it only works when you hook up two computers via ethernet... I could be wrong, though.

Some routers provide Windows domain control as a function, in order to centralize Windows shares. I only recommended removing the router to see if that was the cause. If it was, then it needs to be disabled at the router. Removing it from the network was only meant to eliminate that as the issue. Definately keep the router for IP routing. Also, I learned that a patch cable between two non-switched routers causes half-duplex. Hence the need for a crossover between two in-line PC's. Likewise a crossover between two swithces would cause the same, effectively canceling the "crossover" capabilities of the switch.

Side note: this is why I made the analogy in a previous thread about a switch acting as a crossover. I do know, from experience, that a router without switched ports requires a crossover to the switch in order for DHCP to function on all the ports of the switch. For example...a home-built router with only two NICs. One for incoming WWW, and the other for routing (sharing). Please, uber-net-admins, correct me if I'm way off. I can take it, for I love learning correctly.

Oh I forgot to ask if all of your NICs are also gigabit?

[g0dM@n]

Some routers provide Windows domain control as a function, in order to centralize Windows shares. I only recommended removing the router to see if that was the cause. If it was, then it needs to be disabled at the router. Removing it from the network was only meant to eliminate that as the issue. Definately keep the router for IP routing. Also, I learned that a patch cable between two non-switched routers causes half-duplex. Hence the need for a crossover between two in-line PC's. Likewise a crossover between two swithces would cause the same, effectively canceling the "crossover" capabilities of the switch.

Side note: this is why I made the analogy in a previous thread about a switch acting as a crossover. I do know, from experience, that a router without switched ports requires a crossover to the switch in order for DHCP to function on all the ports of the switch. For example...a home-built router with only two NICs. One for incoming WWW, and the other for routing (sharing). Please, uber-net-admins, correct me if I'm way off. I can take it, for I love learning correctly.

Oh I forgot to ask if all of your NICs are also gigabit?

I think my problem is giving out too much info at once, lol... yes, my NICs are gigabit... I said before... the server, and BOTH of my computers (the two in my bedroom attached to switch#2) are all onboard gigabit. My 2 brothers don't have gigabit just yet, but I have their cards that I bought for them in my closet (when I have the chance i'll install those).

I don't know if I misunderstood you, but NO computers are hooked up to each other with crossover.

To make things easier...

Brother A PC#1 = ME (GIGABIT ONBOARD)
Brother A PC#2 = ME (GIGABIT ONBOARD)
Brother B = middle brother (10/100 ONBOARD)
Brother C = youngest brother (10/100 ONBOARD)
Server = Server (GIGABIT ONBOARD)

(router is 10/100 4 port WIRED, switches are both gigabit)

MAIN Cable Line to house into basement
==> Main split (3 split)
==> Cable modem (7.0dBa), TV Split (3.5dBa), TV Split (3.5dBa)
Cable Modem ==> Router ==> Switch #1
Switch #1 ==> Server (basement); Brother B (2nd floor); Switch #2 (2nd floor)
Switch #2 ==> Brother A PC#1, Brother A PC#2, Brother C

Transfers between Brother A PC#1 and Brother A PC#2 is the only transfer that seems to work faster than before (assuming it's gigabit speed - about 2-3x faster than before I got the gigabit switches installed). If I try any of those two PCs with the Server (which goes through Switch #1 AND Switch #2), it seems to still run normal speed (not gigabit).

Dunno if this info is needed, but before I got these gigabit switches, Switch #1 was just the router, and Switch #2 was a 10/100 8port switch.

As for full duplex and half duplex... I never understood how that stuff works.

Hope that makes it easier to understand. Absolutely no crossover cables involved. I thought the router and switches already do that.

Let me know if I'm still unclear. :)

[hkgonra]

I don't know how to post a screenshot but I have one of my system copying a file over the network here at work, it is chained through 3 switches back to the main switch that the server is connected to. Network utilization is 30% on my 1Gbps connection.
PM me is you want me to e-mail it to you.

[SavageBasher]

Are you guys are sure you can use multiple gigabit switches and still maintain gigabit bandwidth through a connection that goes through BOTH switches???

Yes, or else networking would be retardedly complicated. But do as the one guy said, connect the wire that goes to switch 2 directly to your computer. And make sure all of the connections are coming up as 1000mbit when connected.

[g0dM@n]

Yes, or else networking would be retardedly complicated. But do as the one guy said, connect the wire that goes to switch 2 directly to your computer. And make sure all of the connections are coming up as 1000mbit when connected.
All 3 PCs with the gigabit onboard NIC says 1.0gbps for SURE.

I don't know how to post a screenshot but I have one of my system copying a file over the network here at work, it is chained through 3 switches back to the main switch that the server is connected to. Network utilization is 30% on my 1Gbps connection.
PM me is you want me to e-mail it to you.
Speaking of which... when I CTRL+ALT+DEL and check the NETWORKING tab, I see it say GIGABIT and only do like 9% utilization (ish).

[SavageBasher]

do you have jumbo frames on on all of the nics?

[hkgonra]

My vote goes back to a cable issue then.

[fight556]

My vote also goes to the cabling between the switches, sounds to me like theres enough interference on the line to kick the auto negotiation on the switch link back to 100mb, or it is gigabit, but with lots of errors, which would point to the cabling as well.

[cliffboss]

3. I don't think a crossover would even work between switches. I think it only works when you hook up two computers via ethernet... I could be wrong, though.

between similar device u must use a crossover, unless the switch is smart and can change mdi -> mdix or vice versa

[SavageBasher]

most recent NICs and I'm pretty sure all gigabit switches auto negotiate, I don't have any crossovers anywhere in my network

[hkgonra]

between similar device u must use a crossover, unless the switch is smart and can change mdi -> mdix or vice versa

I have never seen a switch that wouldn't handle that itself, only really old hubs.
I have no doubt they exist but I assume they are pretty rare.

[NWEng]

between similar device u must use a crossover, unless the switch is smart and can change mdi -> mdix or vice versa
QFT. Switch to switch, router to router, PC to PC are all crossover cables. This is due to the TX and RX pairs, essentially with a crossover you are connecting the TX pairs of the switch/router/PC to the RX pairs of the switch/router/PC on the other end of the cable.
Can you verify by the lights that the link connecting the two switches have negotiated to 1gbs?

[Cheator]

One thing we learn in CCNA is that if you have too many repeaters between you and the router, your signal can get degraded. ITs possible that since you have 2 switches, and then a third switch in the router that it is going through, the signal is so degraded it has to lower the speed to ensure decent transfer. For retail style routers and switches, I wouldn't be suprised if this is the case.

Thats my suggestion, anyway

[g0dM@n]

One thing we learn in CCNA is that if you have too many repeaters between you and the router, your signal can get degraded. ITs possible that since you have 2 switches, and then a third switch in the router that it is going through, the signal is so degraded it has to lower the speed to ensure decent transfer. For retail style routers and switches, I wouldn't be suprised if this is the case.

Thats my suggestion, anyway
Well unfortunately I couldn't find a router+gigabit in one... and if there was one, I'm sure it's a LOT more than I paid for my two gigabit switches.

I bought the Netgear gigabit switches from slickdeals like a couple weeks apart... one is the GS608NA, and the other is the GS108NA... i don't even know the diff between the two except the 608 is sexier.

They cost me about $30-35 apiece after rebates. Thought it was a good deal, and the reason I even paid another $30-35 AR for the 2nd one was to save me time so I wouldn't have to rewire everything to one switch. :(

So I guess I should look at the switches lights where they are connected to each other? There would be like a 10mbps light, a 100mbps and a 1000mbps light for each port I assume?

So maybe I have to set up the wire between the two switches as CROSSOVER??!? I thought if I didn't have a wire set up correctly (regular vs crossover) that it wouldn't work @ all??

[torin3]

Most of the switches I've been dealing with on our LAN at work have ports that autodetect what the cable is and set themselves appropriately as needed. I terminate most of our cables here and they are wired straight. If the switch needs crossover, it switches internally to accommodate.

[g0dM@n]

Most of the switches I've been dealing with on our LAN at work have ports that autodetect what the cable is and set themselves appropriately as needed. I terminate most of our cables here and they are wired straight. If the switch needs crossover, it switches internally to accommodate.
well since the switches i bought seem fairly new, i'm sure it does the same.

tho i will check the lights and double-check asap. :) hopefully tonight.

[NWEng]

So I guess I should look at the switches lights where they are connected to each other? There would be like a 10mbps light, a 100mbps and a 1000mbps light for each port I assume?

So maybe I have to set up the wire between the two switches as CROSSOVER??!? I thought if I didn't have a wire set up correctly (regular vs crossover) that it wouldn't work @ all??
Chances are the the lights will be different colors for different connection speeds, I think the Netgear is orange for 100 and green for 1gbs.

I'm a network guy, so I do crossover from switch-switch by default; however, the switch ports may be set to auto-detect cable type and are working properly.

[torin3]

Chances are the the lights will be different colors for different connection speeds, I think the Netgear is orange for 100 and green for 1gbs.


I can confirm that that is the way it works on the Netgear GS605 (v2) switch that is on my desk now. Of course the only green light is to my PC. Over the next several weeks we are going to be upgrading most of our network to gigabit.

[torin3]

I bought the Netgear gigabit switches from slickdeals like a couple weeks apart... one is the GS608NA, and the other is the GS108NA... i don't even know the diff between the two except the 608 is sexier.



Just looked them up. The only difference is one is meant to sit on the desktop, and has a sleek case, and the other is mean to stack somewhere in an office. They both use the same power, both use jumbo frames, and both have auto-uplink ports (switch to handle straight or crossover as needed). The even use the exact same power input. 12V, 1A.

I'm not sure on the lights for the 108, but the 608 will be green for Gigabit, and orange/amber for megabit.

[Cheator]

If you connect both switches to the router it should be fine, at least if I am close to right.

[g0dM@n]

Chances are the the lights will be different colors for different connection speeds, I think the Netgear is orange for 100 and green for 1gbs.

I'm a network guy, so I do crossover from switch-switch by default; however, the switch ports may be set to auto-detect cable type and are working properly.
Well it detects which is upload (internet WAN from router)... so I'm sure it does cross-over detection as well.

In any case, even if it DIDN'T do cross-over connection detection... wouldn't the port just NOT WORK if I had it hooked up incorrectly?

[hkgonra]

Well it detects which is upload (internet WAN from router)... so I'm sure it does cross-over detection as well.

In any case, even if it DIDN'T do cross-over connection detection... wouldn't the port just NOT WORK if I had it hooked up incorrectly?

Correct, it just would not work.

[g0dM@n]

Correct, it just would not work.
so why is everyone having me check that... all connections work, just not as fast as i want, lol... if it was hooked up wrong, it wouldn't be working... meaning, crossover is no longer an issue, right? :)

i'm gonna check the status on each switch now...

[NWEng]

If a connection is not working at it's rated speed then there is more than likely a fundamental problem somewhere in your network. Network troubleshooting is always about verifying the basics (OSI layers 1 through 3 in that order,) before considering the exotic. Latency between two servers can be as simple as a bad network connection that's causing FCS, giant/runt, or CRC errors on an interface. I don't even consider something like routing (as an issue,) until I've verified that the lower layers are working as they should. That's why.

[g0dM@n]

good news and bad news...

The bad news first....
SO IT SEEMS THAT MY LINE FROM THE BASEMENT SWITCH #1 TO SWITCH #2 UPSTAIRS IS CAT 5!!!!

The good news (or so I think)...
That's why I can't do 1000mbps... you need cat5e or cat6... I'm assuming I'm correct, right?

Bad news...
I have to freakin' run lines AGAIN, so might as well try to bypass one switch.

[fight556]

Yes cat5E and cat 6 is rated for gigabit, though have you tested the cable? only pins 1,2,3,6 are used in 100mb ethernet, where as in gigabit ALL the pins are used, so a break somewhere in pins 4,5 or 7,8 would let 100mb ether work just fine, gigabit would not.

[SavageBasher]

yes, you need atleast cat5e, but if I were you, I'd get some cat6 and do it right.

[g0dM@n]

yes, you need atleast cat5e, but if I were you, I'd get some cat6 and do it right.
when would i ever need cat6?? cat5e is so cheap too....

Yes cat5E and cat 6 is rated for gigabit, though have you tested the cable? only pins 1,2,3,6 are used in 100mb ethernet, where as in gigabit ALL the pins are used, so a break somewhere in pins 4,5 or 7,8 would let 100mb ether work just fine, gigabit would not.
are you saying i can MOD the cat5 cable???

[NWEng]

No, you're fine with 5e. In Cat5 (10baseT and 100TX,) only two pair are used for ethernet, 1 TX (pins 1 & 2) and 1 RX (pins 3 & 6.) Cat5e use the other 2 pair as well, for other purposes, which elude me at the immediate moment. However, in any 4 pair cable, all 8 wires are crimped because it's easier (therefore cheaper,) to do all 8, rather than selectively doing 1, 2, 3, & 6; they're just not in use.

[Uqdroma]

Correct, it just would not work.

I totally disagree. If a crossover is needed and not used it will just simply be in half duplex. This means that data can only be sent or received on separate cycles. Full duplex will send and receive at the same time on the same cycle. I hate to break it down to "this is a football" but I just want to clarify that networking is not as cut and dry as a zif socket. Mistakes in cabling won't just render a network device useless. As far as I've learned. Networking makes my head spin, and is prolly the most confusing aspect of modern PCs. So if I'm way off here, I can accept critique from gurus more learned than I, and I will gladly find a new learning source in order to be more helpful in the future.

Review of what I've learned through research aswell as trial-n-error...

Patch cable between two switches=half duplex
Crossover cable between two switches=full duplex
All "powered" switches have crossover capability
A patch cable between the two will cause each crossover (switch) to cancel out creating a half duplex connection. A crossover cable connected to a crossover (switch) will create half duplex...when connected to a third crossover (switch) will put it back into full duplex. This is what I meant to express the whole time.

How that applies to this scenario is too much for me to determine without further education, testing, and/or research. :beer:

[SavageBasher]

Wrong. If you use a patch cable when a crossover it needed, Tx will be linked to Tx, and Rx to Rx. thus no data will be transmitted.

http://pinouts.ws/img/crossover-cable-pinout.gif

[g0dM@n]

mumbo jumbo to me!!! :D

tho i'm interested...

[Spion]

Wrong. If you use a patch cable when a crossover it needed, Tx will be linked to Tx, and Rx to Rx. thus no data will be transmitted.

http://pinouts.ws/img/crossover-cable-pinout.gif
yep, the reason why you can use a patch cable to connect switches is because some of them are smart enough to sense it's a patch cable instead of a crossover and logically change what wires they send-receive data on for that port.

[Neuromancer]

I totally disagree. If a crossover is needed and not used it will just simply be in half duplex. This means that data can only be sent or received on separate cycles. Full duplex will send and receive at the same time on the same cycle. <snip>
Maybe you are correct? Im a moron, only have an a+ and net+ cert...

Try connecting 2 PCs without a crossover cable and tell me it works. I will be AMAZED if it does.

I think thats what people were getting at. With a router in the mix... (which wasnt in this situation) you can PROBABLY connect 2 devices that should have used a CO cable, and got half duplex mode.. only because the router was compensating... OR your were using something other than Etherent. Ring networking or something.. I dont know and dont care.. .Bus and ring networking is outdated. (this seems totally wrong to me though, read the following)


I did my net+ back in 2002 and dont ever recall seeing that you can get by half duplex with using a non crossover cable. All that should happen is collision detection occurs resets and tries to send again... only to hit an obstacle and reset again (send going to send etc). Without googling I would think 1/2 duplex occurs in non-switches/routers.. and is used in trunks and ring networks (old ass technology)

However, I am far from infallible, I misread things all the time... so this could be that case.. .maybe you meant...that crossover cable ARE needed when full duplex is not available?

Just a thought...

[Uqdroma]

Maybe you are correct? Im a moron, only have an a+ and net+ cert...

Try connecting 2 PCs without a crossover cable and tell me it works. I will be AMAZED if it does.

I think thats what people were getting at. With a router in the mix... (which wasnt in this situation) you can PROBABLY connect 2 devices that should have used a CO cable, and got half duplex mode.. only because the router was compensating... OR your were using something other than Etherent. Ring networking or something.. I dont know and dont care.. .Bus and ring networking is outdated. (this seems totally wrong to me though, read the following)


I did my net+ back in 2002 and dont ever recall seeing that you can get by half duplex with using a non crossover cable. All that should happen is collision detection occurs resets and tries to send again... only to hit an obstacle and reset again (send going to send etc). Without googling I would think 1/2 duplex occurs in non-switches/routers.. and is used in trunks and ring networks (old ass technology)

However, I am far from infallible, I misread things all the time... so this could be that case.. .maybe you meant...that crossover cable ARE needed when full duplex is not available?

Just a thought...

I wasn't the one who discounted your post. Re-read who did. Also I was only referring to a patch between two switches. But I did mis-write a little. I shouldn't have stated the first line the way I did. Certainly in a case with two PCs (routers) you would need a crossover. Also a crossover is needed from a switchless router to a switch. Without a crossover signal is gone for sure. I appreciate your education sharing and didn't ever question it. Someone else may have. You may need to re-read the whole thread because throughout I have said my knowledge is intermediate and I'm not afraid to learn. It helps me fill in the parts that still confuse me. No need to posture here as I say again I wasn't the one who questioned your expertise. Again, thank you for your help on behalf of the OP, as I was only suggesting that the two switches connected to each other might need a crossover, which I did state already. For the benefit of the OP's request for help I'll move on from this thread now because no amount of education is worth this kind of posturing.

[g0dM@n]

I wasn't the one who discounted your post. Re-read who did. Also I was only referring to a patch between two switches. But I did mis-write a little. I shouldn't have stated the first line the way I did. Certainly in a case with two PCs (routers) you would need a crossover. Also a crossover is needed from a switchless router to a switch. Without a crossover signal is gone for sure. I appreciate your education sharing and didn't ever question it. Someone else may have. You may need to re-read the whole thread because throughout I have said my knowledge is intermediate and I'm not afraid to learn. It helps me fill in the parts that still confuse me. No need to posture here as I say again I wasn't the one who questioned your expertise. Again, thank you for your help on behalf of the OP, as I was only suggesting that the two switches connected to each other might need a crossover, which I did state already. For the benefit of the OP's request for help I'll move on from this thread now because no amount of education is worth this kind of posturing.
Uqdroma, you better not leave or I'll bring the hammer down on ya!!! :)

Neur0, yes you cannot do PC to PC with a patch cable... how do I know? Well, the first FIRST time I ever tried it, I had NO IDEA there was such thing as a crossover cable... I hooked up both PCs with a patch and wondered why it didn't work, lol.

Other than using a CO cable for PC to PC, I've NEVER, EVER needed one after that... no idea when you would still need a CO cable specifically other than PC to PC... since I'm always using routers and/or switches (routers are switches as well these days)... did I miss anything? :)

[fight556]

most home use switches (and some new Cisco models) have pin sensing for crossover detection. Like devices need crossover, for example switch-switch, hub-hub, pc-pc,and router-pc. though like I said most home grade devices have auto-sense on crossover.

[Uqdroma]

Certainly in a case with two PCs (routers) you would need a crossover. Also a crossover is needed from a switchless router to a switch.

*Quoting myself for correction*

Should read a "switched" router to a switch. I shouldn't get so bothered when taking offense from others. It just makes me rush and make more errors. :beer: I become non-ECC. ;)

[Adragontattoo]

A gigabit network will run with multiple switches.

It would be almost physically impossible in a large network to have no extra switches (the chances of managing to pull that off are totally NOT worth the effort of deciphering).

[Lancelot]

You're talking about Gigabit ONBOARD, but ONLY when all those systems have a dedicated (like Intel CSA?!) bus directly to the chipset you will be able to reach Gigabit burst speeds through RAM, you cannot sustain it cuz current HDD technology is still to slow. Installing Gigabit NICs in PCI slots will never get you near...

[siberia37]

You're talking about Gigabit ONBOARD, but ONLY when all those systems have a dedicated (like Intel CSA?!) bus directly to the chipset you will be able to reach Gigabit burst speeds through RAM, you cannot sustain it cuz current HDD technology is still to slow. Installing Gigabit NICs in PCI slots will never get you near...

On a regular 32-bit PCI this is true but I think a newer 64-bit PCI slot that can do 532 MB/s you will probably be ok. And no you won't be able to keep up if your copying a file from the network to your hard drive, but if your doing network routing which doesn't require data to be written to disk then maybe..

[g0dM@n]

You're talking about Gigabit ONBOARD, but ONLY when all those systems have a dedicated (like Intel CSA?!) bus directly to the chipset you will be able to reach Gigabit burst speeds through RAM, you cannot sustain it cuz current HDD technology is still to slow. Installing Gigabit NICs in PCI slots will never get you near...
so i've heard, but i never knew the exact reasoning.

the thing is... the two gigabit NICs i got for my brothers... i got them for like $1 AR or something close to that... maybe a few bucks AR each.

It's still better than their onboard 10/100 is what I was thinking when I bought them.

[Madwand]

Maybe you are correct? Im a moron, only have an a+ and net+ cert...

Try connecting 2 PCs without a crossover cable and tell me it works. I will be AMAZED if it does.


Prepare to be amazed. Try it. I've done it several times successfully.

Gigabit NICs generally have auto cable detection, so can generally do this without any problems. In fact, this could be a useful workaround / test technique for the OP and others when dealing with switch issues. If in doubt about the switch, just try plugging in the gigabit NICs directly to each other with a standard cable (and manually assign IPs for the test).

In addition to the cat 5 vs. cat 5e/etc. issue, there's a home-made vs. store-bought cable issue. If the OP has done his own termination, then there's a good chance that it's not as good as machine-made termination. These problems often show up as performance problems, not out-right failures. So I'd try a store-bought cable, actually a brand-name cable, and not risk something that a store hand-connected and put in a package on a shelf.

Cat 5 should be fine, but the cat 5e spec. was made to tighten up some of the potential problems in cat 5. For this reason, although a good cat 5 cable might work, you're generally better off getting a cat 5e cable.

[Madwand]

A couple of additional points:

(1) Since you're dealing with completely different computers, there's a chance that the problem's not at the network hardware level, but somewhere else. You should generally use a network performance measurement tool instead of doing a system-level test such as a file transfer in order to isolate these issues.

E.g. using iperf (1.7)

server: iperf -s
client: iperf -c server -l 64k -t 12 -i 3 -r

(2) The Netgear GS60x in my experience is more sensitive to cable issues than some others. What can you do? Not a whole lot... (a) Ensure that you're not using very thick cable ends and are inserting the cables fully, making good contact. (b) Try different cables. (c) Try a different switch. Here swapping the switches wouldn't really help because you'll just move the problem around.

Also ensure that not all the connected ports are blinking during transfers. The GS60x has a tendency to broadcast sometimes when it shouldn't. Now if you had a 100 Mb/s port connected at the same time, it'd be possible for all links to effectively drop down to the 100 Mb/s speed because of the broadcasting issue.

If you're seeing broadcasting, I'd try a power cycle, and try to keep the cabling stable for some time. Switching NICs, etc., IME has a tendency to start the broadcasting.

But don't get me wrong -- if you get the cabling and broadcasting stabilized, the GS60x can give amazing performance for the money.

[g0dM@n]

When I look at the GS608 up in my bedroom (switch#2), I see orange and green lights... one is 1000 and one is 100... the only 100 connection to the switch is the connection from the basement from switch#1, and that wiring from my wall jack was cat5e. From the wall jack all the way to the basement to gigabit switch #1 is what I forgot to check. Turns out it was cat5... which is why I think i was running 100 and not 1000 on that connection.

I have a cat5e long enough in my bedroom right now, but haven't had the chance to go into my attic and drop down the lines to my bedroom and all the way to basement... maybe i'll do it this weekend, but who knows if that'll happen for sure.

and i've no idea what that stuff you typed is:

E.g. using iperf (1.7)

server: iperf -s
client: iperf -c server -l 64k -t 12 -i 3 -r

[Madwand]

I have a cat5e long enough in my bedroom right now, but haven't had the chance to go into my attic and drop down the lines to my bedroom and all the way to basement... maybe i'll do it this weekend, but who knows if that'll happen for sure.

It might be easier to move the gear closer together if necessary, using the same long cables for these tests before you pull wire through walls, etc.


i've no idea what that stuff you typed is:

iperf is a command-line tool that can be used to measure network performance not considering drive speed. Doing separate measurements for this stuff can be very helpful -- there'd be no point tweaking all the networking if you were bottlenecked by the server or its HD's for example, and there is an often incorrect tendency to assume that it's a networking problem when it's something else.

But, as you now mention 100 Mb/s light, that's clearly the first problem to fix, and it is a networking issue here.

It can still be helpful to know about iperf. It's a command-line tool, so you need to invoke it from the command line and specify some parameters there. I gave some example options to start off. Here's a sample output to give you a better idea of what it does:


F:\tools\bench\iperf>iperf -c 192.168.0.167 -l 65535 -t 12 -i 3 -r
------------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 8.00 KByte (default)
------------------------------------------------------------
------------------------------------------------------------
Client connecting to 192.168.0.167, TCP port 5001
TCP window size: 8.00 KByte (default)
------------------------------------------------------------
[720] local 192.168.0.100 port 33293 connected with 192.168.0.167 port 5001
[ ID] Interval Transfer Bandwidth
[720] 0.0- 3.0 sec 337 MBytes 943 Mbits/sec
[720] 3.0- 6.0 sec 335 MBytes 938 Mbits/sec
[720] 6.0- 9.0 sec 338 MBytes 947 Mbits/sec
[720] 9.0-12.0 sec 339 MBytes 948 Mbits/sec
[720] 0.0-12.0 sec 1.32 GBytes 943 Mbits/sec
[896] local 192.168.0.100 port 5001 connected with 192.168.0.167 port 49337
[ ID] Interval Transfer Bandwidth
[896] 0.0- 3.0 sec 340 MBytes 950 Mbits/sec
[896] 3.0- 6.0 sec 339 MBytes 948 Mbits/sec
[896] 6.0- 9.0 sec 339 MBytes 947 Mbits/sec
[896] 9.0-12.0 sec 339 MBytes 948 Mbits/sec
[896] 0.0-12.0 sec 1.33 GBytes 947 Mbits/sec

[g0dM@n]

It might be easier to move the gear closer together if necessary, using the same long cables for these tests before you pull wire through walls, etc.



iperf is a command-line tool that can be used to measure network performance not considering drive speed. Doing separate measurements for this stuff can be very helpful -- there'd be no point tweaking all the networking if you were bottlenecked by the server or its HD's for example, and there is an often incorrect tendency to assume that it's a networking problem when it's something else.

But, as you now mention 100 Mb/s light, that's clearly the first problem to fix, and it is a networking issue here.

It can still be helpful to know about iperf. It's a command-line tool, so you need to invoke it from the command line and specify some parameters there. I gave some example options to start off. Here's a sample output to give you a better idea of what it does:
so i just need to install this program IPERF, and then just start>run>cmd and wallah i can do what you did? :)

[Madwand]

so i just need to install this program IPERF, and then just start>run>cmd and wallah i can do what you did? :)

Essentially, yes -- start the server, then start the client.

I use version 1.7, as I don't get correct results with 2.0.

[g0dM@n]

Essentially, yes -- start the server, then start the client.

I use version 1.7, as I don't get correct results with 2.0.

I'm a couple of years late, but I was wondering how to get this to work. I downloaded v2.0.

I am trying to use the GUI, but I'm not getting any results...
*Edit* Oh I think I have to run the program on 2 diff machines... lemme try. :)
*Edit2* wo0t!!! look below. I used the GUI. The 2 machines communicate through 2 gigabit switches. One is an 8-port GREENnet by TRENDnet, and the other is the same exact one but 5-port. What do you think?
Tried out Madwand's code above (code in post #52 - few posts above):
G:\Installation Files\iPerf\bin>iperf -c 192.168.70.141 -l 65535 -t 12 -i 3 -r
------------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 8.00 KByte (default)
------------------------------------------------------------
------------------------------------------------------------
Client connecting to 192.168.70.141, TCP port 5001
TCP window size: 8.00 KByte (default)
------------------------------------------------------------
[1852] local 192.168.70.132 port 4398 connected with 192.168.70.141 port 5001
[ ID] Interval Transfer Bandwidth
[1852] 0.0- 3.0 sec 217 MBytes 607 Mbits/sec
[1852] 3.0- 6.0 sec 225 MBytes 630 Mbits/sec
[1852] 6.0- 9.0 sec 233 MBytes 652 Mbits/sec
[1852] 9.0-12.0 sec 215 MBytes 602 Mbits/sec
[1852] 0.0-12.0 sec 891 MBytes 622 Mbits/sec
[1944] local 192.168.70.132 port 5001 connected with 192.168.70.141 port 1268
[ ID] Interval Transfer Bandwidth
[1944] 0.0- 3.0 sec 227 MBytes 635 Mbits/sec
[1944] 3.0- 6.0 sec 241 MBytes 675 Mbits/sec
[1944] 6.0- 9.0 sec 232 MBytes 648 Mbits/sec
[1944] 9.0-12.0 sec 214 MBytes 599 Mbits/sec
[1944] 0.0-12.0 sec 915 MBytes 639 Mbits/sec

*Edit3*
Now only getting about 400mbits/sec. I wonder why.
I guess that alone shows that at least I'm doing better than a 10/100 switch since I'm at 400mbits. I wonder how I get 900+ mbits like madwand!! Anyway, gnite.

[ppe1700]

can you explain the setup? what devices the network traverses? i havnt read all the thread so will go back now and edit this post if needed.
is that with a single thread?
the speed you are getting might suggest a half duplex connection.

EDIT:
i skimmed back through the thread. some previous posts are worth completely disregarding.

so, as i understand it, you have the:
cable modem > switch 1 > switch 2 > some pc's. | a file server is on switch 1 also and both switches are gigabit.

now, in practice you should get as near as gigabit transfers over the wire as being directly connected to that device with a faultless setup.

if i were you, i would run the iperf tests between the two pc's in your switch 2 and then compare the results with the same test between the server and your pc upstairs. i know you said it seems faster between upstairs pc's.

someone mentioned that some switches auto detect the need for a crossover and will automatically flip their internal wiring to meet the need. ive also known some switches to state they do this but the process to fail and fall back to half duplex. if both switches have fallen back for half duplex then this is better than one side being full and the other being half fyi.

as a note, some switches also have LED indication whether that port is full duplex or half. you should know that both sides of the link should be full duplex.
[i was going to comment here about gigabit not being able to operate at half duplex, but this is wrong. its actually 10gb that doesnt use half duplex. i mention this because cisco switches with uplink ports do not have negotiation allowed. they are locked at full duplex.]

i would like to mention that it is also possible that your server, or any of your pc's have failed negotiation with the switch and the half duplex point is there and may not be on the switch uplink.
as part of investigation for this, you can go into the NIC properties and make sure that "speed and duplex" is set to auto and nothing else.

let me know how you get on

[g0dM@n]

can you explain the setup? what devices the network traverses? i havnt read all the thread so will go back now and edit this post if needed.
is that with a single thread?
the speed you are getting might suggest a half duplex connection.

EDIT:
i skimmed back through the thread. some previous posts are worth completely disregarding.

so, as i understand it, you have the:
cable modem > switch 1 > switch 2 > some pc's. | a file server is on switch 1 also and both switches are gigabit.

now, in practice you should get as near as gigabit transfers over the wire as being directly connected to that device with a faultless setup.

if i were you, i would run the iperf tests between the two pc's in your switch 2 and then compare the results with the same test between the server and your pc upstairs. i know you said it seems faster between upstairs pc's.

someone mentioned that some switches auto detect the need for a crossover and will automatically flip their internal wiring to meet the need. ive also known some switches to state they do this but the process to fail and fall back to half duplex. if both switches have fallen back for half duplex then this is better than one side being full and the other being half fyi.

as a note, some switches also have LED indication whether that port is full duplex or half. you should know that both sides of the link should be full duplex.
[i was going to comment here about gigabit not being able to operate at half duplex, but this is wrong. its actually 10gb that doesnt use half duplex. i mention this because cisco switches with uplink ports do not have negotiation allowed. they are locked at full duplex.]

i would like to mention that it is also possible that your server, or any of your pc's have failed negotiation with the switch and the half duplex point is there and may not be on the switch uplink.
as part of investigation for this, you can go into the NIC properties and make sure that "speed and duplex" is set to auto and nothing else.

let me know how you get on

I'm very sorry. The OP explains my old home (where my rents and brothers live). I live with my gf for a yr now lol... I'm sry about that.

The setup I'm talking about now is this (GS = Gigabit Switch):
Cable modem -> Netgear DD-WRT Wireless N Router -> GS #1 (8-port).

GS#1 then goes to two PCs (one is my main rig that I'm using as the client for iPerp).
GS#1 also goes to GS#2 (5-port) through a 100ft cat6 cable (soon to be shortened).

The HTPC and my Xbox 360 downstairs are connected to GS#2.

I'm running the iPerp server on my HTPC (which is also my file-server here).

I hope this helps. One other thing I was wondering is what the heck is jumbo frames and is it important?? Not that I want to spend more money to change out these switches (which say they support jumbo frames):
GS#1 - TRENDnet TEG-S80G 10/100/1000Mbps GREENnet Switch (8-port) (http://www.newegg.com/Product/Product.aspx?Item=N82E16833156251)
GS#2 - TRENDnet TEG-S50G 10/100/1000Mbps GREENnet Switch (5-port) (http://www.newegg.com/Product/Product.aspx?Item=N82E16833156250)
I never understood the specs anyway except for the watts, the bandwidth, and the # of ports LOL

[nd4spdbh2]

with gigabit ur gonna max file transfer rates round 250-400mbit ... its not like 100mbit where its easy to get 90% bandwidth usage... iperf will usually show you that your connection is 650-900mbit but in actuallity its less. ... believe me i have gone through everything u have changing settings trying to tweak stuff, gettin dedicated pcie 1x gigabit nics, using cat 6 cat 5e all different types of cables and it usually ends up file transfering round 250-400mbit depending on the computer and hds... it seems as though there is some overhead as i can only get maybe 60-70% the speed of the slower hd in the transfer.

[g0dM@n]

with gigabit ur gonna max file transfer rates round 250-400mbit ... its not like 100mbit where its easy to get 90% bandwidth usage... iperf will usually show you that your connection is 650-900mbit but in actuallity its less. ... believe me i have gone through everything u have changing settings trying to tweak stuff, gettin dedicated pcie 1x gigabit nics, using cat 6 cat 5e all different types of cables and it usually ends up file transfering round 250-400mbit depending on the computer and hds... it seems as though there is some overhead as i can only get maybe 60-70% the speed of the slower hd in the transfer.

Maybe if I bought one of the enterprise gigabit switches... :)

How about a straight connection from PC to PC through gigabit NICs (or onboard)... no switch... have you tried that?

[gangaskan]

when would i ever need cat6?? cat5e is so cheap


the one biggest reason is expansion, cat5e has a max data transfer of gigabit speeds, whereas cat 6 can run up to the 10gig mark with cable length taken into consideration.

its all what you want to do, future upgrading, or cheap and comfortable with gigabit speeds.

[g0dM@n]

the one biggest reason is expansion, cat5e has a max data transfer of gigabit speeds, whereas cat 6 can run up to the 10gig mark with cable length taken into consideration.

its all what you want to do, future upgrading, or cheap and comfortable with gigabit speeds.

If they are to have 10gbit mainstream, it probably would cost a lot for a few yrs. I'll just redo what I have to do in the future. :)

I do have a 100ft cat6 patch cable. I'm thinking of breaking it down into ~60-70ft and the rest for something else... or I could just use the ~70ft of cat5e solid cabling that I have to do a wall run...

[gangaskan]

If they are to have 10gbit mainstream, it probably would cost a lot for a few yrs. I'll just redo what I have to do in the future. :)


just letting you know what the whole reason was :) that's exactly why companies wire cat 6 (or whatever is current and reasonably cost) mostly, due to pulling cable by contractor can be quite expensive because they might charge by footage run, or per cable per run its all up to the contractor and the negotiations with the enterprise they are doing work for.

[nd4spdbh2]

Maybe if I bought one of the enterprise gigabit switches... :)

How about a straight connection from PC to PC through gigabit NICs (or onboard)... no switch... have you tried that?

i dont run a switch on my gigabit network...

i have my htpc and my main rig connected to my server via cat5e and a cat6 cable. Main rig uses onboard marvell yukon pcie1x to servers add in pcie1x marvell yukon nic. Htpc is its onboard realtek pcie1x to servers onboard realtek pcie1x nic. both get right round the same speed. Like i said i have tried MANY things and nothin really changes it...

[Madwand]

with gigabit ur gonna max file transfer rates round 250-400mbit ... its not like 100mbit where its easy to get 90% bandwidth usage... iperf will usually show you that your connection is 650-900mbit but in actuallity its less. ... believe me i have gone through everything u have changing settings trying to tweak stuff, gettin dedicated pcie 1x gigabit nics, using cat 6 cat 5e all different types of cables and it usually ends up file transfering round 250-400mbit depending on the computer and hds... it seems as though there is some overhead as i can only get maybe 60-70% the speed of the slower hd in the transfer.

This is a mostly valid point, in that typical actual file transfer performance over gigabit, without tuning or luck, is around 25 MB/s (= 200 Mb/s), nowhere near 600 Mb/s, let alone 900 Mb/s. The iperf measurements however give some important information, what your network is capable of, without consideration of other bottlenecks including the HDs and OS. In particular, if you're getting say 600 Mb/s under iperf, and only 200 Mb/s during actual file transfers, then there's pretty much no point going out and spending a small fortune on enterprise switches and NICs, because you have just established that the network isn't the main bottleneck in your setup.

It's much harder to maximize gigabit than 100 Mb/s networking, because the bar is higher and other limitations come into play. You'd have to dig up a HD a decade or so old before it became the bottleneck with 100 Mb/s networking, and would have a hard time finding a single HD which is not a bottleneck at some point with good gigabit networking. Such limitations and inefficiencies add up, so that it's very unlikely that you'll hit the HD speed when transferring over gigabit, as that would be an ideal theoretical maximum at that point, and you'd be more likely to maximize gigabit with HD's which hit well over 125 MB/s, which is impractical without decent RAID arrays or fast SSDs, on both ends.

But it can be done, at least with Vista. E.g. a screenshot of a couple of very large file transfers, one at 117.7 MB/s (~940 Mb/s). These were done with RAID on both ends, with consumer hardware, without jumbo frames IIRC.

http://i89.photobucket.com/albums/k203/Madwand0/vista-111-MiBps.png

[g0dM@n]

But it can be done, at least with Vista. E.g. a screenshot of a couple of very large file transfers, one at 117.7 MB/s (~940 Mb/s). These were done with RAID on both ends, with consumer hardware, without jumbo frames IIRC.

http://i89.photobucket.com/albums/k203/Madwand0/vista-111-MiBps.png

That is BOMB!!!!
I just used a stop watch and it took about 18 seconds to transfer a 316mb folder from my raid0 raptors (main rig) to my HTPC (single 1.5tb seagate). That's about 17.5 megabytes/sec (140mbits/sec). :( I was hoping for something 2-3x that.

*Edit*
Okay, I just transferred the same folder from the HTPC (single 1.5tb seagate) to my raid0 raptors and it only took 7 seconds!! That's about 45mbytes/sec (360mbits/sec). Now that's more like it.

*Edit2*
Raid0 raptors to HTPC - 580mb folder. ~24 seconds, 22.2 megabytes/sec
Reverse direction --> ~13 seconds, 44.6 megabytes/sec

Hmmm... I guess the write speed on the 1.5tb is that much slower? Heck I'm sure it's way too complicated to figure it out. Interesting though. :) Btw, lol... look at the stopwatch I used here (http://www.online-stopwatch.com/).

[ppe1700]

thats a high class stopwatch :D

later today if i can i will connect up two laptops back to back and use jperf to get some gigabit throughput. i cant see why the two would not be able to get 900mbit out of a gig connection using a crossover (unless things like ram, and motherboard couldnt do it - im assuming jperf runs completely in ram).

therefore doing the same test with two switches in the way should still be very close to 900mbit. unless the switches are poor. but the understanding is the way switching works is the switch uses a special chip that reads the mac address and this is done at wire speed. therefore the same or very very close to the two talking devices being directly connected with a crossover. i understand it depends on the components used but in theory it should be very very close.

with your 100ft cable... i wouldnt worry too much about this, unless there was something it was running next to introducing interference, ie badly terminated power sockets or plugs or something like that. it would be better if it was shorter, but 100ft is allowed before your supposed to start getting problems. im guessing the cable is coiled? this might introduce interference though.

i would like to see the jperf results from now, compared to what they are after you have shortened the cable to see if it was worth it.

[ppe1700]

here are my JPERF tests between a Dell inspiron laptop with a Gigabit NIC and a HP nx6125 laptop with a gigabit nic.

the machines are connected via a crossover cable. all nic settings are left to auto / auto. nics have established a gig connection, i assume full duplex at this point.


bin/iperf.exe -c 192.168.192.1 -P 1 -i 1 -p 5001 -f k -t 10
------------------------------------------------------------
Client connecting to 192.168.192.1, TCP port 5001
TCP window size: 8.00 KByte (default)
------------------------------------------------------------
[108] local 192.168.192.2 port 26641 connected with 192.168.192.1 port 5001
[ ID] Interval Transfer Bandwidth
[108] 0.0- 1.0 sec 42672 KBytes 349569 Kbits/sec
[108] 1.0- 2.0 sec 42696 KBytes 349766 Kbits/sec
[108] 2.0- 3.0 sec 40824 KBytes 334430 Kbits/sec
[108] 3.0- 4.0 sec 22896 KBytes 187564 Kbits/sec
[108] 4.0- 5.0 sec 37640 KBytes 308347 Kbits/sec
[108] 5.0- 6.0 sec 41184 KBytes 337379 Kbits/sec
[108] 6.0- 7.0 sec 28560 KBytes 233964 Kbits/sec
[108] 7.0- 8.0 sec 17456 KBytes 143000 Kbits/sec
[108] 8.0- 9.0 sec 38824 KBytes 318046 Kbits/sec
[108] 9.0-10.0 sec 43168 KBytes 353632 Kbits/sec
[108] 0.0-10.0 sec 355928 KBytes 290762 Kbits/sec
Done.

bin/iperf.exe -c 192.168.192.1 -P 1 -i 1 -p 5001 -f m -t 10
------------------------------------------------------------
Client connecting to 192.168.192.1, TCP port 5001
TCP window size: 0.01 MByte (default)
------------------------------------------------------------
[108] local 192.168.192.2 port 26642 connected with 192.168.192.1 port 5001
[ ID] Interval Transfer Bandwidth
[108] 0.0- 1.0 sec 39.0 MBytes 327 Mbits/sec
[108] 1.0- 2.0 sec 43.0 MBytes 361 Mbits/sec
[108] 2.0- 3.0 sec 18.5 MBytes 155 Mbits/sec
[108] 3.0- 4.0 sec 35.8 MBytes 300 Mbits/sec
[108] 4.0- 5.0 sec 42.4 MBytes 356 Mbits/sec
[108] 5.0- 6.0 sec 40.6 MBytes 341 Mbits/sec
[108] 6.0- 7.0 sec 19.0 MBytes 159 Mbits/sec
[108] 7.0- 8.0 sec 41.7 MBytes 349 Mbits/sec
[108] 8.0- 9.0 sec 43.0 MBytes 361 Mbits/sec
[108] 9.0-10.0 sec 42.0 MBytes 352 Mbits/sec
[108] 0.0-10.0 sec 365 MBytes 306 Mbits/sec
Done.

bin/iperf.exe -c 192.168.192.1 -P 5 -i 1 -p 5001 -f m -t 10
------------------------------------------------------------
Client connecting to 192.168.192.1, TCP port 5001
TCP window size: 0.01 MByte (default)
------------------------------------------------------------
[148] local 192.168.192.2 port 26647 connected with 192.168.192.1 port 5001
[136] local 192.168.192.2 port 26646 connected with 192.168.192.1 port 5001
[116] local 192.168.192.2 port 26644 connected with 192.168.192.1 port 5001
[128] local 192.168.192.2 port 26645 connected with 192.168.192.1 port 5001
[108] local 192.168.192.2 port 26643 connected with 192.168.192.1 port 5001
[ ID] Interval Transfer Bandwidth
[148] 0.0- 1.0 sec 2.20 MBytes 18.5 Mbits/sec
[136] 0.0- 1.0 sec 3.93 MBytes 33.0 Mbits/sec
[116] 0.0- 1.0 sec 9.97 MBytes 83.6 Mbits/sec
[128] 0.0- 1.0 sec 13.6 MBytes 114 Mbits/sec
[108] 0.0- 1.0 sec 10.8 MBytes 90.6 Mbits/sec
[SUM] 0.0- 1.0 sec 40.5 MBytes 340 Mbits/sec
[116] 1.0- 2.0 sec 7.28 MBytes 61.1 Mbits/sec
[148] 1.0- 2.0 sec 6.56 MBytes 55.1 Mbits/sec
[136] 1.0- 2.0 sec 10.4 MBytes 87.6 Mbits/sec
[128] 1.0- 2.0 sec 9.44 MBytes 79.2 Mbits/sec
[108] 1.0- 2.0 sec 8.11 MBytes 68.0 Mbits/sec
[SUM] 1.0- 2.0 sec 41.8 MBytes 351 Mbits/sec
[148] 2.0- 3.0 sec 6.61 MBytes 55.4 Mbits/sec
[116] 2.0- 3.0 sec 11.2 MBytes 93.8 Mbits/sec
[128] 2.0- 3.0 sec 10.5 MBytes 88.1 Mbits/sec
[108] 2.0- 3.0 sec 9.66 MBytes 81.1 Mbits/sec
[136] 2.0- 3.0 sec 6.35 MBytes 53.3 Mbits/sec
[SUM] 2.0- 3.0 sec 44.3 MBytes 372 Mbits/sec
[136] 3.0- 4.0 sec 10.6 MBytes 89.1 Mbits/sec
[148] 3.0- 4.0 sec 11.4 MBytes 95.2 Mbits/sec
[ ID] Interval Transfer Bandwidth
[116] 3.0- 4.0 sec 6.35 MBytes 53.3 Mbits/sec
[128] 3.0- 4.0 sec 6.66 MBytes 55.9 Mbits/sec
[108] 3.0- 4.0 sec 9.96 MBytes 83.6 Mbits/sec
[SUM] 3.0- 4.0 sec 44.9 MBytes 377 Mbits/sec
[148] 4.0- 5.0 sec 10.6 MBytes 88.6 Mbits/sec
[136] 4.0- 5.0 sec 10.5 MBytes 88.0 Mbits/sec
[128] 4.0- 5.0 sec 10.5 MBytes 88.3 Mbits/sec
[108] 4.0- 5.0 sec 6.91 MBytes 58.0 Mbits/sec
[116] 4.0- 5.0 sec 7.02 MBytes 58.9 Mbits/sec
[SUM] 4.0- 5.0 sec 45.5 MBytes 382 Mbits/sec
[136] 5.0- 6.0 sec 8.86 MBytes 74.3 Mbits/sec
[148] 5.0- 6.0 sec 8.75 MBytes 73.4 Mbits/sec
[116] 5.0- 6.0 sec 9.07 MBytes 76.1 Mbits/sec
[128] 5.0- 6.0 sec 9.50 MBytes 79.7 Mbits/sec
[108] 5.0- 6.0 sec 6.05 MBytes 50.7 Mbits/sec
[SUM] 5.0- 6.0 sec 42.2 MBytes 354 Mbits/sec
[136] 6.0- 7.0 sec 9.52 MBytes 79.8 Mbits/sec
[148] 6.0- 7.0 sec 8.94 MBytes 75.0 Mbits/sec
[116] 6.0- 7.0 sec 8.73 MBytes 73.2 Mbits/sec
[128] 6.0- 7.0 sec 8.97 MBytes 75.2 Mbits/sec
[ ID] Interval Transfer Bandwidth
[108] 6.0- 7.0 sec 8.05 MBytes 67.6 Mbits/sec
[SUM] 6.0- 7.0 sec 44.2 MBytes 371 Mbits/sec
[148] 7.0- 8.0 sec 9.52 MBytes 79.9 Mbits/sec
[136] 7.0- 8.0 sec 8.74 MBytes 73.3 Mbits/sec
[116] 7.0- 8.0 sec 9.05 MBytes 76.0 Mbits/sec
[128] 7.0- 8.0 sec 8.80 MBytes 73.8 Mbits/sec
[108] 7.0- 8.0 sec 9.30 MBytes 78.0 Mbits/sec
[SUM] 7.0- 8.0 sec 45.4 MBytes 381 Mbits/sec
[136] 8.0- 9.0 sec 8.67 MBytes 72.7 Mbits/sec
[148] 8.0- 9.0 sec 8.84 MBytes 74.2 Mbits/sec
[116] 8.0- 9.0 sec 6.30 MBytes 52.8 Mbits/sec
[108] 8.0- 9.0 sec 8.70 MBytes 72.9 Mbits/sec
[128] 8.0- 9.0 sec 8.97 MBytes 75.2 Mbits/sec
[SUM] 8.0- 9.0 sec 41.5 MBytes 348 Mbits/sec
[148] 9.0-10.0 sec 9.73 MBytes 81.6 Mbits/sec
[136] 9.0-10.0 sec 8.98 MBytes 75.4 Mbits/sec
[116] 9.0-10.0 sec 8.43 MBytes 70.7 Mbits/sec
[128] 9.0-10.0 sec 8.76 MBytes 73.5 Mbits/sec
[136] 0.0-10.0 sec 86.6 MBytes 72.6 Mbits/sec
[116] 0.0-10.0 sec 83.4 MBytes 69.9 Mbits/sec
[ ID] Interval Transfer Bandwidth
[148] 0.0-10.0 sec 83.1 MBytes 69.6 Mbits/sec
[108] 9.0-10.0 sec 8.82 MBytes 74.0 Mbits/sec
[SUM] 9.0-10.0 sec 44.7 MBytes 375 Mbits/sec
[128] 0.0-10.0 sec 95.7 MBytes 80.2 Mbits/sec
[108] 0.0-10.0 sec 86.4 MBytes 72.3 Mbits/sec
[SUM] 0.0-10.0 sec 435 MBytes 364 Mbits/sec
Done.

bin/iperf.exe -c 192.168.192.1 -u -P 5 -i 1 -p 5001 -f m -b 1000M -t 10 -T 1
------------------------------------------------------------
Client connecting to 192.168.192.1, UDP port 5001
Sending 1470 byte datagrams
UDP buffer size: 0.01 MByte (default)
------------------------------------------------------------
[144] local 192.168.192.2 port 64180 connected with 192.168.192.1 port 5001
[136] local 192.168.192.2 port 64179 connected with 192.168.192.1 port 5001
[116] local 192.168.192.2 port 64177 connected with 192.168.192.1 port 5001
[128] local 192.168.192.2 port 64178 connected with 192.168.192.1 port 5001
[108] local 192.168.192.2 port 64176 connected with 192.168.192.1 port 5001
[ ID] Interval Transfer Bandwidth
[144] 0.0- 1.0 sec 6.29 MBytes 52.8 Mbits/sec
[116] 0.0- 1.0 sec 6.59 MBytes 55.3 Mbits/sec
[136] 0.0- 1.0 sec 6.56 MBytes 55.0 Mbits/sec
[128] 0.0- 1.0 sec 6.33 MBytes 53.1 Mbits/sec
[108] 0.0- 1.0 sec 6.54 MBytes 54.9 Mbits/sec
[SUM] 0.0- 1.0 sec 32.3 MBytes 271 Mbits/sec
[144] 1.0- 2.0 sec 6.31 MBytes 52.9 Mbits/sec
[136] 1.0- 2.0 sec 6.88 MBytes 57.8 Mbits/sec
[116] 1.0- 2.0 sec 6.87 MBytes 57.6 Mbits/sec
[128] 1.0- 2.0 sec 6.54 MBytes 54.9 Mbits/sec
[108] 1.0- 2.0 sec 6.55 MBytes 54.9 Mbits/sec
[SUM] 1.0- 2.0 sec 33.1 MBytes 278 Mbits/sec
[144] 2.0- 3.0 sec 6.68 MBytes 56.0 Mbits/sec
[116] 2.0- 3.0 sec 6.74 MBytes 56.6 Mbits/sec
[136] 2.0- 3.0 sec 6.89 MBytes 57.8 Mbits/sec
[128] 2.0- 3.0 sec 6.40 MBytes 53.7 Mbits/sec
[108] 2.0- 3.0 sec 6.74 MBytes 56.5 Mbits/sec
[SUM] 2.0- 3.0 sec 33.4 MBytes 281 Mbits/sec
[144] 3.0- 4.0 sec 4.42 MBytes 37.0 Mbits/sec
[136] 3.0- 4.0 sec 4.55 MBytes 38.2 Mbits/sec
[ ID] Interval Transfer Bandwidth
[116] 3.0- 4.0 sec 4.49 MBytes 37.7 Mbits/sec
[128] 3.0- 4.0 sec 4.32 MBytes 36.2 Mbits/sec
[108] 3.0- 4.0 sec 4.44 MBytes 37.2 Mbits/sec
[SUM] 3.0- 4.0 sec 22.2 MBytes 186 Mbits/sec
[144] 4.0- 5.0 sec 5.36 MBytes 45.0 Mbits/sec
[128] 4.0- 5.0 sec 5.59 MBytes 46.9 Mbits/sec
[136] 4.0- 5.0 sec 5.67 MBytes 47.6 Mbits/sec
[108] 4.0- 5.0 sec 5.46 MBytes 45.8 Mbits/sec
[116] 4.0- 5.0 sec 5.46 MBytes 45.8 Mbits/sec
[SUM] 4.0- 5.0 sec 27.5 MBytes 231 Mbits/sec
[144] 5.0- 6.0 sec 5.51 MBytes 46.3 Mbits/sec
[128] 5.0- 6.0 sec 5.74 MBytes 48.2 Mbits/sec
[116] 5.0- 6.0 sec 3.33 MBytes 27.9 Mbits/sec
[108] 5.0- 6.0 sec 5.83 MBytes 48.9 Mbits/sec
[136] 5.0- 6.0 sec 3.30 MBytes 27.7 Mbits/sec
[SUM] 5.0- 6.0 sec 23.7 MBytes 199 Mbits/sec
[144] 6.0- 7.0 sec 5.17 MBytes 43.4 Mbits/sec
[136] 6.0- 7.0 sec 3.23 MBytes 27.1 Mbits/sec
[128] 6.0- 7.0 sec 5.88 MBytes 49.3 Mbits/sec
[108] 6.0- 7.0 sec 5.36 MBytes 45.0 Mbits/sec
[ ID] Interval Transfer Bandwidth
[116] 6.0- 7.0 sec 3.13 MBytes 26.3 Mbits/sec
[SUM] 6.0- 7.0 sec 22.8 MBytes 191 Mbits/sec
[144] 7.0- 8.0 sec 3.67 MBytes 30.8 Mbits/sec
[136] 7.0- 8.0 sec 2.97 MBytes 24.9 Mbits/sec
[116] 7.0- 8.0 sec 2.96 MBytes 24.8 Mbits/sec
[128] 7.0- 8.0 sec 3.79 MBytes 31.8 Mbits/sec
[108] 7.0- 8.0 sec 3.71 MBytes 31.1 Mbits/sec
[SUM] 7.0- 8.0 sec 17.1 MBytes 143 Mbits/sec
[144] 8.0- 9.0 sec 5.60 MBytes 47.0 Mbits/sec
[116] 8.0- 9.0 sec 5.62 MBytes 47.2 Mbits/sec
[136] 8.0- 9.0 sec 5.63 MBytes 47.2 Mbits/sec
[128] 8.0- 9.0 sec 5.29 MBytes 44.3 Mbits/sec
[108] 8.0- 9.0 sec 5.48 MBytes 46.0 Mbits/sec
[SUM] 8.0- 9.0 sec 27.6 MBytes 232 Mbits/sec
[144] 9.0-10.0 sec 6.45 MBytes 54.1 Mbits/sec
[144] 0.0-10.0 sec 55.5 MBytes 46.5 Mbits/sec
[128] 0.0-10.0 sec 56.6 MBytes 47.5 Mbits/sec
read failed: Connection reset by peer
write failed: Successful WSAStartup not yet performed.
select failed: Successful WSAStartup not yet performed.
Done.

bin/iperf.exe -c 192.168.192.1 -P 1 -i 1 -p 5001 -f m -t 10
------------------------------------------------------------
Client connecting to 192.168.192.1, TCP port 5001
TCP window size: 0.01 MByte (default)
------------------------------------------------------------
[108] local 192.168.192.2 port 26656 connected with 192.168.192.1 port 5001
[ ID] Interval Transfer Bandwidth
[108] 0.0- 1.0 sec 40.4 MBytes 339 Mbits/sec
[108] 1.0- 2.0 sec 18.9 MBytes 159 Mbits/sec
[108] 2.0- 3.0 sec 42.9 MBytes 359 Mbits/sec
[108] 3.0- 4.0 sec 26.7 MBytes 224 Mbits/sec
[108] 4.0- 5.0 sec 35.4 MBytes 297 Mbits/sec
[108] 5.0- 6.0 sec 42.9 MBytes 360 Mbits/sec
[108] 6.0- 7.0 sec 15.5 MBytes 130 Mbits/sec
[108] 7.0- 8.0 sec 22.6 MBytes 189 Mbits/sec
[108] 8.0- 9.0 sec 40.2 MBytes 338 Mbits/sec
[108] 9.0-10.0 sec 41.7 MBytes 350 Mbits/sec
[108] 0.0-10.0 sec 327 MBytes 274 Mbits/sec
Done.

bin/iperf.exe -c 192.168.192.1 -P 2 -i 1 -p 5001 -f m -t 1
------------------------------------------------------------
Client connecting to 192.168.192.1, TCP port 5001
TCP window size: 0.01 MByte (default)
------------------------------------------------------------
[116] local 192.168.192.2 port 26659 connected with 192.168.192.1 port 5001
[108] local 192.168.192.2 port 26658 connected with 192.168.192.1 port 5001
[ ID] Interval Transfer Bandwidth
[108] 0.0- 1.0 sec 27.7 MBytes 232 Mbits/sec
[116] 0.0- 1.0 sec 14.5 MBytes 121 Mbits/sec
[SUM] 0.0- 1.0 sec 42.1 MBytes 353 Mbits/sec
[108] 0.0- 1.0 sec 27.7 MBytes 231 Mbits/sec
[116] 0.0- 1.0 sec 14.5 MBytes 121 Mbits/sec
[SUM] 0.0- 1.0 sec 42.1 MBytes 351 Mbits/sec
Done.

bin/iperf.exe -c 192.168.192.1 -P 2 -i 1 -p 5001 -f m -t 20
------------------------------------------------------------
Client connecting to 192.168.192.1, TCP port 5001
TCP window size: 0.01 MByte (default)
------------------------------------------------------------
[116] local 192.168.192.2 port 26661 connected with 192.168.192.1 port 5001
[108] local 192.168.192.2 port 26660 connected with 192.168.192.1 port 5001
[ ID] Interval Transfer Bandwidth
[116] 0.0- 1.0 sec 19.1 MBytes 160 Mbits/sec
[108] 0.0- 1.0 sec 24.4 MBytes 204 Mbits/sec
[SUM] 0.0- 1.0 sec 43.4 MBytes 364 Mbits/sec
[116] 1.0- 2.0 sec 23.3 MBytes 195 Mbits/sec
[108] 1.0- 2.0 sec 22.6 MBytes 190 Mbits/sec
[SUM] 1.0- 2.0 sec 45.9 MBytes 385 Mbits/sec
[116] 2.0- 3.0 sec 13.1 MBytes 110 Mbits/sec
[108] 2.0- 3.0 sec 16.1 MBytes 135 Mbits/sec
[SUM] 2.0- 3.0 sec 29.2 MBytes 245 Mbits/sec
[116] 3.0- 4.0 sec 15.6 MBytes 131 Mbits/sec
[108] 3.0- 4.0 sec 15.7 MBytes 132 Mbits/sec
[SUM] 3.0- 4.0 sec 31.3 MBytes 262 Mbits/sec
[108] 4.0- 5.0 sec 16.6 MBytes 139 Mbits/sec
[116] 4.0- 5.0 sec 16.6 MBytes 139 Mbits/sec
[SUM] 4.0- 5.0 sec 33.2 MBytes 278 Mbits/sec
[116] 5.0- 6.0 sec 16.4 MBytes 138 Mbits/sec
[108] 5.0- 6.0 sec 16.7 MBytes 140 Mbits/sec
[SUM] 5.0- 6.0 sec 33.1 MBytes 278 Mbits/sec
[108] 6.0- 7.0 sec 15.9 MBytes 134 Mbits/sec
[116] 6.0- 7.0 sec 16.0 MBytes 134 Mbits/sec
[ ID] Interval Transfer Bandwidth
[SUM] 6.0- 7.0 sec 31.9 MBytes 268 Mbits/sec
[108] 7.0- 8.0 sec 15.7 MBytes 132 Mbits/sec
[116] 7.0- 8.0 sec 14.5 MBytes 121 Mbits/sec
[SUM] 7.0- 8.0 sec 30.2 MBytes 253 Mbits/sec
[116] 8.0- 9.0 sec 14.1 MBytes 119 Mbits/sec
[108] 8.0- 9.0 sec 15.7 MBytes 132 Mbits/sec
[SUM] 8.0- 9.0 sec 29.9 MBytes 251 Mbits/sec
[116] 9.0-10.0 sec 15.7 MBytes 131 Mbits/sec
[108] 9.0-10.0 sec 14.5 MBytes 122 Mbits/sec
[SUM] 9.0-10.0 sec 30.2 MBytes 253 Mbits/sec
[116] 10.0-11.0 sec 16.5 MBytes 139 Mbits/sec
[108] 10.0-11.0 sec 16.5 MBytes 138 Mbits/sec
[SUM] 10.0-11.0 sec 33.0 MBytes 277 Mbits/sec
[116] 11.0-12.0 sec 15.2 MBytes 128 Mbits/sec
[108] 11.0-12.0 sec 15.0 MBytes 126 Mbits/sec
[SUM] 11.0-12.0 sec 30.2 MBytes 253 Mbits/sec
[108] 12.0-13.0 sec 16.6 MBytes 139 Mbits/sec
[116] 12.0-13.0 sec 16.6 MBytes 139 Mbits/sec
[SUM] 12.0-13.0 sec 33.2 MBytes 278 Mbits/sec
[108] 13.0-14.0 sec 16.8 MBytes 141 Mbits/sec
[ ID] Interval Transfer Bandwidth
[116] 13.0-14.0 sec 16.5 MBytes 138 Mbits/sec
[SUM] 13.0-14.0 sec 33.3 MBytes 280 Mbits/sec
[116] 14.0-15.0 sec 15.3 MBytes 128 Mbits/sec
[108] 14.0-15.0 sec 15.3 MBytes 129 Mbits/sec
[SUM] 14.0-15.0 sec 30.6 MBytes 257 Mbits/sec
[116] 15.0-16.0 sec 17.2 MBytes 144 Mbits/sec
[108] 15.0-16.0 sec 16.9 MBytes 142 Mbits/sec
[SUM] 15.0-16.0 sec 34.1 MBytes 286 Mbits/sec
[108] 16.0-17.0 sec 16.7 MBytes 140 Mbits/sec
[116] 16.0-17.0 sec 16.8 MBytes 141 Mbits/sec
[SUM] 16.0-17.0 sec 33.5 MBytes 281 Mbits/sec
[108] 17.0-18.0 sec 16.5 MBytes 138 Mbits/sec
[116] 17.0-18.0 sec 15.3 MBytes 129 Mbits/sec
[SUM] 17.0-18.0 sec 31.8 MBytes 267 Mbits/sec
[116] 18.0-19.0 sec 12.1 MBytes 102 Mbits/sec
[108] 18.0-19.0 sec 11.9 MBytes 99.8 Mbits/sec
[SUM] 18.0-19.0 sec 24.0 MBytes 201 Mbits/sec
[116] 19.0-20.0 sec 16.3 MBytes 136 Mbits/sec
[108] 19.0-20.0 sec 16.1 MBytes 135 Mbits/sec
[SUM] 19.0-20.0 sec 32.4 MBytes 272 Mbits/sec
[ ID] Interval Transfer Bandwidth
[108] 0.0-20.0 sec 332 MBytes 139 Mbits/sec
[116] 0.0-20.0 sec 322 MBytes 135 Mbits/sec
[SUM] 0.0-20.0 sec 654 MBytes 274 Mbits/sec
Done.


numerous tests show a maximum throughput of just over 350mbit/s... very poor

i manually set the tcp window size using jperf to 1500bytes and then 2500 bytes respectively. i understand that windows overwrites this as it passes through the stack:

bin/iperf.exe -c 192.168.192.1 -P 2 -i 1 -p 5001 -w 2500K -f m -t 20
------------------------------------------------------------
Client connecting to 192.168.192.1, TCP port 5001
TCP window size: 2.44 MByte
------------------------------------------------------------
[116] local 192.168.192.2 port 58036 connected with 192.168.192.1 port 5001
[108] local 192.168.192.2 port 58035 connected with 192.168.192.1 port 5001
[ ID] Interval Transfer Bandwidth
[116] 0.0- 1.0 sec 29.8 MBytes 250 Mbits/sec
[108] 0.0- 1.0 sec 20.0 MBytes 167 Mbits/sec
[SUM] 0.0- 1.0 sec 49.7 MBytes 417 Mbits/sec
[108] 1.0- 2.0 sec 35.5 MBytes 298 Mbits/sec
[116] 1.0- 2.0 sec 8.91 MBytes 74.8 Mbits/sec
[SUM] 1.0- 2.0 sec 44.4 MBytes 373 Mbits/sec
[108] 2.0- 3.0 sec 16.1 MBytes 135 Mbits/sec
[116] 2.0- 3.0 sec 24.9 MBytes 209 Mbits/sec
[SUM] 2.0- 3.0 sec 40.9 MBytes 343 Mbits/sec
[108] 3.0- 4.0 sec 24.8 MBytes 208 Mbits/sec
[116] 3.0- 4.0 sec 12.5 MBytes 105 Mbits/sec
[SUM] 3.0- 4.0 sec 37.3 MBytes 313 Mbits/sec
[116] 4.0- 5.0 sec 5.77 MBytes 48.4 Mbits/sec
[108] 4.0- 5.0 sec 39.9 MBytes 334 Mbits/sec
[SUM] 4.0- 5.0 sec 45.6 MBytes 383 Mbits/sec
[108] 5.0- 6.0 sec 35.8 MBytes 300 Mbits/sec
[116] 5.0- 6.0 sec 9.83 MBytes 82.4 Mbits/sec
[SUM] 5.0- 6.0 sec 45.6 MBytes 383 Mbits/sec
[116] 6.0- 7.0 sec 18.5 MBytes 155 Mbits/sec
[108] 6.0- 7.0 sec 19.2 MBytes 161 Mbits/sec
[ ID] Interval Transfer Bandwidth
[SUM] 6.0- 7.0 sec 37.7 MBytes 316 Mbits/sec
[116] 7.0- 8.0 sec 33.1 MBytes 278 Mbits/sec
[108] 7.0- 8.0 sec 14.4 MBytes 121 Mbits/sec
[SUM] 7.0- 8.0 sec 47.5 MBytes 399 Mbits/sec
[116] 8.0- 9.0 sec 23.4 MBytes 197 Mbits/sec
[108] 8.0- 9.0 sec 23.3 MBytes 195 Mbits/sec
[SUM] 8.0- 9.0 sec 46.7 MBytes 392 Mbits/sec
[116] 9.0-10.0 sec 19.2 MBytes 161 Mbits/sec
[108] 9.0-10.0 sec 25.4 MBytes 213 Mbits/sec
[SUM] 9.0-10.0 sec 44.6 MBytes 374 Mbits/sec
[116] 10.0-11.0 sec 14.5 MBytes 121 Mbits/sec
[108] 10.0-11.0 sec 33.9 MBytes 285 Mbits/sec
[SUM] 10.0-11.0 sec 48.4 MBytes 406 Mbits/sec
[116] 11.0-12.0 sec 27.4 MBytes 230 Mbits/sec
[108] 11.0-12.0 sec 20.5 MBytes 172 Mbits/sec
[SUM] 11.0-12.0 sec 47.9 MBytes 402 Mbits/sec
[116] 12.0-13.0 sec 13.1 MBytes 110 Mbits/sec
[108] 12.0-13.0 sec 32.7 MBytes 274 Mbits/sec
[SUM] 12.0-13.0 sec 45.8 MBytes 384 Mbits/sec
[108] 13.0-14.0 sec 16.4 MBytes 137 Mbits/sec
[ ID] Interval Transfer Bandwidth
[116] 13.0-14.0 sec 31.8 MBytes 266 Mbits/sec
[SUM] 13.0-14.0 sec 48.1 MBytes 404 Mbits/sec
[116] 14.0-15.0 sec 22.3 MBytes 187 Mbits/sec
[108] 14.0-15.0 sec 24.7 MBytes 207 Mbits/sec
[SUM] 14.0-15.0 sec 47.0 MBytes 394 Mbits/sec
[116] 15.0-16.0 sec 37.0 MBytes 310 Mbits/sec
[108] 15.0-16.0 sec 8.01 MBytes 67.2 Mbits/sec
[SUM] 15.0-16.0 sec 45.0 MBytes 377 Mbits/sec
[116] 16.0-17.0 sec 23.5 MBytes 197 Mbits/sec
[108] 16.0-17.0 sec 23.8 MBytes 200 Mbits/sec
[SUM] 16.0-17.0 sec 47.3 MBytes 397 Mbits/sec
[116] 17.0-18.0 sec 36.6 MBytes 307 Mbits/sec
[108] 17.0-18.0 sec 10.7 MBytes 90.0 Mbits/sec
[SUM] 17.0-18.0 sec 47.4 MBytes 397 Mbits/sec
[116] 18.0-19.0 sec 16.7 MBytes 140 Mbits/sec
[108] 18.0-19.0 sec 29.5 MBytes 248 Mbits/sec
[SUM] 18.0-19.0 sec 46.2 MBytes 388 Mbits/sec
[116] 19.0-20.0 sec 21.6 MBytes 181 Mbits/sec
[108] 19.0-20.0 sec 23.6 MBytes 198 Mbits/sec
[SUM] 19.0-20.0 sec 45.2 MBytes 380 Mbits/sec
[ ID] Interval Transfer Bandwidth
[108] 0.0-20.1 sec 478 MBytes 200 Mbits/sec
[116] 0.0-20.2 sec 430 MBytes 179 Mbits/sec
[SUM] 0.0-20.2 sec 908 MBytes 378 Mbits/sec
Done.


this is slightly better with a average of 190mbit per thread. two threads = 380mbit/s total.

hmm.

it is not possible to hard set either machine to 1gb full duplex using their NIC settings..


now
so since this is the max with directly connected laptops, i will now try through a brand new gigabit cisco 2960 switch using straight through cables.
i can verify that each laptop has negotiated full duplex by checking the port duplex and speed auto negotiation settings:

GigabitEthernet0/1 is up, line protocol is up (connected)
Hardware is Gigabit Ethernet, address is 0024.9859.0c81 (bia 0024.9859.0c81)
MTU 1500 bytes, BW 1000000 Kbit, DLY 10 usec,
reliability 255/255, txload 1/255, rxload 1/255
Encapsulation ARPA, loopback not set
Keepalive set (10 sec)
Full-duplex, 1000Mb/s, media type is 10/100/1000BaseTX
input flow-control is off, output flow-control is unsupported
ARP type: ARPA, ARP Timeout 04:00:00
Last input never, output 00:00:01, output hang never
Last clearing of "show interface" counters never
Input queue: 0/75/0/0 (size/max/drops/flushes); Total output drops: 0
Queueing strategy: fifo
Output queue: 0/40 (size/max)
5 minute input rate 0 bits/sec, 0 packets/sec
5 minute output rate 0 bits/sec, 0 packets/sec
60 packets input, 8305 bytes, 0 no buffer
Received 60 broadcasts (0 multicasts)
0 runts, 0 giants, 0 throttles
0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored
0 watchdog, 17 multicast, 0 pause input
0 input packets with dribble condition detected
36 packets output, 5097 bytes, 0 underruns
0 output errors, 0 collisions, 1 interface resets
0 babbles, 0 late collision, 0 deferred
0 lost carrier, 0 no carrier, 0 PAUSE output
0 output buffer failures, 0 output buffers swapped out
Switch#
Switch#
Switch#sh int gi0/2
GigabitEthernet0/2 is up, line protocol is up (connected)
Hardware is Gigabit Ethernet, address is 0024.9859.0c82 (bia 0024.9859.0c82)
MTU 1500 bytes, BW 1000000 Kbit, DLY 10 usec,
reliability 255/255, txload 1/255, rxload 1/255
Encapsulation ARPA, loopback not set
Keepalive set (10 sec)
Full-duplex, 1000Mb/s, media type is 10/100/1000BaseTX
input flow-control is off, output flow-control is unsupported
ARP type: ARPA, ARP Timeout 04:00:00
Last input never, output 00:00:01, output hang never
Last clearing of "show interface" counters never
Input queue: 0/75/0/0 (size/max/drops/flushes); Total output drops: 0
Queueing strategy: fifo
Output queue: 0/40 (size/max)
5 minute input rate 1000 bits/sec, 1 packets/sec
5 minute output rate 0 bits/sec, 0 packets/sec
135 packets input, 21530 bytes, 0 no buffer
Received 135 broadcasts (0 multicasts)
0 runts, 0 giants, 0 throttles
0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored
0 watchdog, 78 multicast, 0 pause input
0 input packets with dribble condition detected
33 packets output, 3184 bytes, 0 underruns
0 output errors, 0 collisions, 1 interface resets
0 babbles, 0 late collision, 0 deferred
0 lost carrier, 0 no carrier, 0 PAUSE output
0 output buffer failures, 0 output buffers swapped out
Switch#


now for the tests: (same as before, 2 threads, tcp window 2500kbyte)
test 1

bin/iperf.exe -c 192.168.192.1 -P 2 -i 1 -p 5001 -w 2500K -f m -t 20
------------------------------------------------------------
Client connecting to 192.168.192.1, TCP port 5001
TCP window size: 2.44 MByte
------------------------------------------------------------
[116] local 192.168.192.2 port 4292 connected with 192.168.192.1 port 5001
[108] local 192.168.192.2 port 4291 connected with 192.168.192.1 port 5001
[ ID] Interval Transfer Bandwidth
[108] 0.0- 1.0 sec 40.3 MBytes 338 Mbits/sec
[116] 0.0- 1.0 sec 11.0 MBytes 92.5 Mbits/sec
[SUM] 0.0- 1.0 sec 51.3 MBytes 431 Mbits/sec
[116] 1.0- 2.0 sec 28.5 MBytes 239 Mbits/sec
[108] 1.0- 2.0 sec 20.4 MBytes 171 Mbits/sec
[SUM] 1.0- 2.0 sec 48.9 MBytes 410 Mbits/sec
[108] 2.0- 3.0 sec 37.8 MBytes 317 Mbits/sec
[116] 2.0- 3.0 sec 11.0 MBytes 92.2 Mbits/sec
[SUM] 2.0- 3.0 sec 48.8 MBytes 409 Mbits/sec
[116] 3.0- 4.0 sec 20.0 MBytes 167 Mbits/sec
[108] 3.0- 4.0 sec 29.0 MBytes 244 Mbits/sec
[SUM] 3.0- 4.0 sec 49.0 MBytes 411 Mbits/sec
[116] 4.0- 5.0 sec 33.0 MBytes 277 Mbits/sec
[108] 4.0- 5.0 sec 15.3 MBytes 128 Mbits/sec
[SUM] 4.0- 5.0 sec 48.2 MBytes 405 Mbits/sec
[116] 5.0- 6.0 sec 33.2 MBytes 279 Mbits/sec
[108] 5.0- 6.0 sec 14.4 MBytes 121 Mbits/sec
[SUM] 5.0- 6.0 sec 47.7 MBytes 400 Mbits/sec
[108] 6.0- 7.0 sec 22.6 MBytes 190 Mbits/sec
[116] 6.0- 7.0 sec 16.6 MBytes 139 Mbits/sec
[ ID] Interval Transfer Bandwidth
[SUM] 6.0- 7.0 sec 39.2 MBytes 329 Mbits/sec
[116] 7.0- 8.0 sec 15.1 MBytes 126 Mbits/sec
[108] 7.0- 8.0 sec 24.1 MBytes 203 Mbits/sec
[SUM] 7.0- 8.0 sec 39.2 MBytes 329 Mbits/sec
[116] 8.0- 9.0 sec 18.2 MBytes 153 Mbits/sec
[108] 8.0- 9.0 sec 22.5 MBytes 189 Mbits/sec
[SUM] 8.0- 9.0 sec 40.7 MBytes 341 Mbits/sec
[108] 9.0-10.0 sec 4.66 MBytes 39.1 Mbits/sec
[116] 9.0-10.0 sec 35.4 MBytes 297 Mbits/sec
[SUM] 9.0-10.0 sec 40.0 MBytes 336 Mbits/sec
[116] 10.0-11.0 sec 32.9 MBytes 276 Mbits/sec
[108] 10.0-11.0 sec 16.6 MBytes 139 Mbits/sec
[SUM] 10.0-11.0 sec 49.5 MBytes 415 Mbits/sec
[108] 11.0-12.0 sec 25.1 MBytes 211 Mbits/sec
[116] 11.0-12.0 sec 21.3 MBytes 179 Mbits/sec
[SUM] 11.0-12.0 sec 46.4 MBytes 390 Mbits/sec
[108] 12.0-13.0 sec 19.8 MBytes 166 Mbits/sec
[116] 12.0-13.0 sec 28.4 MBytes 238 Mbits/sec
[SUM] 12.0-13.0 sec 48.2 MBytes 404 Mbits/sec
[108] 13.0-14.0 sec 17.0 MBytes 143 Mbits/sec
[ ID] Interval Transfer Bandwidth
[116] 13.0-14.0 sec 32.1 MBytes 269 Mbits/sec
[SUM] 13.0-14.0 sec 49.1 MBytes 412 Mbits/sec
[116] 14.0-15.0 sec 15.0 MBytes 126 Mbits/sec
[108] 14.0-15.0 sec 32.3 MBytes 271 Mbits/sec
[SUM] 14.0-15.0 sec 47.3 MBytes 397 Mbits/sec
[116] 15.0-16.0 sec 32.3 MBytes 271 Mbits/sec
[108] 15.0-16.0 sec 16.3 MBytes 137 Mbits/sec
[SUM] 15.0-16.0 sec 48.6 MBytes 408 Mbits/sec
[108] 16.0-17.0 sec 28.1 MBytes 236 Mbits/sec
[116] 16.0-17.0 sec 20.4 MBytes 171 Mbits/sec
[SUM] 16.0-17.0 sec 48.6 MBytes 407 Mbits/sec
[116] 17.0-18.0 sec 29.9 MBytes 250 Mbits/sec
[108] 17.0-18.0 sec 18.7 MBytes 157 Mbits/sec
[SUM] 17.0-18.0 sec 48.5 MBytes 407 Mbits/sec
[108] 18.0-19.0 sec 31.5 MBytes 264 Mbits/sec
[116] 18.0-19.0 sec 17.5 MBytes 147 Mbits/sec
[SUM] 18.0-19.0 sec 49.0 MBytes 411 Mbits/sec
[108] 19.0-20.0 sec 18.6 MBytes 156 Mbits/sec
[108] 0.0-20.0 sec 455 MBytes 190 Mbits/sec
[116] 19.0-20.0 sec 30.0 MBytes 251 Mbits/sec
[ ID] Interval Transfer Bandwidth
[SUM] 19.0-20.0 sec 48.6 MBytes 408 Mbits/sec
[116] 0.0-20.2 sec 482 MBytes 200 Mbits/sec
[SUM] 0.0-20.2 sec 937 MBytes 388 Mbits/sec
Done.


hmm, faster than before'

test 2 = 2 threads, 1500kbyte tcp window

bin/iperf.exe -c 192.168.192.1 -P 2 -i 1 -p 5001 -w 1500K -f m -t 20
------------------------------------------------------------
Client connecting to 192.168.192.1, TCP port 5001
TCP window size: 1.46 MByte
------------------------------------------------------------
[116] local 192.168.192.2 port 4306 connected with 192.168.192.1 port 5001
[108] local 192.168.192.2 port 4305 connected with 192.168.192.1 port 5001
[ ID] Interval Transfer Bandwidth
[108] 0.0- 1.0 sec 41.4 MBytes 347 Mbits/sec
[116] 0.0- 1.0 sec 8.46 MBytes 71.0 Mbits/sec
[SUM] 0.0- 1.0 sec 49.9 MBytes 418 Mbits/sec
[116] 1.0- 2.0 sec 20.0 MBytes 168 Mbits/sec
[108] 1.0- 2.0 sec 27.4 MBytes 230 Mbits/sec
[SUM] 1.0- 2.0 sec 47.4 MBytes 398 Mbits/sec
[116] 2.0- 3.0 sec 14.1 MBytes 119 Mbits/sec
[108] 2.0- 3.0 sec 31.8 MBytes 266 Mbits/sec
[SUM] 2.0- 3.0 sec 45.9 MBytes 385 Mbits/sec
[108] 3.0- 4.0 sec 25.2 MBytes 211 Mbits/sec
[116] 3.0- 4.0 sec 14.4 MBytes 121 Mbits/sec
[SUM] 3.0- 4.0 sec 39.6 MBytes 332 Mbits/sec
[116] 4.0- 5.0 sec 29.3 MBytes 246 Mbits/sec
[108] 4.0- 5.0 sec 18.9 MBytes 158 Mbits/sec
[SUM] 4.0- 5.0 sec 48.2 MBytes 404 Mbits/sec
[116] 5.0- 6.0 sec 34.6 MBytes 290 Mbits/sec
[108] 5.0- 6.0 sec 15.2 MBytes 127 Mbits/sec
[SUM] 5.0- 6.0 sec 49.7 MBytes 417 Mbits/sec
[108] 6.0- 7.0 sec 21.6 MBytes 181 Mbits/sec
[116] 6.0- 7.0 sec 27.8 MBytes 233 Mbits/sec
[ ID] Interval Transfer Bandwidth
[SUM] 6.0- 7.0 sec 49.5 MBytes 415 Mbits/sec
[108] 7.0- 8.0 sec 27.8 MBytes 233 Mbits/sec
[116] 7.0- 8.0 sec 21.7 MBytes 182 Mbits/sec
[SUM] 7.0- 8.0 sec 49.5 MBytes 415 Mbits/sec
[116] 8.0- 9.0 sec 18.4 MBytes 155 Mbits/sec
[108] 8.0- 9.0 sec 30.7 MBytes 258 Mbits/sec
[SUM] 8.0- 9.0 sec 49.1 MBytes 412 Mbits/sec
[116] 9.0-10.0 sec 45.2 MBytes 379 Mbits/sec
[108] 9.0-10.0 sec 3.07 MBytes 25.8 Mbits/sec
[SUM] 9.0-10.0 sec 48.3 MBytes 405 Mbits/sec
[116] 10.0-11.0 sec 27.1 MBytes 227 Mbits/sec
[108] 10.0-11.0 sec 21.6 MBytes 181 Mbits/sec
[SUM] 10.0-11.0 sec 48.7 MBytes 408 Mbits/sec
[108] 11.0-12.0 sec 20.6 MBytes 173 Mbits/sec
[116] 11.0-12.0 sec 29.2 MBytes 245 Mbits/sec
[SUM] 11.0-12.0 sec 49.8 MBytes 417 Mbits/sec
[116] 12.0-13.0 sec 20.3 MBytes 170 Mbits/sec
[108] 12.0-13.0 sec 28.8 MBytes 241 Mbits/sec
[SUM] 12.0-13.0 sec 49.1 MBytes 412 Mbits/sec
[116] 13.0-14.0 sec 32.3 MBytes 271 Mbits/sec
[ ID] Interval Transfer Bandwidth
[108] 13.0-14.0 sec 10.0 MBytes 83.9 Mbits/sec
[SUM] 13.0-14.0 sec 42.3 MBytes 355 Mbits/sec
[116] 14.0-15.0 sec 4.85 MBytes 40.7 Mbits/sec
[108] 14.0-15.0 sec 38.9 MBytes 326 Mbits/sec
[SUM] 14.0-15.0 sec 43.8 MBytes 367 Mbits/sec
[116] 15.0-16.0 sec 18.6 MBytes 156 Mbits/sec
[108] 15.0-16.0 sec 30.8 MBytes 258 Mbits/sec
[SUM] 15.0-16.0 sec 49.4 MBytes 415 Mbits/sec
[116] 16.0-17.0 sec 34.5 MBytes 289 Mbits/sec
[108] 16.0-17.0 sec 14.9 MBytes 125 Mbits/sec
[SUM] 16.0-17.0 sec 49.4 MBytes 414 Mbits/sec
[116] 17.0-18.0 sec 31.8 MBytes 267 Mbits/sec
[108] 17.0-18.0 sec 17.7 MBytes 148 Mbits/sec
[SUM] 17.0-18.0 sec 49.5 MBytes 415 Mbits/sec
[116] 18.0-19.0 sec 32.8 MBytes 275 Mbits/sec
[108] 18.0-19.0 sec 14.8 MBytes 124 Mbits/sec
[SUM] 18.0-19.0 sec 47.6 MBytes 399 Mbits/sec
[116] 19.0-20.0 sec 31.4 MBytes 264 Mbits/sec
[116] 0.0-20.0 sec 497 MBytes 208 Mbits/sec
[108] 19.0-20.0 sec 8.03 MBytes 67.4 Mbits/sec
[ ID] Interval Transfer Bandwidth
[SUM] 19.0-20.0 sec 39.5 MBytes 331 Mbits/sec
[108] 0.0-20.1 sec 449 MBytes 188 Mbits/sec
[SUM] 0.0-20.1 sec 946 MBytes 395 Mbits/sec
Done.


through the switches actually is faster than going direct with a crossover at this point.

now the final test is to connect the devices up this way:
laptop > switch 1 > switch 2 > laptop 2

here are the results:
attempt 1

bin/iperf.exe -c 192.168.192.1 -P 2 -i 1 -p 5001 -w 1500K -f m -t 20
------------------------------------------------------------
Client connecting to 192.168.192.1, TCP port 5001
TCP window size: 1.46 MByte
------------------------------------------------------------
[116] local 192.168.192.2 port 4320 connected with 192.168.192.1 port 5001
[108] local 192.168.192.2 port 4319 connected with 192.168.192.1 port 5001
[ ID] Interval Transfer Bandwidth
[108] 0.0- 1.0 sec 33.6 MBytes 282 Mbits/sec
[116] 0.0- 1.0 sec 16.2 MBytes 136 Mbits/sec
[SUM] 0.0- 1.0 sec 49.8 MBytes 418 Mbits/sec
[116] 1.0- 2.0 sec 23.2 MBytes 195 Mbits/sec
[108] 1.0- 2.0 sec 25.8 MBytes 216 Mbits/sec
[SUM] 1.0- 2.0 sec 49.0 MBytes 411 Mbits/sec
[108] 2.0- 3.0 sec 19.3 MBytes 162 Mbits/sec
[116] 2.0- 3.0 sec 23.4 MBytes 196 Mbits/sec
[SUM] 2.0- 3.0 sec 42.6 MBytes 358 Mbits/sec
[116] 3.0- 4.0 sec 18.3 MBytes 154 Mbits/sec
[108] 3.0- 4.0 sec 19.8 MBytes 166 Mbits/sec
[SUM] 3.0- 4.0 sec 38.2 MBytes 320 Mbits/sec
[116] 4.0- 5.0 sec 17.6 MBytes 148 Mbits/sec
[108] 4.0- 5.0 sec 19.3 MBytes 162 Mbits/sec
[SUM] 4.0- 5.0 sec 36.9 MBytes 310 Mbits/sec
[108] 5.0- 6.0 sec 24.8 MBytes 208 Mbits/sec
[116] 5.0- 6.0 sec 14.1 MBytes 118 Mbits/sec
[SUM] 5.0- 6.0 sec 38.9 MBytes 327 Mbits/sec
[116] 6.0- 7.0 sec 12.5 MBytes 105 Mbits/sec
[108] 6.0- 7.0 sec 33.3 MBytes 279 Mbits/sec
[ ID] Interval Transfer Bandwidth
[SUM] 6.0- 7.0 sec 45.7 MBytes 383 Mbits/sec
[116] 7.0- 8.0 sec 12.5 MBytes 105 Mbits/sec
[108] 7.0- 8.0 sec 33.6 MBytes 282 Mbits/sec
[SUM] 7.0- 8.0 sec 46.2 MBytes 387 Mbits/sec
[108] 8.0- 9.0 sec 17.7 MBytes 148 Mbits/sec
[116] 8.0- 9.0 sec 30.5 MBytes 256 Mbits/sec
[SUM] 8.0- 9.0 sec 48.2 MBytes 405 Mbits/sec
[116] 9.0-10.0 sec 21.5 MBytes 180 Mbits/sec
[108] 9.0-10.0 sec 10.1 MBytes 84.8 Mbits/sec
[SUM] 9.0-10.0 sec 31.6 MBytes 265 Mbits/sec
[116] 10.0-11.0 sec 19.4 MBytes 162 Mbits/sec
[108] 10.0-11.0 sec 17.6 MBytes 147 Mbits/sec
[SUM] 10.0-11.0 sec 36.9 MBytes 310 Mbits/sec
[116] 11.0-12.0 sec 33.4 MBytes 280 Mbits/sec
[108] 11.0-12.0 sec 10.7 MBytes 89.7 Mbits/sec
[SUM] 11.0-12.0 sec 44.1 MBytes 370 Mbits/sec
[116] 12.0-13.0 sec 40.4 MBytes 339 Mbits/sec
[108] 12.0-13.0 sec 7.88 MBytes 66.1 Mbits/sec
[SUM] 12.0-13.0 sec 48.3 MBytes 405 Mbits/sec
[108] 13.0-14.0 sec 21.2 MBytes 178 Mbits/sec
[ ID] Interval Transfer Bandwidth
[116] 13.0-14.0 sec 25.6 MBytes 215 Mbits/sec
[SUM] 13.0-14.0 sec 46.8 MBytes 392 Mbits/sec
[116] 14.0-15.0 sec 7.82 MBytes 65.6 Mbits/sec
[108] 14.0-15.0 sec 40.6 MBytes 341 Mbits/sec
[SUM] 14.0-15.0 sec 48.4 MBytes 406 Mbits/sec
[116] 15.0-16.0 sec 32.2 MBytes 270 Mbits/sec
[108] 15.0-16.0 sec 14.4 MBytes 121 Mbits/sec
[SUM] 15.0-16.0 sec 46.6 MBytes 391 Mbits/sec
[108] 16.0-17.0 sec 23.6 MBytes 198 Mbits/sec
[116] 16.0-17.0 sec 23.8 MBytes 200 Mbits/sec
[SUM] 16.0-17.0 sec 47.4 MBytes 398 Mbits/sec
[108] 17.0-18.0 sec 36.1 MBytes 303 Mbits/sec
[116] 17.0-18.0 sec 6.95 MBytes 58.3 Mbits/sec
[SUM] 17.0-18.0 sec 43.1 MBytes 361 Mbits/sec
[108] 18.0-19.0 sec 30.2 MBytes 254 Mbits/sec
[116] 18.0-19.0 sec 2.19 MBytes 18.4 Mbits/sec
[SUM] 18.0-19.0 sec 32.4 MBytes 272 Mbits/sec
[116] 19.0-20.0 sec 19.6 MBytes 165 Mbits/sec
[116] 0.0-20.0 sec 401 MBytes 168 Mbits/sec
[108] 19.0-20.0 sec 26.8 MBytes 225 Mbits/sec
[ ID] Interval Transfer Bandwidth
[SUM] 19.0-20.0 sec 46.4 MBytes 389 Mbits/sec
[108] 0.0-20.2 sec 466 MBytes 194 Mbits/sec
[SUM] 0.0-20.2 sec 868 MBytes 361 Mbits/sec
Done.


attempt 2

bin/iperf.exe -c 192.168.192.1 -P 2 -i 1 -p 5001 -w 1500K -f m -t 20
------------------------------------------------------------
Client connecting to 192.168.192.1, TCP port 5001
TCP window size: 1.46 MByte
------------------------------------------------------------
[116] local 192.168.192.2 port 37551 connected with 192.168.192.1 port 5001
[108] local 192.168.192.2 port 37550 connected with 192.168.192.1 port 5001
[ ID] Interval Transfer Bandwidth
[116] 0.0- 1.0 sec 34.2 MBytes 287 Mbits/sec
[108] 0.0- 1.0 sec 14.4 MBytes 121 Mbits/sec
[SUM] 0.0- 1.0 sec 48.5 MBytes 407 Mbits/sec
[116] 1.0- 2.0 sec 41.2 MBytes 346 Mbits/sec
[108] 1.0- 2.0 sec 6.30 MBytes 52.9 Mbits/sec
[SUM] 1.0- 2.0 sec 47.5 MBytes 399 Mbits/sec
[116] 2.0- 3.0 sec 44.2 MBytes 371 Mbits/sec
[108] 2.0- 3.0 sec 2.13 MBytes 17.9 Mbits/sec
[SUM] 2.0- 3.0 sec 46.3 MBytes 389 Mbits/sec
[108] 3.0- 4.0 sec 15.7 MBytes 131 Mbits/sec
[116] 3.0- 4.0 sec 28.9 MBytes 243 Mbits/sec
[SUM] 3.0- 4.0 sec 44.6 MBytes 374 Mbits/sec
[108] 4.0- 5.0 sec 26.7 MBytes 224 Mbits/sec
[116] 4.0- 5.0 sec 19.3 MBytes 162 Mbits/sec
[SUM] 4.0- 5.0 sec 46.0 MBytes 386 Mbits/sec
[108] 5.0- 6.0 sec 27.0 MBytes 226 Mbits/sec
[116] 5.0- 6.0 sec 17.9 MBytes 150 Mbits/sec
[SUM] 5.0- 6.0 sec 44.8 MBytes 376 Mbits/sec
[108] 6.0- 7.0 sec 23.1 MBytes 194 Mbits/sec
[116] 6.0- 7.0 sec 23.7 MBytes 198 Mbits/sec
[ ID] Interval Transfer Bandwidth
[SUM] 6.0- 7.0 sec 46.8 MBytes 392 Mbits/sec
[108] 7.0- 8.0 sec 33.9 MBytes 285 Mbits/sec
[116] 7.0- 8.0 sec 0.50 MBytes 4.19 Mbits/sec
[SUM] 7.0- 8.0 sec 34.4 MBytes 289 Mbits/sec
[108] 8.0- 9.0 sec 30.3 MBytes 254 Mbits/sec
[116] 8.0- 9.0 sec 17.4 MBytes 146 Mbits/sec
[SUM] 8.0- 9.0 sec 47.7 MBytes 400 Mbits/sec
[116] 9.0-10.0 sec 17.2 MBytes 144 Mbits/sec
[108] 9.0-10.0 sec 26.6 MBytes 223 Mbits/sec
[SUM] 9.0-10.0 sec 43.8 MBytes 367 Mbits/sec
[108] 10.0-11.0 sec 18.2 MBytes 153 Mbits/sec
[116] 10.0-11.0 sec 12.1 MBytes 102 Mbits/sec
[SUM] 10.0-11.0 sec 30.3 MBytes 254 Mbits/sec
[108] 11.0-12.0 sec 28.6 MBytes 240 Mbits/sec
[116] 11.0-12.0 sec 16.1 MBytes 135 Mbits/sec
[SUM] 11.0-12.0 sec 44.7 MBytes 375 Mbits/sec
[108] 12.0-13.0 sec 44.3 MBytes 371 Mbits/sec
[116] 12.0-13.0 sec 4.52 MBytes 37.9 Mbits/sec
[SUM] 12.0-13.0 sec 48.8 MBytes 409 Mbits/sec
[108] 13.0-14.0 sec 23.7 MBytes 199 Mbits/sec
[ ID] Interval Transfer Bandwidth
[116] 13.0-14.0 sec 24.1 MBytes 203 Mbits/sec
[SUM] 13.0-14.0 sec 47.9 MBytes 401 Mbits/sec
[116] 14.0-15.0 sec 28.2 MBytes 237 Mbits/sec
[108] 14.0-15.0 sec 17.8 MBytes 150 Mbits/sec
[SUM] 14.0-15.0 sec 46.1 MBytes 386 Mbits/sec
[116] 15.0-16.0 sec 29.2 MBytes 245 Mbits/sec
[108] 15.0-16.0 sec 19.3 MBytes 162 Mbits/sec
[SUM] 15.0-16.0 sec 48.5 MBytes 407 Mbits/sec
[108] 16.0-17.0 sec 24.5 MBytes 206 Mbits/sec
[116] 16.0-17.0 sec 22.7 MBytes 191 Mbits/sec
[SUM] 16.0-17.0 sec 47.2 MBytes 396 Mbits/sec
[108] 17.0-18.0 sec 30.6 MBytes 257 Mbits/sec
[116] 17.0-18.0 sec 17.9 MBytes 150 Mbits/sec
[SUM] 17.0-18.0 sec 48.5 MBytes 407 Mbits/sec
[116] 18.0-19.0 sec 11.0 MBytes 92.4 Mbits/sec
[108] 18.0-19.0 sec 36.9 MBytes 309 Mbits/sec
[SUM] 18.0-19.0 sec 47.9 MBytes 402 Mbits/sec
[108] 19.0-20.0 sec 13.8 MBytes 116 Mbits/sec
[108] 0.0-20.0 sec 464 MBytes 194 Mbits/sec
[116] 19.0-20.0 sec 20.1 MBytes 168 Mbits/sec
[ ID] Interval Transfer Bandwidth
[SUM] 19.0-20.0 sec 33.9 MBytes 284 Mbits/sec
[116] 0.0-20.3 sec 430 MBytes 178 Mbits/sec
[SUM] 0.0-20.3 sec 894 MBytes 370 Mbits/sec
Done.


no where near 900mbit (or gigabit speeds)
even when i use multiple threads (i tried 5) the sum is the same, but each individual thread is reduced

no i will simulate a duplex mis match by configuring one side as 1gb full duplex. the mismatch is between the two switches.
ie laptop > switch 1 - duplexmismatch - switch 2 > laptop

here is the interface stats for switch 1

GigabitEthernet0/2 is up, line protocol is up (connected)
Hardware is Gigabit Ethernet, address is 0023.abf6.811a (bia 0023.abf6.811a)
MTU 1500 bytes, BW 1000000 Kbit, DLY 10 usec,
reliability 255/255, txload 1/255, rxload 10/255
Encapsulation ARPA, loopback not set
Keepalive set (10 sec)
Full-duplex, 1000Mb/s, media type is 10/100/1000BaseTX
input flow-control is off, output flow-control is unsupported
ARP type: ARPA, ARP Timeout 04:00:00
Last input 00:00:01, output 00:00:00, output hang never
Last clearing of "show interface" counters never
Input queue: 0/75/0/0 (size/max/drops/flushes); Total output drops: 0
Queueing strategy: fifo
Output queue: 0/40 (size/max)
5 minute input rate 39845000 bits/sec, 3805 packets/sec
5 minute output rate 934000 bits/sec, 1827 packets/sec
2217512 packets input, 2893974993 bytes, 0 no buffer
Received 76 broadcasts (0 multicasts)
0 runts, 0 giants, 0 throttles
0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored
0 watchdog, 62 multicast, 0 pause input
0 input packets with dribble condition detected
1068995 packets output, 68866238 bytes, 0 underruns
0 output errors, 0 collisions, 1 interface resets
0 babbles, 0 late collision, 0 deferred
0 lost carrier, 0 no carrier, 0 PAUSE output
0 output buffer failures, 0 output buffers swapped out


here is the stats for switch two
EDIT: after trying to get them to fail negotiation when i couldnt i tried to manually set one side to half duplex. when i do i get this message (which confirms what i mentioned earlier:

Switch(config-if)#duplex half
Gigabit port is restricted to full duplex


so there you have it.
i had a chat with a colleague and he mentioned the limitations of the pci bus. i looked up this using wikipedia (http://en.wikipedia.org/wiki/PCI_Local_Bus). on that website it states the maximum is 133mega bytes per second which is slightly more than a gig. so in theory i should have been able to get near gig speeds. although i dont know what else is sharing this bus. possibly bluetooth or wireless.

i personally would expect or think two PCs with on board nics to be a little faster though.

[g0dM@n]

Good round-up ppe1700.

Now I'm not entirely surprised to the cisco switch giving you better results, but it is a little puzzling. I guess it helps for better negotiation and packet management...

You know, I guess you'd really have to disable a lot of other things to get true gigabit. Perhaps being on the internet at all can affect it, since other programs are on there, like anti-virus, windows media player, and whatever else in the background. I mean that not only means internet access through the same ethernet port, but it also means the hard drive is performing instructions for those programs as well.

What if you were to use STP cables? :) I don't have... damn I just put in an order with monoprice yesterday. I should have also bought a STP for testing.
*Edit* Actually they only have STP for the 1000ft bundles. I wouldn't wanna buy 1000ft just for testing.

[ppe1700]

yea i dont believe the cisco switches were giving better results, i just think at that time something may have happened to appear to give me better results. you see its basically just averages. i would summarise by saying that going by those results, using one or two switches should not degrade the performance in any way.
i know for a fact the dell latitude laptop i use is complete pants. i used to let vista adjust the vram but it ended up getting to some stupidly high 4gb or so value and every day at a certain time the machine would practically lock up whilst it tried to search the HDD for the ram data it needed. i have 2gb ram, and now i have locked the vram to about 400mb. yes i always get "low virtual memory" warnings but the laptop is actually faster.
what i mean is that may be with better more expensive equipment (ie not "bought off the shelf") may be someone else can get better throughput using two gigabit PCs

the cisco standard says you should only really have a network that is 7 switches wide. this means that data should only really travel 7 switches. but, this is only specified for spanning-tree convergence. if your network is more than 7 wide the documentation states you may need to look into adjusting STP timers to prevent flaping root ports.

another written doc states you should size a switch block according to data flows and traffic patterns... there basically isnt a limit on how many switches you can use, unless it degrades your performance. you would have to have many switches for that to happen. and when this did happen, it probably wouldnt be because the data has to travel a distance, it would be because of all the other devices plugged into the switches broadcasting (because almost every device does some sort of broadcasts, moreso windows machines) and these broadcasts propagating throughout every UP port throught the network.

are you going to be able to run some jperf tests?

[g0dM@n]

I did run some jperf tests, post #55 (http://www.ocforums.com/showpost.php?p=6061510&postcount=55).

If you want me to use a specific command, let me know. I have the jperf server on my HTPC 192.168.70.141 if you want to write up a command that you'd like to see. I think it's running on port 5001 (default). I was jperfing from my main rig (the one in my sig)... my rig goes through a 25ft cat5e to an 8-port giga switch, then 100ft through a cat6 to a 5-port switch, then through a ~2ft patch cable I made with solid cat5e to the HTPC gigabit LAN port.

[gangaskan]

Hmmm... I guess the write speed on the 1.5tb is that much slower? Heck I'm sure it's way too complicated to figure it out. Interesting though. :) Btw, lol... look at the stopwatch I used here (http://www.online-stopwatch.com/).


more than one drive is going to traumph all over a single disk :) speeds are pretty much on hardware.

more importantly, your raid controller and your disks will show you what your network is capable of! also, with those transfer speeds i dont think frame overhead is taken into consideration? do you know PPE? i dont recall if it does or not :P

[g0dM@n]

more than one drive is going to traumph all over a single disk :) speeds are pretty much on hardware.

more importantly, your raid controller and your disks will show you what your network is capable of! also, with those transfer speeds i dont think frame overhead is taken into consideration? do you know PPE? i dont recall if it does or not :P

Frame overhead? And do you mean PPoE? I really don't know much about it...

[gangaskan]

Frame overhead? And do you mean PPoE? I really don't know much about it...

lol, no i ment the frenchie PPE not PPoe :P


frame overhead meaning the extra tagging required in order to send out the packets looking at it now, packet overhead is more of what i was referring too :P

overhead meaning, where its being sent, the sender, what protocol, what TTL, and various things like vlan tagging and encapsulation. all that stuff gets packaged with the frame of data.

[g0dM@n]

lol, no i ment the frenchie PPE not PPoe :P


frame overhead meaning the extra tagging required in order to send out the packets looking at it now, packet overhead is more of what i was referring too :P

overhead meaning, where its being sent, the sender, what protocol, what TTL, and various things like vlan tagging and encapsulation. all that stuff gets packaged with the frame of data.

I'm guessing you mean that that might be the explanation as to why we're not even getting half of gigabit speeds... I guess it makes sense to say that gigabit is the maximum bandwidth including ALL packets and packet info... so if 50% of the bandwidth is part of network maintenance, you only have 50% left to max out.

If that's the case, then I'm sure there will eventually be better management (different protocol).

[gangaskan]

its hard to estimate the % of the packets are overhead only because you can have varying packet sizes, but its safe to say if you have a max payload in a frame (1500 bytes) than you can have 32-42 bytes (untagged vs tagged vlan)

your home will most likely be untagged so its safe to say 32 bits per packet is your overhead. roughly 3-5% if my math is correct!

[ppe1700]

lol, no i ment the frenchie PPE not PPoe :P


frenchie?? now im lost..

in 100mbit the overhead could be what, 10% so in gigabit, the overhead would not be more, just the same as 100mbit... not 50% though

i think basically we have hit the physical limitations of standard equipment with gigabit.
so in theory, if you spent more money on your hardware that was NOT network related, then you could DOUBLE your network performance :/ :screwy:

[petteyg359]

One thing we learn in CCNA is that if you have too many repeaters between you and the router, your signal can get degraded. ITs possible that since you have 2 switches, and then a third switch in the router that it is going through, the signal is so degraded it has to lower the speed to ensure decent transfer. For retail style routers and switches, I wouldn't be suprised if this is the case.

Thats my suggestion, anyway

There's no way a few switches are going to be "too many repeaters". g0dM@n, what router/switches are you using? I don't see it in previous posts. Have you run Atto and other drive benchmarks to see if your hard drive is the limiting factor?

I've got:


DSL ---> WRT350N ---> A ---> C
| \ \
B--\ W C
/| |\
//| /||
CC CC CC


W ---> C


A - D-Link Gb switch
B - SMC Gb switch
C - Computer
W - 802.11n

Every computer on the network gets full gigabit to every other computer, except over the wifi, where it obviously can't do Gb. Obviously that's when dealing with transfers of stuff that's already in memory, disk speed limitations apply, otherwise :)

[TollhouseFrank]

There's no way a few switches are going to be "too many repeaters". g0dM@n, what router/switches are you using? I don't see it in previous posts. Have you run Atto and other drive benchmarks to see if your hard drive is the limiting factor?

I've got:


DSL ---> WRT350N ---> A ---> C
| \ \
B--\ W C
/| |\
//| /||
CC CC CC


W ---> C


A - D-Link Gb switch
B - SMC Gb switch
C - Computer
W - 802.11n

Every computer on the network gets full gigabit to every other computer, except over the wifi, where it obviously can't do Gb. Obviously that's when dealing with transfers of stuff that's already in memory, disk speed limitations apply, otherwise :)

by the 5:4:3 rule, he's fine.

[g0dM@n]

There's no way a few switches are going to be "too many repeaters". g0dM@n, what router/switches are you using? I don't see it in previous posts. Have you run Atto and other drive benchmarks to see if your hard drive is the limiting factor?

I've got:


DSL ---> WRT350N ---> A ---> C
| \ \
B--\ W C
/| |\
//| /||
CC CC CC


W ---> C


A - D-Link Gb switch
B - SMC Gb switch
C - Computer
W - 802.11n

Every computer on the network gets full gigabit to every other computer, except over the wifi, where it obviously can't do Gb. Obviously that's when dealing with transfers of stuff that's already in memory, disk speed limitations apply, otherwise :)

My setup is:

Cable ---> DD-WRT Netgear (broadcast wireless N)
| \
GS1 P
/ | \
GS2 \ \
/ | C C
HTPC 360


GS1 = TRENDnet GREENnet 8-port Gigabit switch
GS2 = TRENDnet GREENnet 5-port Gigabit switch
C = Computers (gigabit)
HTPC = do I have to say? :)
360 = do I have to say? :)
P = USB Print server that goes to my laser printer

The wireless is already part of the router.


*Edit*
What the heck... I made it nice and neat in my txt box, but after the post it doesn't come out right! :(

[gangaskan]

frenchie?? now im lost..


just throwing you for a loop :P in the general area atleast hehe.

[petteyg359]

My setup is:

Cable ---> DD-WRT Netgear (broadcast wireless N)
| \
GS1 P
/ | \
GS2 \ \
/ | C C
HTPC 360


GS1 = TRENDnet GREENnet 8-port Gigabit switch
GS2 = TRENDnet GREENnet 5-port Gigabit switch
C = Computers (gigabit)
HTPC = do I have to say? :)
360 = do I have to say? :)
P = USB Print server that goes to my laser printer

The wireless is already part of the router.


*Edit*
What the heck... I made it nice and neat in my txt box, but after the post it doesn't come out right! :(

Got to count the spaces between characters, not how it looks visually :) You could use a greasemonkey script or something to make the font in <textarea> a fixed width font.

[g0dM@n]

Got to count the spaces between characters, not how it looks visually :) You could use a greasemonkey script or something to make the font in <textarea> a fixed width font.

I guess it's not worth the effort then. :) I dunno many scripts...

[ppe1700]

just throwing you for a loop :P in the general area atleast hehe.

im british not french!

anyways, a frenchie is a kiss of some kind :/ :eek: :screwy:

i have one more thing to add about switches;
when they switch, they SWITCH as in on/off... what i mean by this is that the speed through a switch should be "wire speed" therefore it shouldnt really matter how many you have, it should be the same as running a cable directly from one machine to the other.
now, when traffic is introduced then congestion occurs, but the switching mechanism in quality switches that "switch" in hardware (using chips instead of a cpu cycle(s)) should be as quick as having a cable directly connected.

[g0dM@n]

im british not french!

anyways, a frenchie is a kiss of some kind :/ :eek: :screwy:

i have one more thing to add about switches;
when they switch, they SWITCH as in on/off... what i mean by this is that the speed through a switch should be "wire speed" therefore it shouldnt really matter how many you have, it should be the same as running a cable directly from one machine to the other.
now, when traffic is introduced then congestion occurs, but the switching mechanism in quality switches that "switch" in hardware (using chips instead of a cpu cycle(s)) should be as quick as having a cable directly connected.

Right... until 2 things want to talk to the same machine

[pik4chu]

One thing we learn in CCNA is that if you have too many repeaters between you and the router, your signal can get degraded. ITs possible that since you have 2 switches, and then a third switch in the router that it is going through, the signal is so degraded it has to lower the speed to ensure decent transfer. For retail style routers and switches, I wouldn't be suprised if this is the case.

Thats my suggestion, anyway

Just wanted to add that what Cisco means by "too many" is not 2 lol. its more like 10.. Machines-switch-switch-servers is not an issue.

Built a network recently for a Hospital that involed ~720 network drops connected to Gigabit switches. Full gigabit bandwidth available to all users with no noticable slowdowns, I can confirm this because we had to do a great deal of testing for a number of QoS devices on the same wire and the workstations were pushing data as fast as their hard drives could move it ;) . It takes a lot of equipment to jam up a network like that.

I didnt read much of the rest of the thread but my vote is for the wirinbg between switches being the issue here :) Time to get out that cable tester!

[ppe1700]

Right... until 2 things want to talk to the same machine

even if more than 1 device wants to talk to the same machine. ALL devices still get a directl connection as if it were connected with a crossover cable. the difference is now there is traffic.
switches have buffers and depending on the switches used depends how the data is handled. generally the data is first in - first out. meaning whatever arrived first is sent out first. this is on a per packet basis so you could get many devices all talking at once over the same piece of wire.
imagine someone shuffling cards. like when they split the deck and lay them on the table with the corners bent up and overlapping - then they let go... first in first out with the two halfs split. if that makes sense.

[JimmyG]

Well unfortunately I couldn't find a router+gigabit in one...

The D-Link DIR=655 wireless router has a built in gig switch.

[g0dM@n]

The D-Link DIR=655 wireless router has a built in gig switch.

That comment was over 2 yrs ago, buddy. :) Thx tho!!

[Mike Wolf]

Check the type of CAT cabling. I'd recommend CAT 5e or Cat 6, since it has higher bandwidth. Also you shouldnt be seeing any issues with connecting two gig switches in the same network. Try other switches as well for troubleshooting.