• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Runing Multi FAH On Same Computer

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

DodgeViper

Member
Joined
Sep 21, 2001
Location
WILDCAT COUNTRY
What are the advantages and disadvantages of running say four separate FAH programs in four separate folders with each asigned 1 through 4 ID?
 
Advantages=stops your dialup reconnecting every couple of hours for a new WU, lets you wait until all are finished and have a mass upload/download once a day.

Disadvantages=could miss deadlines, more client overhead=marginally less PPW.

Road Warrior
 
Sometimes when the mass download is doneyou might not get credit immediately but normally the server catches you up. I've used three instances per machine during the week which allows me to ignore them pretty much except for monitoring temps etc. On the weekend I leave the connection on the fastest machine and run one instance per. It gets kind of hectic when you get really small wus. The opposite though is the possibility of big gromacs restarting for whatever reason and folding time being lost. Maximum effort without a broadband connection on every box is hard work; therefore the need for multy instances to lighten the load.
 
From what I've read, anything more then one instance per
CPU is wastefull.

I've not mathmatically proven that, but that's what I've read.
 
Lonely Raven said:
From what I've read, anything more then one instance per
CPU is wastefull.

I've not mathmatically proven that, but that's what I've read.
Actually that is NOT true... running 2 instances on one CPU or 3 on a dually 'should' actually boost your PPW a tiny bit. The reason why is, with only 1 instance running when that WU finishes, the machine sits idle communicating with the servers, sending receiving etc. With multiple instances running when one finishes and is sending/receiving the other instance(s) are still pumping away so you are truely working a WU 24/7.
 
The hand off time on all my work units is about 10-15 seconds.

I'm not sure I'd see that much of an improvement running
3 instances on my dualie...
 
I don't recommend multiple instances for anyone who has a setup that unerringly does everything it should when it should. That's a perfect setup and regretably unachieveable from our end. Stanfords servers have the control over when we get our next unit after download and how fast it will upload. Those who have been around awhile know that sometimes that time can be frustratingly long. It's up to each member to learn as much as he can about the operations both on our end and theirs, and the positives and negatives of his/her setup. We are operators of systems designed (I hope) to glean the most work for Stanford possible. Accordingly, we need to do those things necessary to achieve maximum productivity. Anything less quite frankly is embarrassing.
 
firing up multiple clients is a good idea when Stanford has a planned outage - wouldn't want to run out of beer after the stores close, would you?:beer:
 
Lonely Raven said:
The hand off time on all my work units is about 10-15 seconds.

I'm not sure I'd see that much of an improvement running
3 instances on my dualie...
Hmm, it takes my machines here at my office on cable like 4-5 minutes in between each WU. I don't run multi instances but it does take a good few minutes. I would bet I lose about an hour a week on EACH machine here when sending/receiving.
 
Hmm...I guess I've not been around long enough to see
outages and shortages...maybe it would be wise for me to run
an extra client on the machines that do nothing but fold....

I wonder if that would kill my efficientcy. If you remember I was
the #1 most efficient in that little thing Rav put together...
Granted, I'm running 6 high end machines and only one of them
is used for anything often enough to impact folding...plus I
don't have any low Mhz machines running....

I guess I'll just have to balance it out as I see it over time.
 
Because I don't use a router I manually get and send work (with broadband). I have found that the response time is faster if you monitor the finish and immediately send. In doing this my work increases tremendously. It is time consuming though trying to stay on top of it. At other times I just look to see which one will finish next and leave the net connected to that machine. This is when I notice how long the system waits to get work. I have actually seen a system waiting and shutdown the client and restarted and immediately got work. I don't know what that is all about. I guess what I'm trying to say is that from my experience the more direct involvement I give the higher my production. Another reason that I use multiple instances during the week is that I am away from the farm 12hrs each day and invariably (especially with small gromacs) they are down waiting for me when I get back. I lose about 40% of my current production trying to do it that way.

To answer the question of why I don't use a router. Broadband comes with where I live, free (T1 Lan) but the company no longer allows a network on their network and I haven't learned a workaround.:( If any of you administrators have any suggestions I'm all eyeballs.:D
 
Dude, just take a 500Mhz pentium or something, and
download Smoothwall Linux Router from www.smoothwall.org

The ouside will only see one computer, while inside you could
have 256 computers running. Unless they actually come in
and see what you're doing, I don't see why you couldn't
have this or any other router!

Plus, it took a Linux Newb like me 15 minutes to set up!!
You just download the image (should be quick on your T1)
burn it to CD, boot from CD, and follow the instructions!

Super easy!
 
Unless the address he's getting out on is being port address translated at the edge of the LAN before hitting the web. Then he'd still be bumming. I'm assuming smoothwall will take the one IP address he's allowed, and do the whole "one-to-many" PAT deal.

I'm pretty sure you can't get away with that twice in the same route. He'd need to go groveling to IT for a one-to-one NAT to make it work, at which point they'd know what he's up to...;)
 
JetMech said:
To answer the question of why I don't use a router. Broadband comes with where I live, free (T1 Lan) but the company no longer allows a network on their network and I haven't learned a workaround.:( If any of you administrators have any suggestions I'm all eyeballs.:D
JetMech how would they even know you had a network on their system? Like what was said, a router will only show ONE IP address on the network, you could theoretically have 200 machines behind it but the 'outside' network will only see 1 IP. If they use a MAC address to verify the modem, many routers (my dlink has this) will clone a mac address so the outside network 'thinks' it is seeing the modem and not anything else.
 
I don't think it's so much them knowing as it might be just technically impossible given the environment. It all depends on how he ultimately gets an IP address on the web from the admins on his LAN. My previous point is that if they already have their firewall for the whole LAN doing a one-to many translation (PAT) so that say 100 users all appear to come from the same IP on the web, then he'll have trouble if he throws another device doing PAT behind that. His data will only be routable until it reaches the edge of the LAN. Once it gets translated again, replies from the outside won't know how to get back to him. NAT / PAT as a security and economy measure is cool, but sometimes it svcks...

To verify - from your system, go to network-tools.com, and note the address filled in on the form. That's your address on the Internet. Next, go somewhere else (maybe a few places) on the same LAN, and see if the same test ever gives the same result. If it does, that may be why it won't work when you add your own router doing NAT/PAT. You would need for them to give you your own address as a static NAT (eg. this routable internet address = your router's outside interface and only that) on their firewall to get things cooking. I've also seen configs where there are several Inet addresses farmed out for clients, and all but one of then are one-to-one until the last one. Systems on this last address get the one-to-many PAT deal. It can make it hard to verify unless you just come out and ask them.

Hope this helps
 
Ok this is what I'm seeing (haven't used any of your suggestions yet). I shut down all machines. Connect them to the router (SMC 8 port), shut down and restart router start 1 PC. It has internet access immediately. I go to the status page for the router and it shows the box connected. I browse awhile with lightning fast access. Start second machine. It immediately has lightning fast access. After about 5 minutes the lightning fast degrades to no access on either machine and the status page shows the connection disconnected. Shutting down the two machines and restarting everything and the cycle repeats. Lightning fast to disconnect. I originally had four machines on a linksys 4 port router for about 6 months with four machines and no problem. It was immediately after adding a fifth machine last September that this problem started. Thought it was the router at first but after upgrading to the 8 port still having trouble.
 
Another troubleshooting step:

Do you:

A. Have consistent access to systems on your own personal LAN when you are having trouble with Internet access?

B. Have consistent access to systems on the host LAN that you are attached to (company intranet, internal mail servers, file / print services) when you are having trouble with Internet access?

If both these things hold true, it's more fuel for my "double PAT" idea. Also, just FYI, the quirky stuff you're describing is consistent with what I've seen with this in the past.

Good luck!
 
Back