Wanted: Advice Concerning a New Rig and RAID!

[ChanceCoats123]

Hello everyone, I just had a few questions I wanted to ask and I'm looking for some opinions.

I'm working on selecting parts for a new rig this month/fall. I've pretty much settled on a 5930k w/ Corsair H105, Asrock X99 OC Formula, 32GB G.Skill Ripjaws, (reuse for now, then go to X-Fire in the future) R9 290X, Samsung 512GB 850 Pro for the OS and a few games I play more often. All of this will be powered by an EVGA Supernova G2 1300w unit and stuffed into a Phanteks Enthoo Pro case (no window, I'm trying to get a more simple looking case).

The only part of this build I am unsure of is the storage. I currently have single 1TB and 1.5TB drives in my rig with no RAID or redundancy. I don't mind their performance because I mainly have my movies and music stored on them. However, I am getting a bit annoyed by slow loading times in games and very slow data movement (reading from my HDD to other network connected computers), and because my 64GB SSD (original OCZ Vertex if you believe it!) is just big enough for the OS and a handful of VERY common programs, I don't have space on there.

So here comes my question:

Do I even need to be thinking about RAID here? My current plan is to buy 4/5 1TB WD caviar black drives to run in RAID 5. I figured RAID 5 was a strong contender in terms of space utilization and speed (considering the reads speeds are theoretically nx ). This configuration easily covers my speed requirements (would use Hardware RAID), and also gives me the best bang for my buck from a redundant array (RAID 0 would be best for space, but has zero redundancy).

I've read all about the write hole (and I've thought about buying a UPS for other reasons as well), but I'm not too concerned since this can be avoided by synchronizing the array daily/weekly as well. The next issue is the potential for a rebuild to encounter a URE and corrupt the entire array. I've looked into using RAID 6 instead, but I don't have the money to throw at an enclosure or card capable of 5/6 drive RAID 6. If anyone has other ways for achieving affordable RAID 6, I'm all ears!

For now, I don't really have information which really requires redundancy, but as I'll be a university junior this fall, I'll soon be writing a lot more of my own code, designing circuitry, and other more important school work. This kind of IP is not something I want to sit around on a flash drive or an external drive in my apartment (as a backup of these files). I'd rather have a drive setup I can trust with my data and then maybe a secured backup on another rig at home.

ANYWAY, thanks for taking the time to read that novel and posting in here, I appreciate the advice!

 

[ATMINSIDE]

Definitely go for hardware RAID with a dedicated card and good drives.
I'd be going with WD Red, as they're built with features specific for RAID operation.

My server is running a PERC 6/i with 4x2TB WD Red in RAID5, it's perfect for what I do.

 

[trekky]

for faster speeds I would do a raid 10, my work setup gets ~250 write and ~330 Read
while my raid 5 gets only ~70 write and ~250 read

 

[ChanceCoats123]

Definitely go for hardware RAID with a dedicated card and good drives.
I'd be going with WD Red, as they're built with features specific for RAID operation.

My server is running a PERC 6/i with 4x2TB WD Red in RAID5, it's perfect for what I do.

While the price of the reds is very attractive, I'd rather have a true 7200 rpm drive with the black. The cache is the same size, and the warranty is actually 2 years longer compared to the reds (not the red pro).

for faster speeds I would do a raid 10, my work setup gets ~250 write and ~330 Read
while my raid 5 gets only ~70 write and ~250 read

I was trying to avoid mirrors because they are so damn space inefficient, but I might be able to work RAID 10 with 4 2TB drives. That might not be bad at all.

 

[Automata]

You can't avoid the write hole by rebuilding/"resynchronizing" the array. The data is gone in that situation, and you run the risk of corrupting everything.

Instead of going down this route for RAID, why not use one big drive, and back up the contents. There are a lot of cheap options available: CrashPlan is the one I suggest, and BackBlaze is another very good option. 1TB drives that old are going to be prone to failure, simply due to age and usage. Going with a single, good quality, large drive has a lot lower chance for failure. RAID will save you if a drive fails, but it isn't going to help with user error or other types of corruption.

 

[ChanceCoats123]

You can't avoid the write hole by rebuilding/"resynchronizing" the array. The data is gone in that situation, and you run the risk of corrupting everything.

Instead of going down this route for RAID, why not use one big drive, and back up the contents. There are a lot of cheap options available: CrashPlan is the one I suggest, and BackBlaze is another very good option. 1TB drives that old are going to be prone to failure, simply due to age and usage. Going with a single, good quality, large drive has a lot lower chance for failure. RAID will save you if a drive fails, but it isn't going to help with user error or other types of corruption.

CrashPlan looks really nice, but I'd like to avoid paying for a service I don't really need, or could do myself (obviously not to the same extent)

The reason I don't want to use just one drive is because it won't be nearly fast enough. One of the big benefits I see in raid is that the overall array performance can be much greater than any individual drive ever could. I figure 2tb drives shouldn't be that old, and with 5 year warranties from WD, I feel a bit more safe putting my files on those drives. I typically build a new computer before most of my parts become obsolete anyway, so odds are that I would transfer all this data somewhere else down the road before a drive fails.

 

[ChanceCoats123]

Well I've got a quick update and another quick question (that I think I already know the answer to).

I've decided that I'll probably go with a 1TB 850 Pro (since there is currently a nice $50 off promo code on Newegg) combined with 2 x 2TB WD Red drives in raid 0. I initially thought I would get an external enclosure to hold images of the raid 0 array, but I had an idea I wanted to run past you guys.

Is it possible (and if so, is it recommended) to use hardware raid for the 2TB drives in raid 0, and then use software raid to put the array into raid 1 with a single 4TB drive which basically acts as the same as an external drive with images, but with the benefit of automatic and immediate data backup instead of being on whatever schedule I decide to image the array with.

With that question out there, here is my other-other thought: 2x2TB drives in raid 0, then use a program like Acronis True Image to image the array to a single 4TB red drive on an incremental schedule.

Sorry to keep pestering you guys with my ever-changing ideas, but I really appreciate the help!

 

[Automata]

Is it possible (and if so, is it recommended) to use hardware raid for the 2TB drives in raid 0, and then use software raid to put the array into raid 1 with a single 4TB drive which basically acts as the same as an external drive with images, but with the benefit of automatic and immediate data backup instead of being on whatever schedule I decide to image the array with.

If you want to run something that is incredibly fragile, probably, yes. It is a disaster waiting to happen, though.

What are you attempting to achieve? If you need the reliability of RAID 1, then do RAID 1. Patchwork setups will give patchwork results. Keep it simple.

 

[ATMINSIDE]

What does the Pro do for you that the EVO won't?

 

[ChanceCoats123]

If you want to run something that is incredibly fragile, probably, yes. It is a disaster waiting to happen, though.

What are you attempting to achieve? If you need the reliability of RAID 1, then do RAID 1. Patchwork setups will give patchwork results. Keep it simple.

That's very true. In hindsight, using two different raid setups creates way too many failure points.

I'm trying to achieve a storage solution which is faster than using multiple single drives or JBODing disks together while still keeping myself better protected from drive failure. I don't have any mission critical data which I feel truly needs to be put in a redundant array and then backed up off site. My goal is to simply have my data easily replaceable in the event a drive dies or I have some other kind of hardware failure. As of now, I have zero data protection for my main computer, but since I don't have anything really important, I'm not too worried about a drive failure (which has happened in the life of this rig).

Sorry if I'm rambling. The moral of the story is that I want 4tb of actual storage space which is faster than a single drive, and I want to keep some kind of image of this data so I can recover from hardware/software failure.

What does the Pro do for you that the EVO won't?

This is something I've indeed thought about. Performance wise, the drives will be almost identical. What really warrants the extra ~$90 (in my opinion) is the warranty length of the Pro being 10 years vs the Evo being 5 years. And the TBW value for the Pro is 300TB vs the Evo at 150TB. I know I'll probably never reach those data limits, but the longer I can keep this drive the better. Considering the 1TB size, it's not crazy to think I could move this drive from computer to computer for the next few years/builds.

 

[Automata]

If you need speed, solid state is going to be The Option. What are you doing that requires more than 1tb fast storage? You can't push data off to slower disks after whatever you are doing is done?

 

[ChanceCoats123]

If you need speed, solid state is going to be The Option. What are you doing that requires more than 1tb fast storage? You can't push data off to slower disks after whatever you are doing is done?

Well now that you mention it, I had overlooked the change from 512gb to 1tb for the SSD. The reason I want faster storage was for network data movement between my different computers, but also to help future proof this build down the road. With the 1tb SSD in mind, I could probably just use a single 4tb storage drive. I don't really want to go smaller than 4tb because I'm currently around 2tb of data (music, movies, pictures, programs, games, schoolwork, backups of various devices, etc.) between a 1tb and 1.5tb drive and I would like space to continue my collections.

Edit: Here is my main issue with going to a single drive: I can use two WD red drives in raid 0 to get 4tb of space with better speeds than a single WD black 4tb, but the single costs more (which is a bit counter-intuitive, but comes from the black vs red difference). I'm hesitant to use a single red drive for storage considering the rotational speed of the red drive and the increased I/O latency. If I were to go with a single drive, it would be a black drive, but that brings up the above dilemma. It seems more practical, in my opinion, to put 2x2tb WD red drives into raid 0 and image my SSD and array to a single WD red 4tb with a program like Acronis. Does this make a bit more sense from a reliability and cost stand point to you guys?

 

[ATMINSIDE]

FORGET RAID 0 FOR DATA STORAGE. You're literally asking to lose your data.

 

[Automata]

Does this make a bit more sense from a reliability and cost stand point to you guys?

To me? No.

Mechanical drives years ago could read/write at 125 mb/sec, which is faster than gigabit ethernet. Once you get out of sequential copying, no type of RAID setup with mechanical drives is going to max gigabit. RAID actually gives you a performance penalty to random access times (unless you short stroke, but that is true for single drives as well) because there is an extra layer and access is simultaneous, which means the RAID controller has to wait for both drives to finish the operation before returning to the OS.

Trying to add RAID into desktop systems is a Bad Idea at best. You should only use it if drive failure of a mission critical system is completely unacceptable. I can't think of any home use for this, except a file server.

 

[ChanceCoats123]

FORGET RAID 0 FOR DATA STORAGE. You're literally asking to lose your data.

I promise I'm not being dense deliberately, I've just never really looked into the different storage techniques before. If I am imaging my array on daily basis, then shouldn't I be pretty well protected from an array or drive failure? And the raid 0 is not intended for data storage, it's more intended for data access. I know it's a matter of symantics, but the array is there so I can quickly access and transfer my data. Each day, the image created of said array is really the "storage" part of the whole system.

In which case and I more protected from data loss: Raid 0 drives with daily imaging or single drive with no imaging?

I understand there are reliability issues, and raid 0 is really more like aid0 since there is no redundancy. But if I'm imaging the entire array to a different disk, then I shouldn't I theoretically gain the extra performance of a striped array with an additional backup image (of both my array and the SSD really)?

Edit: Just saw Thid's post. I guess I'll just figure out the best way to find a single drive setup for the build.

 

[ATMINSIDE]

I promise I'm not being dense deliberately, I've just never really looked into the different storage techniques before. If I am imaging my array on daily basis, then shouldn't I be pretty well protected from an array or drive failure? And the raid 0 is not intended for data storage, it's more intended for data access. I know it's a matter of symantics, but the array is there so I can quickly access and transfer my data. Each day, the image created of said array is really the "storage" part of the whole system.

In which case and I more protected from data loss: Raid 0 drives with daily imaging or single drive with no imaging?

I understand there are reliability issues, and raid 0 is really more like aid0 since there is no redundancy. But if I'm imaging the entire array to a different disk, then I shouldn't I theoretically gain the extra performance of a striped array with an additional backup image (of both my array and the SSD really)?

By doing all that mess, you're adding complexity. Don't.
You want it mirrored? Use some version of RAID 1.

Gigabit can be saturated by 1 HDD, so you don't need some massive array for speed.

 

[Automata]

Do you seriously want the un-reliability of RAID 0 - which is especially true for on board RAID - and having to image the stupid thing constantly, or fiddle with a program to back the thing up? You want a system that is set and forget. Backup your critical data online or build a dedicated file server (or both!).

You are building a serious headache with this kind of setup. Keep It Simple.

 

[ChanceCoats123]

Latest plan (just came up with this one!):

1TB 850 Pro - OS + Programs + Games + Data I need quickly
2TB WD Black - Storage
4TB WD Red - Disk Images + Overflow (slow) storage

As a side note, there really isn't much to fiddle with creating disk images. For a $50 one-time-fee, I can purchase a program which runs a schedule to create full and incremental images of my drives whenever I want, help restore said images if I need to, and provide tech support. I agree that cloud storage is definitely a good way to go, but I don't need to pay to keep my movies, songs, and pictures in the cloud (when Google already has most of it). That's just a bit ridiculous to me.

Anyway, let me know what you think of the new(new(new)) plan. This seems a lot more simple to me (and hopefully you two as well), avoids raid, and costs significantly less.

 

[Automata]

That's a lot better and going to give minimal headaches.

You won't need to store your media online, since you can re-rip or whatever easily. Backup your critical data (data you can't reproduce or re-obtain), which is going to be a lot less space, then invest in a backup plan. Blindly backing up everything is also asking for trouble. You need to know your data.

 

[ATMINSIDE]

Latest plan (just came up with this one!):

1TB 850 Pro - OS + Programs + Games + Data I need quickly
2TB WD Black - Storage
4TB WD Red - Disk Images + Overflow (slow) storage

As a side note, there really isn't much to fiddle with creating disk images. For a $50 one-time-fee, I can purchase a program which runs a schedule to create full and incremental images of my drives whenever I want, help restore said images if I need to, and provide tech support. I agree that cloud storage is definitely a good way to go, but I don't need to pay to keep my movies, songs, and pictures in the cloud (when Google already has most of it). That's just a bit ridiculous to me.

Anyway, let me know what you think of the new(new(new)) plan. This seems a lot more simple to me (and hopefully you two as well), avoids raid, and costs significantly less.

Why fuss with a schedule and extra program when RAID 1 will automatically (and instantly) keep a backup?

Still don't see the need for the Pro over the EVO, I've carried my 840 EVO through multiple builds and had it for a few years now.
You're talking about two of the most reliable SSD's on the market.

My plan would be a 1TB 850 EVO with 2x 4TB WD Red in RAID 1.
That said, your newest plan is much better than the previous ones.