Reliability of SSDs?

[theevilsharpie]

Are SSDs reliable enough to safely run in RAID 0 for general use in a production environment?

Backstory:
I'm currently running a 3ware 9650SE RAID controller with four 250GB Seagate Barracuda 7200.8 drives in a RAID 5. The RAID controller has always been a bit flaky, and its most recent problem is locking up for 5-10 minutes during heavy random I/O, like spinning up multiple virtual machines. In addition, there's been a recent spate of log entries indicating that the RAID controller has had to correct parity errors, which the documentation suggests is the result of bad blocks on one or more of the physical drives. The documentation also states that these parity errors are being corrected without any impact to the array, but I've had two instances in the past month where chkdsk has found file system corruption, something I haven't had to deal with since the days of Windows 98. So far, I haven't lost any data I care about, but instability combined with data corruption is enough for me to look into a replacement.

Anyway, I'm looking into the costs of replacing the controller and the failing drive, and the costs are in-line with the cost of two 256GB SSDs. I originally got my RAID setup for performance, and while it performs well (when it works), two striped SSDs would easily outperform mechanical drives for my use case. These would be going into my workstation as well, so having storage that is noiseless and vibrationless is also a benefit. However, I use my workstation to perform compute-intensive tasks that would be impossible to do from my laptop, so my workstation must absolutely be reliable. In other words, being without the use of my workstation for several days in the event of an RMA would be unacceptable. I could build in some fault tolerance by running three 256GB SSDs in RAID 5, but I would have difficulty justifying the costs, and my past experiences with RAID 5 on mobo-integrated RAID controllers have not been positive.

So AT, what say you? The speed of modern SSDs are nice, but mechanical disks matched with a hardware RAID controller with plenty of cache can also be very fast. I need the most reliable option I can get without breaking the bank, so should I throw caution to the wind and go for the SSDs, or should I play it safe and stick with mechanical disk drives?

Cliffs:
1. RAID controller and one or more disks failing
2. Looking into replacement, notice two 256GB SSDs can be had for the same price
3. Interested in running two 256GB in RAID 0, but concerned about reliability; significant downtime not acceptable
4. Your thoughts?

 

[Coup27]

Yes, if you stick to the well regarded SSDs.

Many around here, including me, regard the Samsung 830 series as the most reliable SSD on the market. If you bought two of those and updated to the latest firmware before RAID'ing them together you will have a stable setup.

I presume you have a good backup solution if you are using RAID0 in a production box?

 

[theevilsharpie]

I presume you have a good backup solution if you are using RAID0 in a production box?

Yeah, I back up my data regularly, so I'm not too concerned about data loss. However, backups don't help me if a client needs a same-day design/PoC done and my workstation is out of commission because a drive failed and a replacement is days away.

 

[Coup27]

This is true, however as you already know, everything can and generally does eventually fail.

I don't believe a quality SSD like the 830 has a higher rate than a quality SAS hard drive.

I work for a small company and I started a new IT project today. I will be deploying 12 new computers for my users and all will use an 830. I have one also in my home PC and I can't speak highly enough of them.

 

[theevilsharpie]

This is true, however as you already know, everything can and generally does eventually fail.

True, but I'm used to solid-state components like RAM, CPUs, and mobo ICs lasting substantially longer than mechanical devices like hard drives and fans, which I why I build in fault tolerance for the latter. Not sure which category SSDs fall into.

 

[groberts101]

quick key points.

having "reliable.. and R0" in the same sentence.. would be an oxymoron. lol

however.. "reliable enough.. and R0" is certainly doable as many including myself use them in workstation environments. Although my major data flow is so sporadic it would not be in your same catagory. 6 drives in R0 for over 2 years is nothing to stick your nose up at either. I also run 3 raid cards for storage/backup OS's too and usually R0 everything possible.

The key is to never underestimate the OS's..or the hardware's.. ability to screw you over. Ocassional SE/reimaging can get you back to fresh physically/logically again and avoids many complications over the long run.

And even if you lost a drive?.. restoring a backup image to the remaining SSD again takes little time at all. Just think of it as an.. "R1 array that you manually build on demand". lol

 

[theevilsharpie]

And even if you lost a drive?.. restoring a backup image to the remaining SSD again takes little time at all. Just think of it as an.. "R1 array that you manually build on demand". lol

I have to consider it as a RAID 0, as a single 256GB SSD doesn't have enough capacity to contain my data.

 

[bryanW1995]

Honestly, it sounds like you need the reliability of RAID 5 or RAID 1.

 

[sequoia464]

http://www.tigerdirect.com/applicati...NLxoFT4pGCZ2iA

Get three of these M4's and go raid5 - No tax to So Cal. = $600.00

(courtesy of slick deals)

 

[npaladin-2000]

I suppose they're stable enough for RAID0. If you consider RAID0 stable enough for production. And I can't think of a single reason for trusting something labeled "production" to the redundancy killer that is RAID0. As in Zero Redundancy of Independent Disks.

Honestly, you're better off shelling out a little extra and getting a 512 GB SSD, they can be had for under $600. And they're generally faster than the 256 GB ones anyway, plus you'll get proper TRIMing of the drive.

Whatever you do don't run RAID5 on that 3Ware...I hate those things...

 

[bryanW1995]

256gb m4's are faster than 512gb versions, fyi.

 

[npaladin-2000]

256gb m4's are faster than 512gb versions, fyi.

Not based on the reviews I've read. And besides that, it would make no sense anyway. More NAND = more paths = faster operations.

 

[Burner27]

Not based on the reviews I've read. And besides that, it would make no sense anyway. More NAND = more paths = faster operations.

Yes but more NANDs + same amount of channels as the 256GB drive could bottleneck the controller.

 

[npaladin-2000]

Yes but more NANDs + same amount of channels as the 256GB drive could bottleneck the controller.

Could, but it doesn't in the reviews I've read on sites that have tested both models (unfortunately, AnandTech hasn't). In fact, the only time I've seen the 512 test out slower than the 256 is the Samsung 830. As I said, generally the 512 GB version of an SSD is faster than the 256 GB version.

 

[theevilsharpie]

I suppose they're stable enough for RAID0. If you consider RAID0 stable enough for production. And I can't think of a single reason for trusting something labeled "production" to the redundancy killer that is RAID0. As in Zero Redundancy of Independent Disks.

Honestly, you're better off shelling out a little extra and getting a 512 GB SSD, they can be had for under $600. And they're generally faster than the 256 GB ones anyway, plus you'll get proper TRIMing of the drive.

A single 512GB SSD is not going to give me any benefit over two 256GB SSDs. If either fail, I'm fucked.

Honestly, it sounds like you need the reliability of RAID 5 or RAID 1.

Yeah, it's looking that way. My only problem with RAID 5 is that write performance tends to be absolutely atrocious on integrated fakeraid controllers because of a lack of write cache. However, the low access times on an SSD might blunt that a bit.

Does anyone have any benchmarks of modern SSDs in a RAID 5 array using an integrated mobo controller (preferably an AMD controller)? I see several RAID 5 benchmarks, but thy're all use dedicated hardware controllers.

 

[_Rick_]

A single 512GB SSD is not going to give me any benefit over two 256GB SSDs. If either fail, I'm fucked.

But you reduce p(FAILURE) by around 50%. One device less, is one device less likely to fail.

 

[npaladin-2000]

But you reduce p(FAILURE) by around 50%. One device less, is one device less likely to fail.

You also don't have the RAID controller filtering out the TRIM commands, which will help maintain performance levels.

 

[Anteaus]

It seems if your biggest need is minimizing downtime, then you might be seeking a solution that is unnecessarily complex. A raid 0 or 5 failure is going to require intervention on some level, so neither solution means you can fire and forget.

My suggestion is to get out of raid completely. Spend a little money and buy two of the most reliable SSDs you can find with enough capacity in a single drive configuration. Use one SSD as your primary drive and keep the second one on hand for swap out should the first fail. Keep two backups if you really are worried.

Raid has it's place, but in this particular case I think your actually increasing risk while at the same time actually increasing potential downtime due to drive failure because then your dealing with raid arrays and secondary drive controllers.

With a simple drive swap, you'd be back in business within minutes. In fact, in this particular case I'd still recommend this setup if you stuck with HDDs.

 

[bryanW1995]

You also don't have the RAID controller filtering out the TRIM commands, which will help maintain performance levels.

Until 11.5 IRST finally comes out, right? Besides, modern ssd's with decent GC perform just fine without TRIM.

 

[bryanW1995]

256gb m4's are faster than 512gb versions, fyi.

Not based on the reviews I've read. And besides that, it would make no sense anyway. More NAND = more paths = faster operations.

http://www.anandtech.com/show/4253/the-crucial-m4-micron-c400-ssd-review looks the same here for the m4, so we're both wrong.

http://www.anandtech.com/show/5719/ocz-vertex-4-review-256gb-512gb Vertex 4 is faster with 512gb

http://www.anandtech.com/show/5508/...cherryville-brings-reliability-to-sandforce/2 520 series clearly faster in 240gb configuration

http://www.anandtech.com/show/4863/the-samsung-ssd-830-review 830 both ssd's theoretically is the same, though you say that it is faster with 256gb version


It looks to me like both ~250 and ~500 gb ssd's are in a "sweet spot" for performance, and I assume that we can at least agree that they are both clearly superior to ~120 gb ssd's.

 

[npaladin-2000]

Until 11.5 IRST finally comes out, right? Besides, modern ssd's with decent GC perform just fine without TRIM.

If it ever comes out. And that's assuming he's using the Intel chipset for RAID, and not some add-on RAID chip.

It looks to me like both ~250 and ~500 gb ssd's are in a "sweet spot" for performance, and I assume that we can at least agree that they are both clearly superior to ~120 gb ssd's.

True...unfortunate that the 128 GB ones are starting to hit the mass sales point of $100.

 

[theevilsharpie]

With a simple drive swap, you'd be back in business within minutes. In fact, in this particular case I'd still recommend this setup if you stuck with HDDs.

Two 512GB SSDs would cost over $1,200. For that price, I could purchase a new controller and four 1TB drives, and still have a bit left over.

As for being "back in business within minutes," you realize that it would take several hours to restore 450+GB of data from a backup, right? Even if I were to go with two 512GB SSDs, I'd configure them in RAID 1 so that if one fails, I'd still have use of my workstation while I RMA it.

 

[npaladin-2000]

Two 512GB SSDs would cost over $1,200. For that price, I could purchase a new controller and four 1TB drives, and still have a bit left over.

As for being "back in business within minutes," you realize that it would take several hours to restore 450+GB of data from a backup, right? Even if I were to go with two 512GB SSDs, I'd configure them in RAID 1 so that if one fails, I'd still have use of my workstation while I RMA it.

It all depends on how much the speed is worth to you versus the lack of downtime. Correct me if I'm wrong but it sounds like downtime isn't an option at all, reliability is super important, and you would like speed, but you have trouble justifying expenditures beyond a certain point? If that's so you may have to make a tradeoff, and it sounds like speed is the least important of the three elements?

 

[theevilsharpie]

It all depends on how much the speed is worth to you versus the lack of downtime. Correct me if I'm wrong but it sounds like downtime isn't an option at all, reliability is super important, and you would like speed, but you have trouble justifying expenditures beyond a certain point? If that's so you may have to make a tradeoff, and it sounds like speed is the least important of the three elements?

Speed is important. Otherwise, I'd throw my hardware controller in the trash and use my mobo's built-in RAID.

To clarify, I'm satisfied with my current level of performance, but I'm obtaining that performance by using a hardware RAID controller that appears to be in the early stages of failure, and I'm considering all my options for fixing the situation.

 

[BrightCandle]

Raid 5 with 3 256 GB drives is your best trade off. Write performance is somewhat reduced but you get some redundancy to reduce your downtime at the lowest cost point while maintianing the amount of space you need. You aren't going to benefit from trim at this point so you need to get an SSD with great trim support. Raid 1 for 512GB is more expensive and wont bring anything more in performance and all other solutions are going to cost hours on the failure of a disk.

To get safer or faster you are going to have to trade off against additional cost. Personally I have found a single SSD quite a lot more reliable than HDD's. I have had 4 SSDs running for over 2 years with a single failure while in the same time with 12 hard drives I have had 13 failures. In many ways I am disappointed with the high failure rate of SSDs, they should be acting like solid state components like CPUs but right now they are moderately better hard drives in failure rate.