13

I am thinking of upgrading from a Western Digital 10k RPM hard drive to using two Intel-X25 solid state drives in RAID 0 configuration. I will be using this for hosting the operating system and installing applications. But all data is currently and will remain stored on a separate Seagate Barracuda 7.5k RPM hard drive.

My goal is to significantly improve performance of the OS (Vista Ultimate) and startup of applications. I have two concerns...

1, Are Intel-X25 drives reliable enough to use in a hard working machine especially when you have two of them in RAID 0 and so double the chance of failure.

2, I hear that solid state drives have a limited number of write cycles before they stop working and so is heavy desktop usage going to cause them to expire within a couple of years? My expected upgrade cycle means I would probably replace my machine after 2 years anyway.

2
  • retag with raid?
    – Alex B
    May 3, 2009 at 13:15
  • retagged as suggested. May 4, 2009 at 4:58

9 Answers 9

13

All drives fail. Either through wear, or though manufacturing fault. By placing those drives in a RAID0 configuration you have doubled the chances of a drive fault destroying your data.

5
  • 4
    Not only are chances increased due to a lack of redundancy, but you are now using a device that has a shorter lifespan, and possibly a shorter MTBF as well. Over the short term, this may not be an issue; but for long-term storage, it's trouble-in-a-box(tm). +1 for pointing out the fragility of this arrangement. May 20, 2009 at 7:20
  • 2
    actually if the probability of one drive failing is p, then the probability that one of two will fail is 2p-p^2 (which is almost 2p if p is small) Jun 4, 2009 at 16:59
  • @smoofra, what is the calculation for 3 drives? 3p - p^3 ? Jun 4, 2009 at 23:50
  • @ Dave Cheney: 3p - 3p^2 + p^3
    – dave4420
    Jul 1, 2009 at 9:11
  • 6
    You guys aren't thinking in SSD-specific terms. Contrary to how traditional platter-based drives work, the lifetime of a SSD is directly related to the number of write cycles it executes. Since two drives in RAID-0 are only doing 1/2 the amount of writing, you are likely doubling your MTBF compared to a single drive. Of course, you are still twice as likely to experience other failures. But unlike mechanical drives there's a not a lot else that can fail on SSDs. Conceivably, yes, the controller chip could die or a capacitor could blow. Backup, backup, backup.
    – John Rose
    Jul 1, 2009 at 17:25
5

Dodgy SSD's tend to halt for up to a second which is fatal to RAID usage, but if you're using the Intel ones, don't fret and go ahead and RAID them, they'll probably live longer than your 2 years target until the flash wears. Just don't forget to update the firmware

3

It depends on the level of use. There's easily an order of magnitude difference, or more, in the number of writes that magnetic media will tolerate vs. flash media. Because of the much lower write capacity of flash, SSDs remap logical blocks to different physical blocks in a process called "wear leveling". The X25m has some known problems with wear leveling issues causing dramatic performance loss with use. I don't know if the X25e has a similar problem or if the wear leveling is less aggressive due to the higher write-capacity of the flash cells.

Personally, for now, I'd trust SSDs on a workstation but not on an important high-capacity server.

3

As you're talking about your system and applications drive, then reliability shouldn't be a concern, right?

You have no important data on it, if it breaks - just replace the drive(s) and restore the system from the last system backup. Easy, fast and shouldn't cost you more than an hour or so.

Generally SSDs (while flaky in performance some times) have a much more reliable mode of operation - and can warn you when it starts to break down. Of course, interface electronics and other stuff can suddenly stop working, as always. And as noted in other answers, performance may degrade with wear - replace them in time.

3

I'd never use standard spinning-disk HDDs in RAID-0 for data I cared about, but you may find that SSDs actually have a longer lifespan in RAID-0 because the number of writes going to each drive is reduced.

There are a number of optimizations you can use to reduce the amount of disk writes that Windows performs.

In a nutshell:

  • Move the pagefile to another drive, or disable (Windows can act funky if you disable it entirely)

  • Disable features like prefetch

  • Consider disabling the "last accessed" file attribute, as this will cause a disk write even if you are only reading a file

  • Disable your web browser's disk cache or move it to a standard HDD

This is a good source for SSD optimization tips. While the product-specific information is obviously only applicable to OCZ's products, there are a lot of Windows SSD optimization tips on this forum that would be applicable to any SSD. http://www.ocztechnologyforum.com/forum/forumdisplay.php?f=186

3

From my experience, and the views of others the Intel X25's are a pretty reliable drive and should be as reliable (if not more) as standard hard disks. Using RAID 0 does increase the chance of failure, but thats the chance you take with the increased speed.

As for the write cycles, it depends on the model you get, the X25-M uses multi level memory cells which are limited to 10,000 cycles, whereas the X25-E uses single level and is limited to 100,000 write cycles.

2

I would not consider using Vista with SSD's. Vista was not optimized for use with SSD's (nor was XP) and has some serious performance bottlenecks. You would be better off with SSD's on Windows 7 which has shown to have significant SSD perofrmance improvements.

SSD's may be susceptible to static electical disturbance. As long as you don't care if you lose everything (say you have good backups and can restore from an image) then RAID0 is fine and should give a nice boost, though i've never seen any benchmarks to confirm this with SSD's.

1

Although the theory of operation of an SSD should make it more reliable than a hard drive, the technology and resulting products are immature.

You should expects SSDs to have a failure rate similar to any electronic component. Since the consequences of data loss are often more severe than the consequences of downtime due to a failed CPU, redundancy makes more sense. Realistically, consumer-grade SSDs will have a greater failure rate than expected because they're so poorly made, and high-end SSDs will have a greater than expected failure rate because they're bleeding edge technology.

Even if the data is not important, the downtime and work caused by recovering the system can be significant. I worked in one environment where a handful of sysadmins manage many hundreds of systems with exclusively redundant storage. There was at one point a belief that it was acceptable to have nonredundant storage for unimportant easily-recoverable data. Luckily, the lesson as to why this is wrong was properly learned. Recovering a system from a drive failure tied up a sysadmin for hours. The cost savings in the price of a drive was nothing compared to the waste in paying someone to recover the system.

-1

Remember: RAID0 is NOT RAID. The "redundant" bit is not present.

RAID originally stood for "Random Array of Inexpensive Disks" and in this context RAID0 is indeed RAID. I'm not sure when people started using Redundant as the R, but it's incorrect.

1
  • [citation needed] Wikipedia suggests that you're incorrect: en.wikipedia.org/wiki/RAID Specifically, see the reference at the bottom that says "Originally referred to as Redundant Array of Inexpensive Disks, the concept of RAID was first developed in the late 1980s by Patterson, Gibson, and Katz of the University of California at Berkeley. (The RAID Advisory Board has since substituted the term Inexpensive with Independent.)"
    – Bill Weiss
    Nov 20, 2011 at 15:16

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .