I just decided to start this thread in response to multiple posters in seemingly every third (or so) thread on these storage forums.
Why would anybody recommend setting up a RAID 0 for a desktop computer (not a workstation) intended for office / gaming / internet browsing use? (99% of desktop users out there)
Now, before you RAID0 fanboys start responding rashly, please finish reading this post.
1) The most common argument given for RAID 0 is that "it is way faster." To quote one poster, "two 7200 RPM drives in RAID0 are faster than a single Raptor 150GB." I think the main question with that argument is, what "faster" really means. Sure, empirical tests indicate that you get faster transfer speeds. That's all fine and dandy, but several tests designed to compare RAID 0 performance to single drive performance indicate that this increase in transfer speed under ideal conditions translates into little or no improvement in the real life applications' (i.e. those things we really run on our desktop computers) performance.
2) Drawbacks: the most obvious one is that a fault of one drive renders the entire array faulty. The results of such a fault may vary drastically; if you had the array partitioned, you may lose that partition, or worst of all you may lose the entire array if the partition data gets lost. In response to many of the posters here, by mere consequence and a most elemental application of the probability theory, this makes
The Probability That RAID 0 array is degraded = N * Probability of A Single Drive Failing
Or simply if you are crazy (and wealthy) enough to set up a 4-drive RAID 0, it will be 4 times as likely to fail as one drive within that array (assuming they are all the same, which is the preferred scenario anyway).
Another drawback is portability. Most controllers, to the best of my knowledge and I may be mistaken here, use incompatible striping algorithms. And so, if your controller goes out, or for whatever reason you want to port your array to a different controller (i.e. new motherboard for built-in controllers or a better competitive add-on controller), you may not be able to just unplug the drives and plug them back in.
Here's a link that tends to suggest that it is improbable that “…RAID 0 offers increased performance for all or even most applications...”:
http://faq.storagereview.com/SingleDriveVsRaid0
With that said, can anyone present empirical tests that back up their claim that RAID 0 results in any significant performance benefit (for real life scenarios and not theoretical benchmarks) that outweigh the drawbacks of such a setup?
I think this should be a requirement for anyone recommending it to other people who come here for good information and advice.
Why would anybody recommend setting up a RAID 0 for a desktop computer (not a workstation) intended for office / gaming / internet browsing use? (99% of desktop users out there)
Now, before you RAID0 fanboys start responding rashly, please finish reading this post.
1) The most common argument given for RAID 0 is that "it is way faster." To quote one poster, "two 7200 RPM drives in RAID0 are faster than a single Raptor 150GB." I think the main question with that argument is, what "faster" really means. Sure, empirical tests indicate that you get faster transfer speeds. That's all fine and dandy, but several tests designed to compare RAID 0 performance to single drive performance indicate that this increase in transfer speed under ideal conditions translates into little or no improvement in the real life applications' (i.e. those things we really run on our desktop computers) performance.
2) Drawbacks: the most obvious one is that a fault of one drive renders the entire array faulty. The results of such a fault may vary drastically; if you had the array partitioned, you may lose that partition, or worst of all you may lose the entire array if the partition data gets lost. In response to many of the posters here, by mere consequence and a most elemental application of the probability theory, this makes
The Probability That RAID 0 array is degraded = N * Probability of A Single Drive Failing
Or simply if you are crazy (and wealthy) enough to set up a 4-drive RAID 0, it will be 4 times as likely to fail as one drive within that array (assuming they are all the same, which is the preferred scenario anyway).
Another drawback is portability. Most controllers, to the best of my knowledge and I may be mistaken here, use incompatible striping algorithms. And so, if your controller goes out, or for whatever reason you want to port your array to a different controller (i.e. new motherboard for built-in controllers or a better competitive add-on controller), you may not be able to just unplug the drives and plug them back in.
Here's a link that tends to suggest that it is improbable that “…RAID 0 offers increased performance for all or even most applications...”:
http://faq.storagereview.com/SingleDriveVsRaid0
With that said, can anyone present empirical tests that back up their claim that RAID 0 results in any significant performance benefit (for real life scenarios and not theoretical benchmarks) that outweigh the drawbacks of such a setup?
I think this should be a requirement for anyone recommending it to other people who come here for good information and advice.