To RAID or not to RAID???

[djnes]

...that is the question. Tonight, I'm reformatting my main system. I've read the various posts and articles claiming RAID doesn't do much in a workstation environment, but when I run benchmarks with PCMark2004, HDTach, and Sisoft Sandra, the benchmarks show a huge difference? Am I missing something, or is this another case of synthetic benchies not matching real world performance. Basically, I'm trying to decide if I should leave my 2 Raptors in RAID0 or split them into 2 drives?

[blackrino9]

This may be the opportunity to put this issue to rest once and for all.

You should post all the bench's before and after with the two separate configs as well as your real-world experience.

[Shemazar]

Well, Anandtech seems to say no to using two Raptors in RAID-0.

Article: Western Digital's Raptors in RAID-0: Are two drives better than one? (http://www.anandtech.com/storage/showdoc.aspx?i=2101)

And from personal experience... recovering from a RAID-0 failure sucks, bad.

[arkamw]

I see a few conflicting issues. First, the risk of data loss. However, if you keep everything backed up there shouldn't be any real problem with that.

Second, check out these articles: Overclockers (http://www.overclockers.com/articles1063/) and Anandtech (http://www.anandtech.com/storage/showdoc.aspx?i=2101&p=1) for some results (has HardOCP done one of these yet?). They both reach the conclusion that "real-world" performance in terms of load times don't change much, if at all. However, they also both say that anything that is very IO heavy might see a benefit. I think it would depend very much on your application and whether you need the write times to be better.

But, from the MaxPC Speed Issue, (Jun04), adding another drive in RAID0 almost doubled the read speed (they performed no other tests). The 16k stripe did the best for them.

I think that it's a tough call. If I were to build a new system right now, I think that I would go with two hard drives (maybe the Hitachis with the IO optimizations) in a non-RAID array with a good backup system (still don't fully trust the Hitachi/IBM drives, waiting to see about them). If real-world doesn't make that much of a difference, what's the point?

Cheers.

edit: heh, got beat to the linkage

[djnes]

This may be the opportunity to put this issue to rest once and for all.

You should post all the bench's before and after with the two separate configs as well as your real-world experience.

That's a good idea. I'll try it without RAID and see how it goes. I'm not sure what apps are available to me for "real world" performance though.

[arkamw]

I'm not sure what apps are available to me for "real world" performance though.

How about just non-RAID / RAID comparisons of load times for Windows (this one might be hard because of the RAID intialization), different games (Far Cry, UT2k4, Neverwinter Nights, Sacred, etc), applications (any of the Office stuff, Photoshop, Premiere, compilers?, etc). And maybe the feel of how fast things load, larger movies, pics, etc.

I'd be curious as to the results.

Cheers.

[evilmicah]

Hey, it's my first post here.

FWIW - I run a 1.25TB Raid5 array using a Supertrak SX6000 using six 250GB WD's. It's really not a matter of the speed, it's more a matter of having a large usable disk with a level of fault tolerance. I just upgraded the array with the 250GB drives yesterday. Previously I was using 120's, but found myself almost out of space.

I don't really think the performance of RAID should be a strong selling point, instead I would consider fault tolerance. Of course if you are considering striping for the speed of access, I would recommend changing block sizes and benchmarking with sandra to find optimal performance. I've run raid 0,1,0+1, and 5. Never really bothered with Benchmarking, just had the drives around and figured why not.

If you manage to do some comparative benchmarks I would be interested in seeing the results.

[Leadman584]

I've been running Raid 0+1 via the siI 3114 built onboard to my MB, tomorrow(well actually it's today) I get my Netcell Raid 5 card. Hybrid RAID 3, but who cares. Current data throughput is 82-85MB/S Hope to climb into the mid 120's with RAID 5 array. RAID 0 is never a good idea unless you religiously back it up daily. New card supposed to e-Mail me if a HD fails, gotta wait and see.

[Yogi]

But, from the MaxPC Speed Issue, (Jun04), adding another drive in RAID0 almost doubled the read speed (they performed no other tests). The 16k stripe did the best for them.


Maximum PC has never really known much about hard drive performance. They rely on simple file copys from a ramdrive and on synthetic benchmarks. I really like the magazine and all, but thay haven't got a clue when it comes to hard drives and real world performance.

Not only has Anandtech looked into this in depth, but www.storagereview.com has also declared that raid 0 is useless in a single user enviroment. If there is any source of hard drive information and knowledge I can trust completely, it's storagereview.

If you love to compare synthetic benchmarks and can stand the risks, go ahead and use raid 0, but for real world workstation performance, you will see almost zero benefit.

[arkamw]

Maximum PC has never really known much about hard drive performance. They rely on simple file copys from a ramdrive and on synthetic benchmarks.

While the merits of how they test things can be argued, I threw it in there for a different type of comparison. It should be noted that in the quoted article, only synthetic-type benchmarks were used (HDTach) and only one result of the tests displayed.

Cheers.