Wow, RAID-0 gets owned!

[scottman]

What about 2 x 76gb raid 1? or even 2 36gb r1? Yeah, you basically lose half your storage but your read/write increases dramatically and you've got total redundancy incase of failure.

I tested some identical drives, 2 in r1 and 1 by itself and the gain in my tests was ~40%.

Scott

 

[imported_Phil]

Originally posted by: scottman
What about 2 x 76gb raid 1? or even 2 36gb r1? Yeah, you basically lose half your storage but your read/write increases dramatically and you've got total redundancy incase of failure.

I tested some identical drives, 2 in r1 and 1 by itself and the gain in my tests was ~40%.

Scott

Yes, of course you did.

RAID-1 does not yield higher read/write speeds. Both the drives are writing the same information at the same time, meaning that you get the speed of the one disk on it's own.

That's why it's called mirroring.

 

[tweeve2002]

Originally posted by: DopeFiend

Originally posted by: scottman
What about 2 x 76gb raid 1? or even 2 36gb r1? Yeah, you basically lose half your storage but your read/write increases dramatically and you've got total redundancy incase of failure.

I tested some identical drives, 2 in r1 and 1 by itself and the gain in my tests was ~40%.

Scott

Yes, of course you did.

RAID-1 does not yield higher read/write speeds. Both the drives are writing the same information at the same time, meaning that you get the speed of the one disk on it's own.

That's why it's called mirroring.

Actully you can get higher read times with RAID 1 because it reads off both disk instead on just one.
Write times stay just about the same.

 

[imported_Phil]

Originally posted by: tweeve2002

Originally posted by: DopeFiend

Originally posted by: scottman
What about 2 x 76gb raid 1? or even 2 36gb r1? Yeah, you basically lose half your storage but your read/write increases dramatically and you've got total redundancy incase of failure.

I tested some identical drives, 2 in r1 and 1 by itself and the gain in my tests was ~40%.

Scott

Yes, of course you did.

RAID-1 does not yield higher read/write speeds. Both the drives are writing the same information at the same time, meaning that you get the speed of the one disk on it's own.

That's why it's called mirroring.

Actully you can get higher read times with RAID 1 because it reads off both disk instead on just one.
Write times stay just about the same.

First, read this, where it says it is theoretically possible. So, I'll change my response to "no, scottman didn't see 40% increases".

 

[Sunner]

Originally posted by: DopeFiend

Originally posted by: tweeve2002

Originally posted by: DopeFiend

Originally posted by: scottman
What about 2 x 76gb raid 1? or even 2 36gb r1? Yeah, you basically lose half your storage but your read/write increases dramatically and you've got total redundancy incase of failure.

I tested some identical drives, 2 in r1 and 1 by itself and the gain in my tests was ~40%.

Scott

Yes, of course you did.

RAID-1 does not yield higher read/write speeds. Both the drives are writing the same information at the same time, meaning that you get the speed of the one disk on it's own.

That's why it's called mirroring.

Actully you can get higher read times with RAID 1 because it reads off both disk instead on just one.
Write times stay just about the same.

First, read this, where it says it is theoretically possible. So, I'll change my response to "no, scottman didn't see 40% increases".

Reads will be faster if the controller supports read load balancing.

 

[imported_Phil]

Originally posted by: Sunner

Originally posted by: DopeFiend

Originally posted by: tweeve2002

Originally posted by: DopeFiend

Originally posted by: scottman
What about 2 x 76gb raid 1? or even 2 36gb r1? Yeah, you basically lose half your storage but your read/write increases dramatically and you've got total redundancy incase of failure.

I tested some identical drives, 2 in r1 and 1 by itself and the gain in my tests was ~40%.

Scott

Yes, of course you did.

RAID-1 does not yield higher read/write speeds. Both the drives are writing the same information at the same time, meaning that you get the speed of the one disk on it's own.

That's why it's called mirroring.

Actully you can get higher read times with RAID 1 because it reads off both disk instead on just one.
Write times stay just about the same.

First, read this, where it says it is theoretically possible. So, I'll change my response to "no, scottman didn't see 40% increases".

Reads will be faster if the controller supports read load balancing.

A 40% increase? I find it hard to believe, to be honest 🙂

 

[Sunner]

Theoretically a 100% increase.
Why do you find that hard to believe? Two disks, equal data, you have to read blocks 1-10, why does it seem unreasonable to read blocks 1 and 2 from disks 1 and 2 simultanously, and so on?

 

[NOX]

Originally posted by: shuttleteam
LOL, RAID0 has its place. Too bad the majority of the users are not using it for what it was intended for, then point fingers or say it's useless or overrated. :roll:

Cheers!

Best post yet!

 

[Alkaline5]

Originally posted by: Tostada
Go to StorageReview and you'll see that a Promise FastTrack card will give you a noticable boost in many things.

I don't think anyone has mentioned this yet:

Currently posted on: StorageReview's Main Page
Though the idea that RAID0 does not significantly assist desktop performance is not new to SR, AnandTech has recently corroborated the idea with results of their own. Anand has confirmed that no significant gain is realized when striping a pair of Raptors off of an ICH5 controller. Interestingly, the feedback associated with Anand?s article reads like a laundry list of the same irrational objections SR had to overcome. Anand is free, of course, to link to SR?s articles which, ironically, address virtually every objection raised by his RAID0 proponents.

"Many things" apparently don't include anything I would want to do.

 

[Sivar]

Yep, StorageReview's recent front-page post essentially states that SR has long stated RAID0 is not worthwhile on the desktop and that Anandtech's conclusion is right on.

RAID 1 is, of course, still a great way to defend your data in case of hard drive failure, especially when combined with a regular backup, since RAID won't help for problems like file system corruption, accidentally deleting files, Windows virii, etc.

 

[Sivar]

Originally posted by: Sunner
Theoretically a 100% increase.
Why do you find that hard to believe? Two disks, equal data, you have to read blocks 1-10, why does it seem unreasonable to read blocks 1 and 2 from disks 1 and 2 simultanously, and so on?

For tasks like video editing where the video filter is disk-bound rather than CPU-bound (as many filters are too CPU intensive for modern CPUs to keep up with drives), RAID 0 can indeed have a good 90%+ performance increase. The same goes for editing of huge music files--if the task is disk bound. Such editing is quite linear in its drive access pattern and allows the combined sustained transfer rate of both drives to spread their wings. The problem is that close to zero other tasks, like gaming or office applications, have an even remotely similar access pattern.
I don't think that Anandtech or StorageReview have ever outright stated that RAID0 is "useless", just that it is useless on Joe User's desktop, and almost every other computer save for the most specific of access patterns.

Desktop hard drive performance today has little to do with access time, [RAID level], or STR (sustained transfer rate), and a lot to do with firmware optimization and the areal density of the drive's platters.

edit: added portion in square brackets [].

 

[Sivar]

Originally posted by: DopeFiend

Originally posted by: scottman
What about 2 x 76gb raid 1? or even 2 36gb r1? Yeah, you basically lose half your storage but your read/write increases dramatically and you've got total redundancy incase of failure.

I tested some identical drives, 2 in r1 and 1 by itself and the gain in my tests was ~40%.

Scott

Yes, of course you did.

RAID-1 does not yield higher read/write speeds. Both the drives are writing the same information at the same time, meaning that you get the speed of the one disk on it's own.

That's why it's called mirroring.

A smart RAID 1 controller can read every other chunk of a file from either drive, similar to RAID 0. This can improve STR.
I am sure that an STR benchmark (like HDtach) on such a controller could show a 40% performance improvement on read speeds, but again this has almost nothing to do with real-world performance most of the time (because STR just doesn't matter for most applications).

Write speeds would, of course, be the same as a single drive or perhaps slightly slower.

Edit: Correction: Er, RAID1 does not double STR (Sustained Transfer Rate), but can improve read performance for a series of queued requests on different cylinders. The queueing can be at the OS level, so does not need a TCQ/NCQ-aware controller.

 

[cyong]

It's somewhat unfair to judge RAID 0 performance like this. Using these sorts of benchmarks to determine RAID 0 performance is like sticking a Radeon X800 XT in your computer and rebooting Windows XP over and over again, claiming it's a waste because it doesn't speed up boot times.

(If you really don't know, a Radeon X800 XT is a recent ATI video card released at $500 that is approx twice as fast as any previous generation video card... I really don't know why I'm explaining this, but I personally live amongst non-computer people😛 Sorry, won't do it again😛)

Quite honestly, RAID 0 does provide some pretty good performance gains, but will not show in any of the tests provided:

1. for you gamers out there (seeing that gamers are the ones who spend lots of $$ on computer equipment) the reason why load times of some games do not see significant benefit from RAID 0 is because:
i. your video card has 128-256 MB of ram and thus even a lousy laptop hard drive could fill that at 20 MB/s in under 10 seconds
ii. other things that are loaded with a level just go into your system ram, which is likewise most likely only 512-1024 MB (which programmers try to optimize usage of anyway by not loading all the textures in the level all at the beginning anyway)
iii. high resolution textures usually require some degree of decompression upon "loading" a level which involves more cpu/graphics card usage than hard drive bandwidth

2. In response to all those business benchmarks using excel/word, tiny i/o processes are not really capped by hard drive bandwidth (as opposed to a few large i/o processes). They should have used some benchmarks involving photoshop or adobe premiere.

3. the main benefit of RAID 0 is load time... this can be seen in a speedier boot-up time (assuming you have your RAID 0 set up right). Unfortunately it's less obvious in program applications since the extra speed is only seen the first time you load the program (since after that the files are still cached in RAM).

My conclusion is that Anandtech has done the best they could with the existing benchmarking programs they had, but their conclusion is extremely biased against RAID 0.

Many (non-gaming) people today view a computer's speed as how fast it can start a program. There is nothing more satisfying than clicking on a button and seeing that startup screen snap up like it was waiting for you. Usually that only happens the second time you start the program (as the files are already in RAM) but RAID 0 brings us one step closer to having that all the time.

It's really late now, and I think I'm losing the point that I wanted to make when I started typing😛

My final conclusion? Every time you hear that "clickety clack" sound from your computer, you could be doing whatever you're doing faster if you had RAID 0. Anytime you don't hear it, RAID 0 will not help you worth a darn because your hard drives aren't even spinning. Period.