* * * Requesting all RAID experts * * *

[silicon demon]

yours truly here,

i posted the following, verbatim, in the motherboards forum. only one person responded (thanx, Zap). i would like more input. hope you guys and gals can help:


"(2) questions. Isn't it true that the RAID 2 configuration has gone the way of the dodo bird? ....meaning RAID 2 is no longer utilized? if so, WHY?'

To my knowledge, RAID 2 allows for extremely high transfer rates, and uses a Hamming code for error correction. It is similar to RAID 0 striping, except that it stripes at the bit level rather than the block level. I'd be very interested in using RAID 2 if it's still an industry standard. '

Second question: Does anyone out there have any experience with using RAID 0 across more than 2 disks? (3-disk....4-disk RAID 0 configuration?). I understand that the risk of system failure goes up if a disk goes bad, but how much gain in performance can you get by using a multi-disk (i.e., more than 2 disks) RAID 0 config? '

I guess i don't care about disk failure "too" much since i Ghost (Norton) quite regularly. "

 

[Fullmetal Chocobo]

RAID 2 , to my understanding, does not exist in the consumer market, and mainly exists on the propietary levels.

I've setup a RAID 0 array using 5 Hitachi 250gb SATA hds. I benchmarked it, and then converted the array to a 5 disk RAID 5 array, and the performance was about the same. And this is on a PCI-X 64-bit, 133-MHz bus card, so bottlenecks were likely not an issue. RAID 0 is good for specific applications on 2 disks. Those applications include when there is another disk for storage and backup duties (IE two RAID 0 disks, and a third for storage). After you start having 3 disks in the RAID array, moving to RAID 5 is more effective.

 

[Vegito]

http://pics.bbzzdd.com/users/forcesho/raid0-1.jpg
http://pics.bbzzdd.com/users/forcesho/raid0-2.jpg
http://pics.bbzzdd.com/users/forcesho/raid0-3.jpg
http://pics.bbzzdd.com/users/forcesho/raid0-4.jpg

 

[silicon demon]

thanx, bro..

i've read that: "RAID 5 writes are expensive in terms of disk operations and traffic between the disks and the controller"......[and that]..."RAID 5 implementations suffer from poor performance when faced with a workload which includes many writes which are smaller than the capacity of a single stripe"......because it uses block-level striping.

from what i understand (i am a novice), RAID 5's a type of disk striping--like RAID 0--except that it also writes a parity stripe to help out if a disk goes bad. it would seem to me that that parity stripe alone increases write time, and would eat up disk space (or, at least, MFT allocation).

And i don't understand why RAID 5 is more effective than RAID 0 if you use 3 disks or more?

 

[Madwand1]

RAID 5 basically costs you one drive for parity, and that cost gets proportionately less as you increase drives. It also typically costs you performance during writing. However, if you have a large number of drives connected, then you have a ton of data capacity, and the risk you take with all this data going poof, and backup time, etc., is high. Hence RAID 5 is generally perferred. RAID 0 must be as fast or faster, but it doesn't deserve the "R" in RAID.

 

[silicon demon]

i understand your statment about the "R".

To my understanding, the pros you get in a basic (i.e., 1-5) RAID setup are: 1.) redundancy; or 2.) faster read times.

i thought that having a 2-disk RAID 0 config. basically halved your disk read time, and that having a 3-disk array cut your time by one-third, etc., etc. Is my understanding of RAID correct?

(I do realize that RAID 0 offers no redundancy.) And doesn't that "parity" stripe in RAID 5 refer to redundancy? (yes? no?)

What i really want to know is can you get blazing disk access by building a RAID 0 config across 3 disks or more.

 

[Vegito]

Originally posted by: silicon demon
What i really want to know is can you get blazing disk access by building a RAID 0 config across 3 disks or more.

Yes but its all minimal gains. I have 7 drive array in RAID 0, I done from 2 to 7 RAID 0s and each time it goes up by a few percent.. not like triple or quadruple...

 

[ribbon13]

disc access times do not decrease unless you have a HBA with intellingent read ahead that happened to have that block in RAM. in fact, on average they increase. the transfer rates however can increase significantly.

 

[Matthias99]

Originally posted by: ribbon13
disc access times do not decrease unless you have a HBA with intellingent read ahead that happened to have that block in RAM. in fact, on average they increase. the transfer rates however can increase significantly.

:thumbsup:

For read-heavy workloads, RAID5 and RAID0 will perform similarly -- and as the number of drives in the array increases, the read performance gets closer and closer (but the chance of having a failure in either array goes up). Since both are striped arrays, STR increases almost linearly with the number of drives (to the limit of the controller, at least).

For write-heavy workloads, RAID5 will perform worse than RAID0. For a write of less than a full stripe, you have to: read in the rest of the stripe from the drive you are writing to, then read in the corresponding parity stripe, then calculate the new parity stripe's contents, then write both stripes back to disk (for a full-stripe write, you only have to read in the parity stripe). With RAID0, you just write the data where it is supposed to go, no extra reads or calculations required. However, unless you have a VERY write-heavy workload, the reliability improvement in RAID5 is usually deemed worthwhile.

If you have a workload that is heavy on writes and needs protection, but also needs high STR, you should look at RAID01/10 (also sometimes called "0+1" or "1+0"). The write penalties are not as bad as a RAID5, and STR is high -- but you sacrifice half your disk space like in a RAID1 setup.

 

[Madwand1]

Originally posted by: silicon demon
i understand your statment about the "R".

To my understanding, the pros you get in a basic (i.e., 1-5) RAID setup are: 1.) redundancy; or 2.) faster read times.

i thought that having a 2-disk RAID 0 config. basically halved your disk read time, and that having a 3-disk array cut your time by one-third, etc., etc. Is my understanding of RAID correct?

(I do realize that RAID 0 offers no redundancy.) And doesn't that "parity" stripe in RAID 5 refer to redundancy? (yes? no?)

What i really want to know is can you get blazing disk access by building a RAID 0 config across 3 disks or more.

Yes -- the parity redundancy IMO is the coolest idea in RAID -- redundancy with just 1 additional drive! OK, striping is cool too in general. Just for the sake of argument, I could also argue that RAID 1 isn't really "RAID" in the original sense, and it's certainly no intellectual achievement. However, I digress..

Sadly, RAID is no Panacea. (I don't even think it's Greek.) If everything was so simple, and if the performance gains were so universally great, I'd expect to see RAID mentioned on every hard drive manufacturer's front page. My small sampling showed it on none of them.

The following shows some actual scaling benchmarks with arrays of various sizes. It's RAID5, but I think you could roughly convert them to RAID 0 equivalents by removing one drive. Of course, that's not entirely correct, and only a rough guess. Reality usually bites theory.

http://www.tomshardware.com/2004/07/23/...sts_with_up_to_eight_drives/index.html

 

[boomerang]

So refreshing to read knowledgeable posts about RAID. I'm so tired of reading posts where a legitimate RAID question is answered with a blanket statement that RAID is "stupid".

Thumbs up to everyone in this thread. :thumbsup:

 

[alpha88]

Please remember this as well:

There are two types of speed, Low Latency and High Bandwidth.

RAID (in general), increases Bandwidth, but compromises (raises) Latency.

So for bandwidth intensive tasks, like working with video, RAID is great.

For things like gaming, latency is often more important, so RAID 0 will hurt your performance.

 

[silicon demon]

much thanx to everyone who responded to this thread.

i'm only posting here because because i'm truly trying to learn, and to fully exploit (in a good sense) the beauty of forum communities--and their collective wealth of knowledge-- in order to decrease my learning curve.

my questions are never ego-centric. and i don't debate simply to seek an opportunity to sound intellectual. i don't mind telling everybody i am the true neophyte....such a novice at all this!

some of the responses were just a tad over my head, but that's okay. that raises the bar for me a little. also, i hope you all can continue to be patient, because i AM going to ask and ask and ask until i fully understand...and i'm liable to run a thread 'til AT's servers start smokin.'

anytime you guys see me post at anadtech, i fully welcome your feedback, however critical, and i appreciate it.

 

[Arcanedeath]

If you really want to know alot about raid check out www.Storagereview.com it has pretty much all you want to know about the various raid levels

 

[Tostada]

Originally posted by: Vegito
http://pics.bbzzdd.com/users/forcesho/raid0-1.jpg
http://pics.bbzzdd.com/users/forcesho/raid0-2.jpg
http://pics.bbzzdd.com/users/forcesho/raid0-3.jpg
http://pics.bbzzdd.com/users/forcesho/raid0-4.jpg

That system must be really good at booting Windows!

What's the matter? Are you embarrassed to run any other benchmarks besides the near-useless Sequential Read Speed?

 

[Tostada]

Originally posted by: boomerang
So refreshing to read knowledgeable posts about RAID. I'm so tired of reading posts where a legitimate RAID question is answered with a blanket statement that RAID is "stupid".

Thumbs up to everyone in this thread. :thumbsup:

Plenty people around here are knowledgable about RAID arrays, and I think most of them will still tell you it's not worth the trouble in many cases except for the redundancy aspect. It's not necessarily "stupid," but it is largely pointless for most users.

I challenge anybody to refute the guys at StorageReview:

The point? Dont assume RAID 0 offers increased performance for all or even most applications... and dont assume that transfer rates reflect application-level performance.

 

[Vegito]

Originally posted by: Tostada

Originally posted by: Vegito
http://pics.bbzzdd.com/users/forcesho/raid0-1.jpg
http://pics.bbzzdd.com/users/forcesho/raid0-2.jpg
http://pics.bbzzdd.com/users/forcesho/raid0-3.jpg
http://pics.bbzzdd.com/users/forcesho/raid0-4.jpg

That system must be really good at booting Windows!

What's the matter? Are you embarrassed to run any other benchmarks besides the near-useless Sequential Read Speed?

I did that for demo, i dont even use it as raid 0, i have a large raid system and i do demo benches for people that likes to know what config may look like

http://pics.bbzzdd.com/users/forcesho/Server01.jpg
http://pics.bbzzdd.com/users/forcesho/Server02.jpg

thats my raid system and its not some raptor 150s, its mostly 15k drives

 

[Tostada]

Originally posted by: Vegito
I did that for demo, i dont even use it as raid 0, i have a large raid system and i do demo benches for people that likes to know what config may look like

http://pics.bbzzdd.com/users/forcesho/Server01.jpg
http://pics.bbzzdd.com/users/forcesho/Server02.jpg

thats my raid system and its not some raptor 150s, its mostly 15k drives

Unfortunately, that has nothing to do with this thread. The guy asked a question about RAID 2 and you respond with a post devoid of anything but JPGs of your sequential transfer rate. Give me a break.

Seriously. You never even said what kind of RAID it was or listed what any of the drives were. You're certainly not contributing, and you're doing a pretty poor job of bragging.

 

[Vegito]

Originally posted by: Tostada

Originally posted by: Vegito
I did that for demo, i dont even use it as raid 0, i have a large raid system and i do demo benches for people that likes to know what config may look like

http://pics.bbzzdd.com/users/forcesho/Server01.jpg
http://pics.bbzzdd.com/users/forcesho/Server02.jpg

thats my raid system and its not some raptor 150s, its mostly 15k drives

Unfortunately, that has nothing to do with this thread. The guy asked a question about RAID 2 and you respond with a post devoid of anything but JPGs of your sequential transfer rate. Give me a break.

Seriously. You never even said what kind of RAID it was or listed what any of the drives were. You're certainly not contributing, and you're doing a pretty poor job of bragging.

Second question: Does anyone out there have any experience with using RAID 0 across more than 2 disks? (3-disk....4-disk RAID 0 configuration?). I understand that the risk of system failure goes up if a disk goes bad, but how much gain in performance can you get by using a multi-disk (i.e., more than 2 disks) RAID 0 config? '

He say second question about raid 0, the pics clearly shows RAID 0, even the file name says RAID 0

 

[Vegito]

quote:

--------------------------------------------------------------------------------
Originally posted by: silicon demon
What i really want to know is can you get blazing disk access by building a RAID 0 config across 3 disks or more.

--------------------------------------------------------------------------------
Yes but its all minimal gains. I have 7 drive array in RAID 0, I done from 2 to 7 RAID 0s and each time it goes up by a few percent.. not like triple or quadruple...



and I answer his question how from 2 drive up to 7, you get minimal gains.. so what if i got these crap sitting around.. i gave most of it away to people here for FREE so bite me

 

[Vegito]

& BTW

http://www.anandtech.com/storage/showdoc.aspx?i=2682&p=3

With two of these 160GB 7200.9 units in a RAID-0 setup, we see that the operations per second increase about 55%. We performed this test as a brief look at how RAID affects performance...and to prove to the skeptics that setting up two or more drives in RAID-0 at least will increase the performance. Here, we see that happening. Let's see if this trend continues.

I dont write, i'll only link it and you'll have to make your on judgement call on RAID.

 

[Tostada]

Originally posted by: Vegito
& BTW

http://www.anandtech.com/storage/showdoc.aspx?i=2682&p=3

With two of these 160GB 7200.9 units in a RAID-0 setup, we see that the operations per second increase about 55%. We performed this test as a brief look at how RAID affects performance...and to prove to the skeptics that setting up two or more drives in RAID-0 at least will increase the performance. Here, we see that happening. Let's see if this trend continues.

I dont write, i'll only link it and you'll have to make your on judgement call on RAID.

Yeah, because AT's hard drive articles are so much more legitimate than StorageReview!

Forget the guys who have been benchmarking this stuff for years -- Purav Sanghani is on the case!

Take a long look at those real world tests in the AT article. They've chosen benchmarks of zipping files and copying folders. Those are the things that should get a big boost from RAID 0, but only the copy folder tests do, and certainly not double. Of course if you were interested in just copying folders, you would get better performance by splitting the RAID and copying between separate physical drives.

Obviously this is no news to you. You know that RAID doesn't help much as far as performance in the majority of things. That's why the benchmarks you're showing are for sequential transfer.

 

[Chaotic42]

Here's a lot of info on RAID levels if anyone is interested:

Link

 

[silicon demon]

okay. i guess i didn't know what a behemoth understanding RAID would be. i've just spent the last couple of hours reviewing everyone's links/ references, and it's possible that i'm even more cloudy on the subject now than i was initially.

Tostada...you played devil's advocate so i checked your link out first: It was a generally pessimistic review of RAID 0. one quote from the article was:

There are certain uncommon situations where RAID 0 can significantly improve system performance. For example, editing of large audio or video files is sometimes limited by the maximum sequential transfer rate of the hard drives, but it is far more common for the processor to be limiting factor. Generally, if you frequently make simple edits to large media files, RAID 0 can potentially improve your productivity.

i found this interesting because my primary application is pro audio (saying this might help everyone tailor their responses for me!). I don't know how to say in technical terms exactly what the computer's tasks are, but i might play several audio files simultaneously, with many instances of a "reverb" effect (multiple 'vsts' in audio lingo) in use at the same time.

Vegito....conversely your link was a generally optimistic view of RAID 0. And it was an AT article. It was fairly convincing but in direct contrast to Tostada's link....

Madwand 1...your link covered RAID 5 instead of 0: It was a generally optimistic review. what i gathered, in summation, from that article is that with RAID 5, more disks offer better performance. But....[Question: Are those RAID 5 performance marks better than a single drive without RAID?]

i only want to use RAID (whichever type) if it will give me better performance than a single drive. i don't care about redundancy; i care about speed. I back-up to external drives regularly so data security isn't an issue to me.

i have a 2-disk RAID 0 setup now, but i didn't configure it, and was even more of a novice when i first got this computer--i didn't even know what i had. What i do like about my present setup is that I can combine drives of various oddball sizes (capacities) into one array, then create tailor-sized logical partitions.

So guys..............i'm about to build my first computer (myself! ), and i have not taken this matter lightly. i've been researching for some time now; but it's disappointing to me that i can't even find personal resolution on whether to use RAID or not, and what type!

As i said, digital music creation will be my primary application. i assume that parallel processing will be more important than sequential processing. i also assume that access times and transfer times will be more important than write times. {Please correct any of my assumptions if they are wrong.} And i'm still very open to more input.

(P.S. I read everyone's reference material. Chaotic 42, thanx for your link; and Ribbon 13, your rig sounds like it's awesome.)