Is my RAID 0 working right?

L31mr0d

Ok I've got 2 of these HDD set up in a RAID 0: WD Caviar SE16 250GB

and i've attached an screencap from HDtach, a hard drive speed test.

I can never seem to get higher than 278 MB/s. Surely if my HDDs are rated as 300 MB/s, and they are in a RAID 0 they should be performing at least greater than 300 MB/s? I've done all kinds of defragmenting and disk cleaning and this is the best I get.

WizZard

Yes, it's working right. That's a decent speed.

Rated to 300MB/s, not capable of it though due to a few different factors - spindle speed, where data is placed on disk etc.

djmarkus

Your theoretical bandwidth is 300mb/s, a fast hard drive in real life is closer to 70 mb/s. multiply that by two. 140 mb/s.

Anti

agreed. 270 is fecking fast as it is

po0k

I can never seem to get higher than 278 MB/s. Surely if my HDDs are rated as 300 MB/s, and they are in a RAID 0 they should be performing at least greater than 300 MB/s? I've done all kinds of defragmenting and disk cleaning and this is the best I get.

Jeeeeeeesus.

<snip>

L31mr0d

what speeds would you expect from one 300 MB/s drive not in a RAID? 140 MB/s like djmarkus said. Whats the point in rating them at that if in reality they can't actually perform at that speed. What kind of speeds would you get out of a raptor, would they be about the same?

conzy

Well this is what I score with 1 74gig raptor



So you are doing very well....

Stupid expensive raptor:D I should have went for more storage....

*edit* Hmmmm, 2 raptors in Raid 0 only score 91mb/s as an average read speed, so how the hell are 2 7200rpm drives in Raid0 faster than to 10000rpm drives in Raid0?

Cuddlesworth

Those figures cant be right. I cant even get scsi 15k drives to read and write that fast.

Edit: I see your average read time is 99mbs. Thats pretty good.

Anti

300mbs = burst

SouperComputer

advertised speeds refer to the theoretical bandwidth of the interface. For example, I dont know of any U320 SCSI's that can put out 320MB/s

How the drive actually peforms depends on the logicboard and HDA.

Thats a pretty good figure you have there, be happy with it. Only problem with RAID 0 is the real-world seek times affect peformance more than one would expect.

*edit* Hmmmm, 2 raptors in Raid 0 only score 91mb/s as an average read speed, so how the hell are 2 7200rpm drives in Raid0 faster than to 10000rpm drives in Raid0?

The aureal density of the higher drive means better throughput. Its a synthtic benchmark though, so again, in the real world, the low phyical latency of the Raptors should mean better effective throughput.

Anti

you make SyxPak angry

L31mr0d

The only thing I use to increase drive speed is Diskeeper Pro because it features I-FAAST, that gave me around a 21% increase in access time, which made HDtach go up from approx. 250 MB/s to 278 MB/s. Other than that I don't know what to tell you as to why it is so fast.

gline

anti wrote:

300mbs = burst

thats exactly it, you will never get a sustained rate that high
im sure there is a chart of hdd speeds if you google it

po0k

L31mr0d - do your games load comfortably? Are you happy with how your rig is performing? Did you expect to go from average loading times to lightspeed just by going raid0? The drives are still drives, and are still based largely on the same tech from 20 years ago. There have been small incremental improvements over the years, but they're still magneto-mechanical devices that resemble a vinyl record player.

Synthetic benchmarks mean sfa when it comes to real-world usage. The disk I/O subsystem performance is governed by alot of factors. It is far from simple. Using online benchmarks as a basis for your performance expectations is not as good an idea as it sounds. Your machine and usage patterns are practically unique.

Things like RAID0 exist specifically because hard-drive performance has remained largely stagnant when compared to the advancements in drive capacity.

Have you done any research on storage systems and technologies?

Assuming that because a drive states SATA300 (which is NOT SATAII) compliance it can transfer data that quickly all the time is incredibly näive.

These are a starting point, but are by no means exhaustive explanations, or neccessarily entirely accurate:
http://en.wikipedia.org/wiki/RAID
http://en.wikipedia.org/wiki/SATA

Also, most "Raid" setups you see people with are glorified dumb controllers using software and your CPU's horsepower to run the show.

'Real-World' performance numbers are often going to be a good bit below the benchies, but if you're not roaring at the screen to hurry up, then do you really care what speed it's going at, aside from cock-waggling on t'intarnets?

gline

SyxPak wrote:

, then do you really care what speed it's going at, aside from cock-waggling on t'intarnets?

hehe

true as long is there are no problem with your setup and it loads in a reasonable time, there is no problem

po0k

Apologies if that post came across a little curt, but it really annoys me to see threads like this on this board when a few seconds of intelligent/relevent searching would yield far better and more accurate results.

http://en.wikipedia.org/wiki/SATA#Practical_Benefits

50MByte/sec * 2 drives ~= 100MB/sec, give or take 20-30MB.

There's a very very big distinction between "theoretical interface bandwidth" (read: specs the marketing folks concentrate on) and "real-world throughput".

If you want to actually tweak your setup, perhaps consider adjusting the acoustic setting on the drives (more noise = more 'performance', and more heat), this can probably be set by tools from the drive manufacturers or a plethora of tweaking apps. I've only ever set it with hdparm under *nix.

L31mr0d

SyxPak wrote:

Apologies if that post came across a little curt, but it really annoys me to see threads like this on this board when a few seconds of intelligent/relevent searching would yield far better and more accurate results.

I wasn't looking for data, I was looking for opinions and comparisons with other people on the boards as to what other people are getting, and I KNOW there is a big difference between "theoretical interface bandwidth" and "real world throughput", but if this test was reading as 10 MB/s it would show as a "real world" effect on the performance of my drive. So it is not completely perfunctory, and I do notice a drop in performance when my drive becomes fragmented and my HDtach scores drops.

Your reply is frankly unwarranted, as I am not boasting about my uberperformance, I am simple asking if it is normal with other peoples RAID setups or drive speeds, and plenty of other people post synthetic benchmarks for various hardware testing or overclocking and I dont see you spewing out about "cock-waggling on t'intarnets" with them. How did every other poster in this thread manage to either compare my speed with theres or explain what to expect whilst also being civil?

CivilServant

Random access time, Cpu utilization and average read times are the only stats that matter. 99.6MB/s sounds about right that would make it around 50MB/s for each drive. Burst speed really means nothing because you might get that for a split second and then drop down to average read speeds.

L31mr0d

Ok well I looked into AAM and it turns out it wasn't turned on for my drives. I got a few other benchmark programs. This one (Disk Bench) seems pretty good for doing real world tests on your drive.

I looked up about AAM and turning it fully on seems to have a marginal effect on the access time, but nearly completely removes the noise your HDD makes when performing a seek (mine makes a kinda scratching noise) so I used a program called Roadkils Disk Speed to measure the change in access time with AAM on and with it off.

The first attached image is before enabling AAM, and the second is with it fully on. There is a bit of a performance drop, but I guess i'll live with it as my drives are a lot quieter now.

CivilServant

AAM sounds interesting. What hard drives do you have?

L31mr0d

I've linked to them at the top of this thread, they are western digital Caviars. The utility I used to change the AAM is the "Hitachi Feature Tool", it works with most drives that support AAM.

CivilServant

L31mr0d wrote:

I've linked to them at the top of this thread, they are western digital Caviars. The utility I used to change the AAM is the "Hitachi Feature Tool", it works with most drives that support AAM.

Going to give that a try later. Thx!

DemonOfTheFall

I'm sorry to say it but.....

What the hell is the point in making a power-guzzling, noisy but fast RAID-0 setup with far over double the chance of losing all your data, then turning on AAM to make it into a power-guzzling, not-so-noisy but not any faster RAID-0 setup which still has far over double the chance of losing all your data ?

RAID-0 is a stupid useless piece of e-penis extending crap. Whatever tech sites and forums made people think that it is useful for desktop computers should be shut down...

L31mr0d

Agreed, for my next setup I won't be running a RAID 0. I set it up purely to try it out, and see for myself how it performed, and how it worked. It was more to get experience in setting up a RAID as i'd never had one before. But i've got it now, and I won't be formatting anytime soon, so I'm just going to leave it as is, and there is only a slight performance drop when AAM is enabled, not enough to cancel out the 20%+ performance increase a RAID gives you.

Cuddlesworth

Raid 1 on the other hand..............

astrofool

Well I have a Raid 0, with a third hard disk for backup, and I wouldn't go back to a single HD anymore (like the move to dual core). I've never lost all my data due to a Raid failure, yet I also have everything constantly backed up, so were it to go tits up, there won't be anyone about to say I told you so

I don't however go tweaking the drives with various programs, as most programs are geared toward benchmark improvements rather than real world performance. (I develop on the machine alot).

po0k

L31mr0d wrote:

Your reply is frankly unwarranted, as I am not boasting about my uberperformance

I never said you were. I was infering that you didn't even bother to check what you *should* be getting with 2 bog standard IDE (PATA/SATA are both IDE) drives in a fakeraid setup. How can you worry that it isn't fast enough if you don't even know what to expect from 2disk raid0 array in the first place.

This is what most people (techies/modders/enthusiasts/overclockers) would consider basic knowledge and therefore which, in light of your puzzlement at such "low" speeds, makes this thread redundant.

It grinds my gears, as I feel this forum is in danger of being repeatedly swamped by the same questions instead of interesting new stuff.

But pehaps I am being rather selfish.

I am simple asking if it is normal with other peoples RAID setups or drive speeds, and plenty of other people post synthetic benchmarks for various hardware testing or overclocking and I dont see you spewing out about "cock-waggling on t'intarnets" with them. How did every other poster in this thread manage to either compare my speed with theres or explain what to expect whilst also being civil?

Because they're not all as cranky as I can be.
And I am civil. I am also of an engineering mind. The cock-waggling was actually my attempt at softening my reply somewhat. It was intended to better illustrae my point about realising what's "good enough", instead of constantly fretting over benchmark 'scores'. They're to be used as a guide. Not an absolute metric.

You just happened to post a question which I thought was lazy, especially on a forum such as this.

I'm running a raid0 array myself. It's fast enough, but I place little faith in the reliability of hard-drives. I built it last summer, and as part of the install process I read up on raid0 and digested what info I could find beforehand so I knew the process and what to expect ath the end of it.

Were I in foul humour I would draw parallels between your post and that of someone who couldn't set the clock on their VCR. But I'm not, so have a smiley instead

gline

SyxPak wrote:

I feel this forum is in danger of being repeatedly swamped by the same questions instead of interesting new stuff.

i completley agree, though im not totally without blame in this department myself but the tweaking/modding section seems to be flooded with basic questions which i feel should be in the general computer section

rant over

L31mr0d

What is the use of these forums if I can't ask questions to people that already know the answers. With an engineering mind, like myself, you should know that reinventing the wheel each time something needs to be done is pointless. I had a query, I asked it and got a reply in 5 minutes. Other people already had the answer so there was no point in doing all the research.

Not everybody knows about a RAID and how they should perform, the same with other questions that are asked. Sure there seems to be the same questions being asked but its only because there are new people joining the forums, and when somebody else comes along with a question about what to expect with a RAID, I can now answer them.

Maybe instead of just mouthing off about the state of the forum and telling people to do their own research when you know the answer to their questions you might actually post a thread or two concerning your "interesting new stuff" and stop the forum being swamped.

po0k

Ok.
Mittens on.

L31mr0d wrote:

What is the use of these forums if I can't ask questions to people that already know the answers. With an engineering mind, like myself, you should know that reinventing the wheel each time something needs to be done is pointless.

Irony Instance #1.

I had a query, I asked it and got a reply in 5 minutes. Other people already had the answer so there was no point in doing all the research.

Thiis brings us to Irony Instance #2.
Like typing "SATA" and "RAID" into wikipedia and google? WHich would take all of 5-10secs depending on your typing ability, and would've given you far more information. Why should you even expect a reply when the information is already so incredibly easily available?

Not everybody knows about a RAID and how they should perform, the same with other questions that are asked. Sure there seems to be the same questions being asked but its only because there are new people joining the forums, and when somebody else comes along with a question about what to expect with a RAID, I can now answer them.

Double Whammy of Irony. The same questions are repeatedly asked by people who don't bother to use the internet. Between ARPAnet and eCommerce (shiver), the 'net was primarily a research tool. These "new users", as you put it, can be expected to pretty quickly learn the benefits of searching for the information they desire.
Your good self, however, have been here for a while. Perhaps I expect too much cop-on?

Maybe instead of just mouthing off about the state of the forum and telling people to do their own research when you know the answer to their questions you might actually post a thread or two concerning your "interesting new stuff" and stop the forum being swamped.

Because I still cling to the seemingly vain hope that this small spat may actually have a positive impact and improve the SNR of this board.

There's a reason it's seperate from the Computers / Technology forum. It's for the power user. The geek. The techie. The cut above the cruft.

You may see this as being elitest and arrogant. But maybe you should stop and try and see how things look from the other side of the fence.

And just on a point of information, I did give you information. AAM and the wiki links - which were just a reaffirmation and structured form of what the replies to your question contained in the first place - both of which you could have found quicker than typing out the thread in the first place.

This thread is an example of what's making the internet clog up with crap.
As if MySpace and Bebo weren't enough.

After reading this I hope you finally understand why I'm cranky.

Gone West

erm, to sum up...
Use AAM. Get your hands very dirty. Measure the speeds yourself. Forget crappy online benchmarks. Also grab yourself a proper hardware RAID card.

Theres a lot more satisfaction to be gotten from learning the things you need/want to know yourself, rather than asking people just for the snippets of information you need right now.
Linking you to wikipedia pages is akin to "teach a man to fish..."