New Posts  All Forums:Forum Nav:

Raidin' 0 Striping

post #1 of 14
Thread Starter 
Can anyone tell me about the increase in R/W speeds (HD access speeds) when going to a RAID 0 array.

This is why I am asking. I currently have a 40G 5,400 rpm drive installed, and I was tinkering with the idea of putting in my 2 40G 4,500 rpm drives if I was going to get a decent performance increase... I have these two (exact) same drives sitting around and wanted to see if it was worth it (i.e. increasing performance without shelling out $s).

post #2 of 14
The way I've had it explained to me, RAID 0 splits all the information going to and from the drives equally between them, so each is only having to read/write half of the information. This, in effect, would double working RPMs, since you have two 5400 rpm drives writing half of the information each, which would be like one single 10,800 rpm drive.

The downfall to this is, of course, that if something should go wrong with one of the drives, both of them go down as a result, since each is holding a part of your system.

-Hal
post #3 of 14
Hal, In theory that is about correct; however, these drives only use one channel. This results in a speed increase of about 50 - 60% rather than the "whole banana". It's worth a try as it is much faster than the single drive. Hal's cautions is entirely correct.
post #4 of 14
If you want some hard numbers, I did the testing with my 8890 when switching from RAID 1 to RAID 0 (I had issues, so I had an opportunity to test with a single drive). Here's the numbers from HDTach:

2x Toshiba MK6022GAX 60GB 5400RPM

Without RAID 0 - Single Disk (not even RAID 1)
Burst: 89.0MB/s
Sequential: 31MB/s to 17MB/s
Average Read: 25.6MB/s
Random Access: 18.0ms
CPU: 2%

With RAID 0 (Stripe)
Burst: 102.6MB/s
Sequential Read: 64MB/s to 35MB/s
Average Read: 50.3MB/s
Random Access: 20.0ms
CPU: 4%
post #5 of 14
Thread Starter 
I too would like to make a before and after test to post results...

How did (do) you test & log the results?
post #6 of 14
How'd I do the tests? Here's the process I used:

Install ONLY XP Pro w/SP2 + all drivers on Sager 8890 without any RAID.
Install HDTach. ( available free from: http://www.simplisoftware.com/Public...request=HdTach )
Run HDTach with the long bench and quick bench modes. (Results were nearly identical)
Scribble down results.

then....
Install ONLY XP Pro w/SP2 + all drivers on Sager 8890 with RAID 0 (striping)
Install HDTach.
Run HDTach with the long bench and quick bench modes.
Scribble down results.
Post results!
post #7 of 14
Thanks for doing the test and posting the results dprischa.
There has been a lot of bs flying around when it comes to the issue of what RAID 0 actually gives you in terms of speed improvement.

The defence rests.
post #8 of 14
I am hoping to do some RAID striping on my older 8886 to see if I can squeeze a little better performance out of it, but I'm not sure how similar the HD's need to be...do they have to be identical (same model numbers) or just the same size and speed (or even just the same size, etc)??

And RAID is set simply through the BIOS right? Or is there some messing with jumpers?
post #9 of 14
The easiest answer to your question is: No, they do not have to be identical, or even similar, drives.

The stripe will be the size of the smallest disk * the number of disks in the stripe. The "extra space" on the larger disks will be lost, totally unusable.

Example: 20G + 40G HDs = 40G stripe + 20G unusable.

The performance issue will be the interesting portion because you have to take into account spindle speed, access times, and HD cache of each drive. Sometime you may get lucky and the differences won't affect each other much. I believe that no matter what the differences are in the specs, you WILL see a fairly significant boost in performance. I saw nearly 100% with mine. My sequential read times doubled, but my random access times increased by 2ms. Ignoring all the numbers, my 8890 "feels" and is much faster with the stripe.

The best solution is 2 identical disks, but any combination will work.

I have the 8890, so I'm not sure about the settings on the 8886, but I had to set my jumpers under the keyboard to enable RAID, then I had to reset the BIOS, then create the stripe in the RAID utility (after BIOS). It should be in your manual, which for the 8890 was written quite well and easy to follow.

Oh, and remember to do regular backups once you use the stripe because the probability of losing a disk in a 2 disk stripe is not twice as likely, it is 4x more likely...
post #10 of 14
I see in your sig that you have "RAID 1", I've seen a lot of people talking about "RAID 0"...what is the difference? Also you mention the RAID utility..does that come with the laptop or do you need a separate program?
post #11 of 14
RAID-0 is striping. Make two drives look as one and divide data between the two drives.

RAID-1 is imaging. Make two drives look as one with a copy of all data on the first drive on the second drive.

If you have 2-60GB drives. RAID-0 results in 1-120GB Drive; while RAID-1 results in 1-60GB Drive (imaged).

There are some performance differences; however, I won't go there.
post #12 of 14
Unforntunately Tryanous only the 8890 and later models (8790 and 9860) have the RAID controller chip built into them. I just checked the 8886/8887 service and users manuals and the chip and RAID enable switches are missing .

The driver for the RAID or ATA setup was supplied as two floppies with the 8890. You install them when Windows requests additional drivers during install.

On the issue of reliability, I quote:
Quote:
Originally Posted by dprischa
Oh, and remember to do regular backups once you use the stripe because the probability of losing a disk in a 2 disk stripe is not twice as likely, it is 4x more likely...
I know it might "feel" like it is 4x more likely but in reality the figures are less than that. For a full mathematical analysis of MTBF etc etc on the RAID setup you can read it here.
post #13 of 14
I'm not sure I agree with either of you....

Aussie, in the post you reference, you state that the probability of no failure for one drive over 2.5 years is 97.83%; for two drives it's 95.71%. No issues with your calculations.

The failure rate on one drive is therefore 2.17% and for two drives 4.29%. The probability of failure goes up by 97.7%. The risk of having a failure does, in fact almost double.

To say it differently, if you have 100 non-raid drives you would expect a touch over 2 failures in 2.5 years. If you had 100 two-drive RAID 0 setups, you's expect just over 4 failures in 2.5 years.

The risk of having a failure does, in fact, almost double because it's gone from 2% to 4%.
post #14 of 14
That is a common misconception. The issue is the probablility of a failure which is different to risk.

Just because the probability of failure goes from 2% to 4% your risk does not double.
New Posts  All Forums:Forum Nav:
  Return Home