If your thinking of using Raid 0

[Odos]

Hi all, a long and boring post if your not interested in Raid 0 you should stop reading now 

If your still here I'll begin. For years I have always used Raid 0 on my home machine ( general usage ) because I believed it was faster. I know all the arguments about data safety etc but I keep all important stuff backed up anyway so all I was after was speed. Over time I've read all the discussions on the different forums about how raid makes no difference to the everyday user but there was never any hard evidence. Anyway this last week I've had some free time so decided to do some rough and ready real life testing, and some of the results surprised me.

Before I start let me just say I'm making no claims or assertions for the use of raid with regards to anyone else. There are too many variables involved to be able to make any such claim. The results below are only relevant to me and my computer/configuration.

First of all the system I used for testing, well the relevant bits anyway.

Asus A8N SLI Premium motherboard
AMD 64x2 4800 cpu
4 gig of DDR400 memory
XP Home Edition for the testing
1 x 200gig Sata II Maxtor drive not used during tests
2 x 400gig Sata II Samsung drives in Raid 0 on M/B's Nvidia controller ( Sata II )
2 x 500gig Sata II Hitachi drives in Raid 0 on M/B's SIL3114 controller ( Sata I )

In order to have a standard test I created a test folder which was 3.59gig in total size and contained 2 multi-media vob files each 1 gig in size and a folder containing random files from one of my application drives. This second folder contained 23,634 files of varying sizes from just a few bytes to a few hundred meg. My reasoning being that in the real world the files you access in daily usage are of all sizes.

I created a bare winXP OS image which had nothing installed except the drivers required to run the tests. The drives/arrays were formatted between each test and a single 10gig partion on each drive/array created in the same mid platter area each time. The single 200gig drive that was not used during testing was used to hold the test folder between formats.

The tests consisted of timing with a stop watch how long it took to copy the test folder from one array to the other and then back again. This was done twice each time so for each configuration I had 4 result times which I averaged out to get the results below.

The Results -- all times in seconds

          Single Drives No Raid
   Fat32                 NTFS
Cluster Size         Cluster Size
32K = 105             32K = 110
16K = 100             16K = 105
8K = 97               8K = 105
4K = 97               4K = 110

           Raid 0 8K Stripe Size
   Fat32                 NTFS
Cluster Size         Cluster Size
32K = 125             32K = 128
16K = 121             16K = 124
8K = 120              8K = 137
4K = 114              4K = 140

           Raid 0 16K Stripe Size
   Fat32                 NTFS
Cluster Size         Cluster Size
32K = 90              32K = 97
16K = 84              16K = 90
8K = 80               8K = 89
4K = 80               4K = 89

           Raid 0 32K Stripe Size
   Fat32                 NTFS
Cluster Size         Cluster Size
32K = 90              32K = 92
16K = 84              16K = 90
8K = 80               8K = 89
4K = 79               4K = 89

           Raid 0 64K Stripe Size
   Fat32                 NTFS
Cluster Size         Cluster Size
32K = 94              32K = 97
16K = 87              16K = 93
8K = 83               8K = 89
4K = 83               4K = 89


The conclusion for me is that raid does make a difference in real life, but if you get the stripe and cluster size combination wrong then the difference is a negative and not positive. The other surprise ( for me at least ) was the difference in speed the file system itself made. In every test Fat32 beat NTFS easily.

As I stated earlier I'm not saying anyone should use raid but if anyone is thinking about it then this info might be useful. If it does nothing else it should highlight that you should do your own tests to find the optimal settings for your particular system.

Cheers
      Tony

[pacifika]

See these two excellent articles from CodingHorror:
You Want a 10000 RPM Boot Drive
Desktop RAID: Oversold?

[mrapoc]

I have just raided my 320gb seagate 7200.10 (32mb cache total) together for a whopping 640gb storage (200gb backup)

I must say its a lot faster (although a pain to setup) - I chose the stripe size best for a multimedia pc (largest iirc) and its brill

[Lance]

Thanks for this Tony, its a really helpful post! Must have been time consuming to do and i think you deserve a karma for all your efforts!

By the way, thats a monster of a system you have!

[Odos]

Thanks for the Karma Lance and yep, took me two days in total ( taking out time for taking the other half shopping  ).

Cheers
     Tony

[Rik]

Hi Tony

Great post, very interesting results. I've had the same drives configured as RAID0 and separate, discrete, units on the same machine and, as your tests suggest, I found little or no benefit from the RAID.

NTFS is always going to be slower than FAT, it's a far more robust and secure FS. On balance, though, I prefer a slight loss in speed for the benefits it offers over FAT.

[siege2]

on my laptop which i put in a bigger Hard drive 120 GB on ide I get read average read speeds off 27.3 MBS on my PC I get 115 average MBS in Raid 0 SATA maxtors *2
my collegue thinks it's very fast using that config now for 2 years now.
PS a single PC drive averages out 55 MBS oh and using HDTACH its free for home use...

my point :-)

[Rik]

OTOH, your laptop is probably a 5400rpm drive, maybe even a 4200?

[siege2]

I know its a 7200 rpm I put it in but it all depends on the interface mostly !
good morning Rik

[Rik]

'Morning Siege.

I assume it's IDE on the lappy?

[siege2]

yep it is..

[Rik]

You know, I've always wondered why SCSI wasn't more successful as an interface. It had so much potential, but never took off for the average user (completely OT, I know...).