RAID HD benchmark in sandra

[Kaliman]

i just benchmarked my RAID setup with two 60 gb 60 gpx IBM HDs and it came out lower than what it was being compared to. The compared score was around 36000 and mine was about 34300 how can i fix this?

 

[FatBurger]

My Maxtors use 'Acoustic Management' to make them quieter (and slower). I don't think IBMs have anything like that, but it's a thought.

Use HDTach (don't know where to find it). It's supposed to be more accurate for those sorts of things. Also, Promise RAID controllers work better on IBM drives than HighPoint. I don't know which you have.



<font color=blue>Quarter pounder inside</font color=blue>
<font color=red>Change the Sig of the Week!!!</font color=red>

 

[Matisaro]

Make sure you are using the smallest stripe size, smaller is faster, but with more cpu overhead.

~Matisaro~
"Friends don't let friends buy Pentiums"
~Tbird1.3@1.55~

 

[CALV]

You can get hdtach from <A HREF="http://www.calvsplace.cwc.net" target="_new">My site </A> in the download bit, I DO NOT hava a crack for it on there, but the unreg'd version will perform read testing on windows 98/me, it WONT perform write testing, and wont work on 2k. (who cares about write testing???!!), However if you really, really need to "register it" to get all the features then mail me calv@fuckyou.co.uk , or just do a search for a c**ck


Next time you wave - use all your fingers

 

[Solstice]

Stripe size?
If you mean Block size smaller is slower

 

[MeTaLrOcKeR]

Just a question for you CALV....

How did you manage to get an @fuckyou e-mail address ???!???

-MeTaL RoCkEr
My <font color=red> Z28 </font color=red> can take your <font color=blue> P4 </font color=blue> off the line!

 

[MeTaLrOcKeR]

Post deleted by MeTaLrOcKeR

 

[CALV]

just go to www.fuckyou.co.uk and sign up for one, fuckyou aint my isp! its web based but web based anon email is sometimes handy, besides, this one has an entertainment value too!


Next time you wave - use all your fingers

 

[MeTaLrOcKeR]

Thats exactly why i wanted to know.....the humor in the address is just to funny...LoL....
Imagine putting an @fuckyou e-mail address on your resume ?
Oh man....LoL

-MeTaL RoCkEr
My <font color=red> Z28 </font color=red> can take your <font color=blue> P4 </font color=blue> off the line!

 

[Matisaro]

No, stripe size, block size, smaller is faster.(according to storagereview.com.) A large block size is inefficient diving up the data(ie you have to wait for hdd a to write 64k of data before sending the next bit to hdd B) however, very small stripe sizes lose this bonus in increased latencys on the read. 8kb offers the best performance/latency ratio.

~Matisaro~
"Friends don't let friends buy Pentiums"
~Tbird1.3@1.55~

 

[FatBurger]

Then why do they use a stripe size of 64k in their benchmarks?



<font color=blue>Quarter pounder inside</font color=blue>
<font color=red>Change the Sig of the Week!!!</font color=red>

 

[Matisaro]

http://www.pcguide.com/ref/hdd/perf/raid/concepts/perfStripe-c.html

some snippets

"Decreasing Stripe Size: As stripe size is decreased, files are broken into smaller and smaller pieces. This increases the number of drives that an average file will use to hold all the blocks containing the data of that file, theoretically increasing transfer performance, but decreasing positioning performance. "

"Increasing Stripe Size: Increasing the stripe size of the array does the opposite of decreasing it, of course. Fewer drives are required to store files of a given size, so transfer performance decreases."


just got fatburgers reply in my email, I have no idea why they use 64kb, however, my results with 2x 45gig 75gxp drives are about 10% faster than theirs with similar hardware. (I use 8k stripes)

~Matisaro~
"Friends don't let friends buy Pentiums"
~Tbird1.3@1.55~

 

[Matisaro]

"Note: The improvement in positioning performance that results from increasing stripe size to allow multiple parallel accesses to different disks in the array depends entirely on the controller's smarts (as do a lot of other things in RAID). For example, some controllers are designed to not do any writes to a striped array until they have enough data to fill an entire stripe across all the disks in the array. Clearly, this controller will not improve positioning performance as much as one that doesn't have this limitation. Also, striping with parity often requires extra reads and writes to maintain the integrity of the parity information, as described here."

the highpoint controler is as vanilla raid as you can get, it has no special optimisations, so a smaller stripe size will outperform a larger one.

~Matisaro~
"Friends don't let friends buy Pentiums"
~Tbird1.3@1.55~

 

[Ncogneto]

Larger stripe sizes are better at handling large transfers Ie if you are continuosly transfering very large files you want a larger stripe, while smaller stripes are quicker when tranfering small chunks. Once again it is all application dependent on what you are doing.

Video editing?? Ha, I don't even own a camera!

 

[FatBurger]

Ncogneto is right. Look <A HREF="http://www.dell.com/us/en/esg/topics/power_ps4q00-jordan.htm" target="_new">here</A>.

ROFLMAO
Me and Matisaro just provided the same link.



<font color=blue>Quarter pounder inside</font color=blue>
<font color=red>Change the Sig of the Week!!!</font color=red><P ID="edit"><FONT SIZE=-1><EM>Edited by FatBurger on 08/24/01 10:44 AM.</EM></FONT></P>

 

[Matisaro]

I just read anandtech's raid review, it appears they found that 64k was the best average stripe size, but it IS application dependant. however it looks like a 8k stripe size beat 128k most of them time. It is strange how different tech sites say different things. (however all the results I looked at were with hardware raid cards, I will look for highpoint raid results tonight.

~Matisaro~
"Friends don't let friends buy Pentiums"
~Tbird1.3@1.55~

 

[Kelledin]

One thing I've found from benchmarking with both Sandra and Winbench is that they seem to produce strange (and apparently inaccurate) results with RAID arrays. Storage Review found this out about WinBench when they first tried benchmarking RAID arrays and switched to IOmeter instead.

Kelledin

"/join #hackerz. See the Web. DoS interesting people."

 

[CALV]

I have to say, that sandra seems very inaccurate in a lot of ways, though earlier versions I used to use a lot (back in the days of using quake 2 timedemo and a pair of voodoo 2's, however recently I dont believe much that sandra tells me, it benchmarks my ram (currently running at 140) less than pc 100 and also says my hard drive is slower than a 5200 8 gig (its a 40 gig 7200 ibm), AND it tells me my sc read speed is 23x max (erm its a plextor ultraplex40 scsi, and every other benchmark says MINIMUM 17x, max 40x).

Next time you wave - use all your fingers

 

[Oni]

If I wanted to do video capture with a RAID-0 setup would I want 64k or 128k stripes? and also where can I get the newest drivers for a high point on board raid controller?

PS: I've got two IBM 40 gig 60GXP drives in the mail right now for use with my Epox 8K7A+ motherboard.