RAID Question

[TSR105]

Hello

I can get to the speed of 133 megabytes per second (MB/s)
This is my speed result

Write speed

********************29-6-2002 1:46:02********************
Disk: d: [video] Write speed: 17048 kb/sec
Write mode: 65536 blocks - Test file size: 500 MB
************************************************** ************

Read speed

********************29-6-2002 1:46:40********************
Disk: d: [video] Read speed: 36247 kb/sec
Read mode: 65536 blocks - Test file size: 500 MB
************************************************** ************

Can sombody please help me




************************************************** ************
Mainboard Asus P4S533
CPU INTEL 2.0 GHZ
PCI Controller: Promise Ultra 133 TX2
2x Matrox 80 GB D740X 7200 RPM ATA133
VGA: MATROX MARVEL G450 eTV
RAM: 512 MB DDR333


THX from TSR105

[Gurm]

36MB/sec. sustained read is pretty good.

There is no such thing as a drive, OR RAID ARRAY, that will really deliver ATA-133. Just doesn't exist.

- Gurm

[Reverend Maynard]

The only program I have for measuring HDD speed is HDTach. Dunno how accurate it is, but it is nice to compare and tweak my home systems. I ran tests on both my arrays and here is what I got:

AMI raid 0 (built on the mobo) with 2 IBM 40 gig 7200 drives gave me an average transfer rate of 41 MB/sec

Promise FastTrak 66 (modded) with 2 Maxtor Di740 40 gig drives gave me a 38 MB/sec average transfer.

Again my testing program differs from yours, but for rough estimates sake your array seems about right.

[TSR105]

It is still too slow !

Look here what the card can do

http://www.tecchannel.de/hardware/817/11.html

http://www.viaarena.com/?PageID=66#intro

THX too all

[Gurm]

That's buffered speed. Honestly, you aren't gonna get that. In truth, with a Promise card and two high-speed drives, you ought to be sustaining closer to 50. Maybe it needs tweaking or newer drivers or a newer BIOS. But your speeds aren't bad, honestly. Those DRIVES only deliver 25MB/sec. sustained, so your theoreticla maximum sustained speed is 50MB/sec. - but RAID0 isn't just twice as fast, although it CAN be twice as fast. Are you following me? There is no way in hell you can get 133MB/sec. because the drives deliver less than half of that.

- Gurm

[TSR105]

This will not make me happy

hi Gurm

This will not make me happy ,because i want to capture video with virtualDub in res 480x576 and 704x576 but it is raining dropped frames ,so that's why i us the Promise Ultra 133 TX2 to get more write speed .
I have try everything in the last 3 weeks still nothing only stress .

Sorry for my bad english ,but i'm try to do my best

[Gurm]

Sorry man - your write speed remains somewhat constant. raid0 doesn't increase write speed.

BUT, framesizes around 640x480 shouldn't overtax any dedicated drive or array... your write speeds DO seem a bit low.

- Gurm

[TSR105]

THX Gurm

THX for youre help

i'm stoped with it for now , going to do somthing els

[Dr Mordrid]

Originally posted by Gurm
Sorry man - your write speed remains somewhat constant. raid0 doesn't increase write speed.
- Gurm

The average write speed on most of my TX4 raids (admittedly with 4 drives) runs somewhere north of 90 mb/s UNBUFFERED using Maxtor D740X drives. This using various benchmarks.

They seem to be handling 65 mb/s capture with no problem, signifying that there is enough real world performance & overkill for the task, so I don't see how you can say that RAID0 doesn't increase write speed

No...you can't ask specifics

Dr. Mordrid

[Gurm]

That TX4 is a much beefier array than TX2. And RAID0 doesn't TECHNICALLY increase write speed - or at least it shouldn't. If your write speed is increased, it's because of something else that the TX4 is doing.

- Gurm

[Dr Mordrid]

All the TX4 is doing different than previous Fasttraks is that is uses two Fasttrak100 ASIC's and a PCI bridge chip to provide 4 master IDE's instead of 2 masters and 2 slaves typical of Fasttraks.

As far as write speeds go I've seen 60+ mb/s with Fasttrak100 & TX2's using two Maxtor D740X drives, so >single drive speed is nothing new there. The ratios with a master/slave card like the FT100 run like this;

2 drives (two masters): 2x single drive write speed
3 drives (+ 1 slave): 2.5x single drive write speed
4 drives (+ 2 slaves): 2.8-2.9x single drive write speed

With the TX4 things get more linear with 3 and 4 drive arrays because they're all masters, but because nothing is perfect a 4 drive TX4 array maxes out a few mb/s below the PCI throughput saturation for the system involved.

Dr. Mordrid



Dr. Mordrid

[LvR]

Gurm, I gotta go with the doc on this one.

Even my much hated HPT370 (by most hardware gurus around here) as well as my Fastrack100 shows a speed increase roughly equal to what doc said there - 2 of the D740X drives in a "Raid 0" (Striping) config gives me around 58M/s, while 3 returns about 71M/s.

You perhaps thinking of a "JBOD" setup? - your access speeds will stay roughly constant and equal to the access speed of a single disk there

[Dr Mordrid]

That's what it sounds like to me too. JBOD would give >capacity but no speed difference.

With RAID0 you have interleaved writes to multiple disks at once, which is how you get the speed increase.

The U of Texas has an interesting page on RAID here;

RAID in general: http://www.utexas.edu/cc/vms/about/raid.html

RAID0: http://www.utexas.edu/cc/vms/about/raid0.html

Dr. Mordrid

[Gurm]

This is one of the reasons that IDE RAID is very confusing. Write multiplexing is NOT part of the original RAID specifications.

Now, as to the original poster - he is certainly getting speeds below what many people experience, but not totally out of the question, depending on a number of factors.

- Gurm

[Dogbert]

Dumb newbie question:

What does it take to compress video in realtime to anything like Mpeg1 and what does it take for Mpeg2/Divx ?

IMHO, if you can compress on the fly, disk access becomes much less relevant, especially with you have enough memory in your computer.