Recent RAID0 vs. Not in Single Use Workstations?

[Doc]

About to build a new workstation.
AMD FX53 with 1GB RAM on (probably) MSI Neo2 Plat board.

Use is streaming video, video editing and some minimal graphics.
Not photoshop.

Reading that RAID0 is NOT providing the "great" real life performance
improvement once thought and need to investigate.

Drives will be Maxtor MaxLine III 250 GB/16MB cache (3 of 'em) and
*was* planning on a single for system C: and then the other two in
a RAID1 for archiving, etc... basic storage on D:

Realize that an INTEL chipset ICH6 would provide tagged queuing, but
that means a PIV and the FX53 is coming in cheaper for the top end
and faster so really leaning towards the nVidia chipset board. Seeing
very little 'single use' difference for tagged queuing anyway.

Is the community in "agreement" (oh yeah...) on the nebulous benefits of
RAID0 for single workstation use? Refuting comments greatly appreciated.
Thanks.

 

[Arno Wagner]

About to build a new workstation.
AMD FX53 with 1GB RAM on (probably) MSI Neo2 Plat board.

Use is streaming video, video editing and some minimal graphics.
Not photoshop.

Reading that RAID0 is NOT providing the "great" real life performance
improvement once thought and need to investigate.

Drives will be Maxtor MaxLine III 250 GB/16MB cache (3 of 'em) and
*was* planning on a single for system C: and then the other two in
a RAID1 for archiving, etc... basic storage on D:

Realize that an INTEL chipset ICH6 would provide tagged queuing, but
that means a PIV and the FX53 is coming in cheaper for the top end
and faster so really leaning towards the nVidia chipset board. Seeing
very little 'single use' difference for tagged queuing anyway.

Is the community in "agreement" (oh yeah...) on the nebulous benefits of
RAID0 for single workstation use? Refuting comments greatly appreciated.

I have seen a speed-up of factor 2 with Linux software RAID0 on
reading, but not on writing, which was slower. This was just a
test-set up and some time ago. I see about 2 times disk speed
when reading from software RAID5 with >= 3 disks, but again, writing
is slower, maybe half the native disk speed, regardless of number of
drives.

I recently tried to use an Adaptec 2810AS SATA RAID contoller (pice
of trash) under Linux and it was slower than RAID5 in software.

Arno

 

[Odie Ferrous]

Is the community in "agreement" (oh yeah...) on the nebulous benefits of
RAID0 for single workstation use? Refuting comments greatly appreciated.
Thanks.

I certainly am.

When I used to use Photoshop 7 editing large (100MB plus) files was
inherently quicker on a RAID 0 array. I was, however, using SCSI
(Adaptec 2100S) which was a lot quicker than the equivalent ATA RAID 0.

My experience is that RAID 0 will make an enormous difference to any
application where disk activity is frenzied.


Odie

 

[Shailesh Humbad]

About to build a new workstation.
AMD FX53 with 1GB RAM on (probably) MSI Neo2 Plat board.

Use is streaming video, video editing and some minimal graphics.
Not photoshop.

Reading that RAID0 is NOT providing the "great" real life performance
improvement once thought and need to investigate.

Drives will be Maxtor MaxLine III 250 GB/16MB cache (3 of 'em) and
*was* planning on a single for system C: and then the other two in
a RAID1 for archiving, etc... basic storage on D:

Realize that an INTEL chipset ICH6 would provide tagged queuing, but
that means a PIV and the FX53 is coming in cheaper for the top end
and faster so really leaning towards the nVidia chipset board. Seeing
very little 'single use' difference for tagged queuing anyway.

Is the community in "agreement" (oh yeah...) on the nebulous benefits of
RAID0 for single workstation use? Refuting comments greatly appreciated.
Thanks.

Where did you hear RAID-0 is not providing 'great' performance? I
have 2x raid-0 raptor 36GB drives and I'm luvin' it. My computer is
only an XP 1.8Ghz, but with the drives, it seems faster than a 3Ghz
P4. I'll never go back to a single drive. Many of the tasks that are
disk-intensive run as much as twice as fast. On my single-user
workstation, some examples are:

searching eudora email box (100Megs of mail)
starting up WinXP
shutting down WinXP
Loading the dozen or so programs i run daily.
large database insert/update queries
picture browsing (1.8MB/file)
recording video from ati tv player
playing back video
searching the file system for files
enqueing My Music into winamp
installing any new programs
uninstalling programs
loading large games, like half-life

Nebulous benefits? Probably true for an office PC that runs a single
application all day and stays on 24/7, but not in my situation. I
wouldn't see any benefit from tagged queueing though.

 

[Ron Reaugh]

About to build a new workstation.
AMD FX53 with 1GB RAM on (probably) MSI Neo2 Plat board.

Use is streaming video, video editing

There is a MAJOR advantage to RAID 0 in video editing.

and some minimal graphics.
Not photoshop.

Reading that RAID0 is NOT providing the "great" real life performance
improvement once thought and need to investigate.

Those claims are misleading. They refer to the average user doing Office
etc with low intensity disk I/O.

Drives will be Maxtor MaxLine III 250 GB/16MB cache (3 of 'em) and
*was* planning on a single for system C: and then the other two in
a RAID1 for archiving, etc... basic storage on D:

Realize that an INTEL chipset ICH6 would provide tagged queuing,

Assume NOT implemented in drivers yet.

but
that means a PIV and the FX53 is coming in cheaper for the top end
and faster so really leaning towards the nVidia chipset board. Seeing
very little 'single use' difference for tagged queuing anyway.

Tagged command queuing is not important for video editing nor most
workstation usage.

Is the community in "agreement" (oh yeah...) on the nebulous benefits of
RAID0 for single workstation use?

RAID 0 has distinct advanatges for many power users.

 

[Ron Reaugh]

I certainly am.

For video editing RAID 0 rules.

When I used to use Photoshop 7 editing large (100MB plus) files was
inherently quicker on a RAID 0 array. I was, however, using SCSI
(Adaptec 2100S) which was a lot quicker than the equivalent ATA RAID 0.

That's completely false.

My experience is that RAID 0 will make an enormous difference to any
application where disk activity is frenzied.

Right.

 

[Odie Ferrous]

For video editing RAID 0 rules.


That's completely false.

It's not, actually. At the time I had just purchased a new Adaptec
2400A IDE RAID card and was getting poor performance from RAID 0.

I finally managed to get Adaptec to do a straight swap for their 2100S,
a SCSI RAID card.

The difference was noticeable.

Granted, using 10K rpm high speed IDE drives will sway the argument more
in IDE's favour, but taking reliability issues into account, SCSI wins
hands down for me.

But I suppose you would require a lot of experience of both setups
before passing valid comment. I do have that experience.


Odie

 

[Doc]

Ron Reaugh typed this:

There is a MAJOR advantage to RAID 0 in video editing.


Those claims are misleading. They refer to the average user doing
Office etc with low intensity disk I/O.


Assume NOT implemented in drivers yet.


Tagged command queuing is not important for video editing nor most
workstation usage.


RAID 0 has distinct advanatges for many power users.

Hmm, okay. Well, this is interesting. I have a 3 drive Maxtor Highpoint
rocket raid 404 in my current system but it is not the C drive. So the
quicker loading and and shutting down of XP would not be noted. Indeed I
am really only using the drives for data and storage retrieval.

I *DO* note that I had a heck of a time 'installing' some applications
(the latest Word Pro from lotus for example) onto a RAID array. Not
sure, and wasn't then, WHY. Finally copied the files to the C drive and
did a direct install from there... weird behavior also effected Delorme
mapping app installation.

So given that the CONSENSUS here is to use RAID0 for video and other
such application use, I'll defin. reconsider the design. It was the
review at STORAGEREVIEW.com and a few other locations that piqued my
questions. Thanks.

 

[Folkert Rienstra]

It's not, actually. At the time I had just purchased a new Adaptec
2400A IDE RAID card and was getting poor performance from RAID 0.

I finally managed to get Adaptec to do a straight swap for their 2100S,
a SCSI RAID card.

The difference was noticeable.

Liar/Fantast:

" I have 2 x 10K rpm SCSI drives in a RAID 0 configuration.
The performance, even with 128MB of cache memory, is slower
than a single IDE drive running off the motherboard."

"My RAID 0 array consists of 3 x Fujitsu 10K rpm drives on an Adaptec
2100s controller. I get 100MB/s writes**, which is probably ok for most."

"I've done some benchmarking with my Adaptec and 128MB of cache
memory. Data is written at around 350 - 450 MegaBYTES per second."

Granted, using 10K rpm high speed IDE drives will sway the argument more
in IDE's favour, but taking reliability issues into account, SCSI wins
hands down for me.

But I suppose you would require a lot of experience of both setups
before passing valid comment. I do have that experience.

Your experience is obviously all made-up.

 

[Odie Ferrous]

Liar/Fantast:

" I have 2 x 10K rpm SCSI drives in a RAID 0 configuration.
The performance, even with 128MB of cache memory, is slower
than a single IDE drive running off the motherboard."

At that stage, because of problems, that was the case.

I suppose you know better - I mean, you were there, right?

"My RAID 0 array consists of 3 x Fujitsu 10K rpm drives on an Adaptec
2100s controller. I get 100MB/s writes**, which is probably ok for most."

"I've done some benchmarking with my Adaptec and 128MB of cache
memory. Data is written at around 350 - 450 MegaBYTES per second."

1. Burst.

2. Burst to cache memory. According to drive test util.

Your experience is obviously all made-up.

You're a funny little man, Folkert. Why do you need to spend your time
going through archive material?

Shame - you should put your energy into more fruitful things.

Anyway, you do make me laugh so!!!


Odie

 

[Arno Wagner]

" I have 2 x 10K rpm SCSI drives in a RAID 0 configuration.
The performance, even with 128MB of cache memory, is slower
than a single IDE drive running off the motherboard."

That is possible, if something is misconfigured or broken. Otherwise not.

"My RAID 0 array consists of 3 x Fujitsu 10K rpm drives on an Adaptec
2100s controller. I get 100MB/s writes**, which is probably ok for most."

Given that I get 70MB/s sustained linear writes on a software RAID5
under Linux in a PCI system (4 disks on the chipset, 3 on PCI) I would
say this one is possible. However I would say not with standard PCI.
The bus ist just not fast enough.

"I've done some benchmarking with my Adaptec and 128MB of cache
memory. Data is written at around 350 - 450 MegaBYTES per second."

Agreed. Unless the benchmarks were smaller than the buffer memory,
write buffering was turned on and the bus was at least PCI-X.
A completely meaningless result nonetheless, if true.

Arno

 

[Eric Gisin]

That is possible, if something is misconfigured or broken. Otherwise not.

There are some hardware RAID cards that perform this poorly.

Given that I get 70MB/s sustained linear writes on a software RAID5
under Linux in a PCI system (4 disks on the chipset, 3 on PCI) I would
say this one is possible. However I would say not with standard PCI.
The bus ist just not fast enough.

Clueless. Adaptec and LSI Ultra160 cards can do 120MB/s with RAID 0.

 

[Folkert Rienstra]

That is possible, if something is misconfigured or broken. Otherwise not.


Given that I get 70MB/s sustained linear writes on a software RAID5
under Linux in a PCI system (4 disks on the chipset, 3 on PCI) I would
say this one is possible.

However I would say not with standard PCI.

Standard PCI is what is on the 2100S.

The bus ist just not fast enough.

Actually it *is* fast enough.

Agreed. Unless the benchmarks were smaller than the buffer memory,
write buffering was turned on and the bus was at least

PCI-X.

Nope, 64-bit 66MHz PCI would accomodate that nicely.