Software RAID5 with 2.6.0-test
Hi,
First off, I am never going to try building a software IDE RAID 5
on Linux again in an important production system - my experience says that it just isn't possible to get that many IDE channels to work together reliably. I know there are some people who have managed it, and I did hear there was a race condition and a patch to fix it - but in the end in a production system with a serious size of RAID throwing in a 3-ware or similar is a hell of a lot simpler and just works.
As for speed; well I'm not sure. I think the comparisons of
100MHz processors vs the speed of your xeon are bogus. I would hope at least some of the work on the hardware raid controllers is done in hardware (XORing blocks of data isn't exactly hard in hardware), and in addition the amount of CPU used is going to be limited by memory bandwidth (and associated cache pollution?) before the clock rate of the processor gets involved I would have thought.
I'd love to see some real benchmarks to prove me wrong however!
Dave
-----Open up your eyes, open up your mind, open up your code -------
/ Dr. David Alan Gilbert | Running GNU/Linux on Alpha,68K| Happy \ \ gro.gilbert @ treblig.org | MIPS,x86,ARM,SPARC,PPC & HPPA | In Hex / \ _________________________|_____ http://www.treblig.org |_______/
In article <20031018115049.GB760@gallifrey>, Dr. David Alan Gilbert <[email protected]> wrote:
| I'd love to see some real benchmarks to prove me wrong however!
Okay, I'm scheduled to build a system next month, give me the name of a good controller which will do four drives and which can be used as a dumb controller as well, and I'll grab six drives for the build and run a test (baring a change in schedule, obviously).
I'll run as many RAID configs as make sense, whatever the hardware supports. And also four way RAID-1 for a heavy read application, to see if the software RAID does better than the controller, if it can do that at all.
Signature · 3 lines
bill davidsen <[email protected]> CTO, TMR Associates, Inc Doing interesting things with little computers since 1979.
Quoted 9 lines · bill davidsen
In article <20031018115049.GB760@gallifrey>, Dr. David Alan Gilbert <[email protected]> wrote:
| I'd love to see some real benchmarks to prove me wrong however!
Okay, I'm scheduled to build a system next month, give me the name of a good controller which will do four drives and which can be used as a dumb controller as well, and I'll grab six drives for the build and run a test (baring a change in schedule, obviously).
I'd try the 3ware. AC and others have reported good throughput and, at least compared to SCSI, they aren't that expensive and claim to support JBOD. I just checked on pricewatch and the 7506-4 can be had for about $250US and the eight port for about $375. Older models can be had for less. When the economy picks back up enough to improve my bank ballance i'm planning to buy a few.
Quoted 4 lines
I'll run as many RAID configs as make sense, whatever the hardware supports. And also four way RAID-1 for a heavy read application, to see if the software RAID does better than the controller, if it can do that at all.
Signature · 5 lines
________________________________________________________________ J.W. Schultz Pegasystems Technologies email address: [email protected] Remember Cernan and Schmitt