Performance of Drive Extender vs Software RAID 5
-
Monday, April 30, 2007 8:12 PMI currently have a server running Windows 2003 Standard configured with software RAID 5.
I've run a couple of hard drive benchmarking tools and am curious if the performance would increase or decrease using WHS?
C: is a single Drive
E: is a software RAID 5 drive
Passmark gives the following marks:
C:
Disk - Sequential Read: 12.4 MBytes transferred per second
Disk - Sequential Write: 15.1 MBytes transferred per second
Disk - Random Seek + RW: 2.07 MBytes transferred per second
E:
Disk - Sequential Read: 25.4 MBytes transferred per second
Disk - Sequential Write: 1.20 MBytes transferred per second
Disk - Random Seek + RW: 1.09 MBytes transferred per second
FreshDiagnose gives the following:
C:
Write: 14 MB/s
Read: 10 MB/s
E:
Write: 3 MB/s
Read: 188 MB/s
The numbers above seems to vary between each program, but I have run the test several times using each program and they are consistent withing each program.
The CPU is an AMD Athlon 1.1 GHz with 512 MB RAM
I have another machine I could move up to and it is a P4 1.8 with 512 MB or RAM,
but I'm curious as to how performance compares between drive extender and software RAID 5.
thanks,
tom
Answers
-
Monday, April 30, 2007 9:24 PMModerator Well, you can't configure WHS to use software RAID at all. I promise that you'll break Drive Extender in a heartbeat if you try.
So I'm not sure I understand the point of your question.
But WHS should deliver performance on a par with a single disk for both reads and writes. That's because it pretty much is delivering data off a single disk.
RAID will shine for reads, and may deliver data fast enough to saturate a gigabit connection to your LAN. It will suck for writes. Both of which are borne out by your numbers.
Something your numbers leave out is the performance if your array is in a degraded state. Software RAID will deliver very bad performance in that case, because it has to calculate the data on the missing drive "on the fly" from the parity information on the other drives.
All Replies
-
Monday, April 30, 2007 9:24 PMModerator Well, you can't configure WHS to use software RAID at all. I promise that you'll break Drive Extender in a heartbeat if you try. So I'm not sure I understand the point of your question.
But WHS should deliver performance on a par with a single disk for both reads and writes. That's because it pretty much is delivering data off a single disk.
RAID will shine for reads, and may deliver data fast enough to saturate a gigabit connection to your LAN. It will suck for writes. Both of which are borne out by your numbers.
Something your numbers leave out is the performance if your array is in a degraded state. Software RAID will deliver very bad performance in that case, because it has to calculate the data on the missing drive "on the fly" from the parity information on the other drives. -
Monday, April 30, 2007 10:00 PMI am aware that I can't run with a RAID configuration. I was curious how the data that is replicated will perform. So the speed of replicated data is the same as data that is one a single disk, correct?
I'm looking to get data protection and faster speeds. My Read speeds are great and Writes speed suck.
thanks,
tom -
Monday, April 30, 2007 10:25 PMModeratorThe speed of replicated data is approximately that of a single drive, yes (I think it should be a bit slower, and with a bit higher latency for the initial response, but it's a good enough approximation). You'll find that modern drives deliver acceptable performance for just about anything that the home user is likely to want to do.
If you want blindingly fast reads, and decent writes, you want a hardware RAID controller with on-board parity calculation and preferably a large cache. Even a degraded array will typically perform as well as (or better than) a single drive on reads. Note that there are no motherboard RAID solutions that offer that hardware parity calc, so motherboard RAID has the same issues as software RAID.
.png){kind=spacer}
