mbreitba

So I'm getting on the virtualization bandwagon. I'm deploying a supermicro bladecenter and starting off with 1 blade (with a spare for parts). I am using Promise iSCSI for storage (it's sitting in my spare bedroom, it's sweet), and I wondered what everyones thoughts were on how to benchmark the iSCSI, and what RAID configuration I should be choosing.

I'll be using the Promise Vtrak M610i as my initial storage platform, populated with 500GB Seagate Barracuda ES.2 drives. I'm leaning towards a 14 drive RAID10 array, with two hot-spare drives. The only concern I have is that I'll end up with not enough space. 3.5TB pre-formatted capacity isn't all that much for the amount of virtualization that I will be doing.

What kind of tests can I do to check I/O throughput in RAID5 and RAID6 modes to determine if I can go with either of those RAID levels? I'd really love to have a 15 drive RAID6 array (6.5TB w/ one hot spare) if I can get away with the parity overhead calculations. I'll be running webservers off of this iSCSI box, and based on current CPU and memory usage trends, I'll be able to put somewhere between 10-16 VM's on each blade.

Any thoughts? (PS - the bladecenter should be coming in next week, the promise box is already in, the drives, CPU and memory will be here in the next day or two. I'll take pics of all of it when it's all here - 32GB of RAM should be an interesting sight to behold

mbreitba

Also - windows performance tests would be appreciated. I'm using windows 2003 datacenter x64 for my virtualization host OS, and Windows Virtual Server 2005 R2 for my virtualization software. I'll be moving to 2008 and Hyper-V when that's available for the host OS, but most of the guest OS's will stay Windows 2003 Web edition.

Edit - added photo of the drive array

mbreitba

Got the memory and processor in for the first blade. 32GB doesn't even look like that much memory. I was underwhelmed.

mbreitba

Got the HDD's in today - 16 - 500GB Seagate Barracuda ES.2 drives..mmmm....tons of I/O...

CL3P20

ohmehgoo. Look at all of the RAID goodness!

I would consider using SiSandra to bench the HDD's out in the various RAID arrays..HDDTach is another that some like for I/O throughputs.

As far as the type of array.. I know there are obvious advantages/disadvantages to having different types of array setups..but the characteristics and differences between those have more to do with the RAID controller itself than anything else.
I would 'google-around' and see if you can find any threads on that particular RAID controller.. and its qwerks. Other than that.. only testing is going to determine the affects of a different array/striping, on your controller.

__________________

..:Q6600 Club:..

INTEL 45

2x1 Dominators: 540mhz, 4-4-4-8
e7400: 4.5ghz
16.5k-3D '06 8800GS

My photo albums
DCS Inc_client build: Jr Fox

DuckieHo

Impressive.

There's no way to tell performance other than benchmarking. I am 100% sure you won't find comparable benchmarks to your system. HDTach will not be a good indicator of your storage performance since this is a multi-user environment.

__________________
To answer most of your questions: (1) a fridge cannot cool a PC (2) 64-bit OS for over 3.4GB (3) If a PCIe card fits, it should work (4) Resolution, not screen size (5) If you have a question, it is not news (6) Report, not respond to Spam (7) Single/Non-Modular Rail PSUs are not always better than Multi-Rail/Modular

mbreitba

Yeah, the worst part about this is that the 610i is pretty new, and it's the first one to integrate RAID6 capabilities. We've got a m500i and a m300i, and those both work extremely well in RAID5 and RAID10, but RAID6 is a whole new ballgame.

My best friend used to run a hardware review site (HPC Hardware Guide Hardware.Pairnet.Com - way back in the day) and we're planning on doing a full benchmarking suite and writeup on the performance of the array. We've found with the more datacenter oriented hardware, there is very often no good reviews that really give you an in-depth look at the hardware and it's performance. Hopefully we'll change that for at least this piece of hardware.

OCZedd

Wow, can you just chuck them all in Raid0 and bench just to see how crazy high your throughput would be. Also I would recomend raid0+1 or raid10 over raid5, as the parity over will kill it imo/

mbreitba

I can test in RAID0, but the throughput is going to be limited by the gigabit ethernet for sure. Even teaming both of the iSCSI connections together, I'll be limited to 2Gbit/sec (about 250MB/sec). The real screamer is the DAS 14 drive SCSI320 array that we've got. Unfortunately, DAS has gone the way of the dodo, and for clustering and virtualization, having a SAN is the way to go.

Just out of curiosity, has anyone used this tool to benchmark drives, and is it pretty decent? http://www.microsoft.com/Downloads/d...displaylang=en

CL3P20

*off topic.. sorta.

Can anyone here "nutshell" what exactly the main differences are between RAID_0+1 and RAID_10 are?

__________________

..:Q6600 Club:..

INTEL 45

2x1 Dominators: 540mhz, 4-4-4-8
e7400: 4.5ghz
16.5k-3D '06 8800GS

My photo albums
DCS Inc_client build: Jr Fox