8 Spinpoint F3 1TB's in hardware RAID 10

[El_Capitan]

I'm upgrading my Virtualization server, parts come in tomorrow, Saturday, and Monday. I'm using the 3ware 9650SE-8LPML KIT PCI Express x4 SATA II Controller Card with a BBU-MODULE-04, along with 8 Spinpoint F3 1TB's in RAID 10.

 

I'll also be comparing my 4 Spinpoint F3 1TB's in ICH10R RAID 10.

 

Any free hard drive benchmark software out there other than the HD Tach? One that benches IOPS and Writes would be good to have.

 

Ok... let me know what to test. Here's a benchmark that's pretty impressive:

Samsung Spinpoint F3 8x1TB RAID10

 

Edited May 8, 2010 by El_Capitan

[The Smith]

There's HDTune, AS SSD, ATTO.

 

I'm eager to see the results!

[Nemo]

There's the Intel NAS Performance Toolkit. It measures performance under real-world scenarios instead of just IOPs, which is more tuned to actual drive performance. It gives you an easy and repeatable test to compare different setups and represents a more realistic testing scenario than running other benchmark testing programs. Your test platform (not the server) has to be running an Intel processor though.

[El_Capitan]

There's HDTune, AS SSD, ATTO.

 

I'm eager to see the results!

 

There's the Intel NAS Performance Toolkit. It measures performance under real-world scenarios instead of just IOPs, which is more tuned to actual drive performance. It gives you an easy and repeatable test to compare different setups and represents a more realistic testing scenario than running other benchmark testing programs. Your test platform (not the server) has to be running an Intel processor though.

Sweet, I'll try those out. Thanks! Will be back later tonight to give the results (hopefully all drives are good).

[krazypoloc]

There's the Intel NAS Performance Toolkit. It measures performance under real-world scenarios instead of just IOPs, which is more tuned to actual drive performance. It gives you an easy and repeatable test to compare different setups and represents a more realistic testing scenario than running other benchmark testing programs. Your test platform (not the server) has to be running an Intel processor though.

It appears that this needs to be done on a raw drive with no data is that correct?

[hardnrg]

RAID 10 would barely use the processing power of the 9650SE... you'd need to have parity XOR calculations (e.g. RAID5) to show a signficant advantage of hardware RAID vs onboard/host/software RAID

 

I'd use RAID5, 50, or maybe 5+HotSpare or 6... you can get very fast RAID10 performance without hardware RAID cards... if you make use of RAID 5, 50, or 6, you would have more speed and capacity

[El_Capitan]

RAID 10 would barely use the processing power of the 9650SE... you'd need to have parity XOR calculations (e.g. RAID5) to show a signficant advantage of hardware RAID vs onboard/host/software RAID

 

I'd use RAID5, 50, or maybe 5+HotSpare or 6... you can get very fast RAID10 performance without hardware RAID cards... if you make use of RAID 5, 50, or 6, you would have more speed and capacity

True, however, the advantage of a BBU and write caching makes it a worthwhile investment particularly for RAID 10. While RAID 5 will offer up to 10-15% faster performance over RAID 10 in 8KB random reads, it does horribly (up to 60-65% worse in 8KB random writes). Since I primarily host VM servers with write intensive databases (except for my reports VM server), it's great for what I need. I also don't like the fact that if one drive fails in RAID 5, it can degrade performance by 50%, while if that happens in RAID 10, there would be no degradation at all.

[The Smith]

PCMark Vantage HDD test is great also.

 

Edit: D'Oh, you asked for free benchmarks.

Edited May 8, 2010 by The Smith

[El_Capitan]

PCMark Vantage HDD test is great also.

 

Edit: D'Oh, you asked for free benchmarks.

Well, I don't mind paying for a benchmark if it isn't too pricey. I was debating about buying the full license for HD Tune Pro.

 

Anyways, my first benchmark is up. Close to 600MB/s in Reads and Writes with 3.6TB space.

[hardnrg]

Is it really that bad for 8KB random writes on the 9650SE for degraded-RAID5? Degraded-RAID6 write performance seems almost normal here, although admittedly quite likely an average block size.

 

I see now why you chose RAID10, but if you have the time/patience to test, I am very curious to see the difference between 8-drive RAID10, 5, and 6, normal/degraded/rebuilding... there doesn't appear to be any reviews of this card.

[El_Capitan]

Is it really that bad for 8KB random writes on the 9650SE for degraded-RAID5? Degraded-RAID6 write performance seems almost normal here, although admittedly quite likely an average block size.

 

I see now why you chose RAID10, but if you have the time/patience to test, I am very curious to see the difference between 8-drive RAID10, 5, and 6, normal/degraded/rebuilding... there doesn't appear to be any reviews of this card.

Yeah, I'm curious to see the difference between different RAID configurations, too. Especially 5, 6, and 50 - although there's a lot to think about using those configurations (for what type of environment you'll be using it for, and setting the best RAID Stripe size and Allocation Unit size). The only reviews of the card are on Newegg, and it's been pretty decent as far as reviews.

[ebarone]

Roughly how much did the controller card and the eight hard drives cost, and how much space do you use on them?