Raid 10 benchmarks?

[4dm]

I am building a new Conroe system, and with the stellar deal (IMO) on the spinpoint 160GB sata II drives on newegg, I decided I want to take advantage of my ICH8R southbridge's built in raid 10 controller. (Using the GA-965P-DQ6 mobo)

I have had raid 0 arrays in the past, but been completely frustrated by them when they fail. I figure a raid 10 array is just what I need to satisfy my performance wants without the frustration of instant failure/rebuilding when one drive goes bad.

My question is, does anyone have any benchmarks on SATA II drives on raid 0 versus raid 10 performace?

The only thing I could find benchmarking was this site here, but I would believe the performance differences would be a lot larger with SATA II drives.
http://www.astro.ufl.edu/~ken/perf_tests.html

Does anyone have any thoughts or benchmarks to share?

thanks,
-4DM

 

[dBTelos]

Link.

 

[BDawg]

Just build a RAID 1 and be done with it. The article shows that there's no real world difference is RAID 0, 1, 0+1, 10, and single drive.

 

[4dm]

Thanks dBTelos. That is exactly what I was looking for.

BDawg, thanks for the laugh, you keep thinking that. I suggest you look over the benchmarks instead of coming to your own conclusions.

 

[RebateMonger]

Originally posted by: 4dm
BDawg, thanks for the laugh, you keep thinking that. I suggest you look over the benchmarks instead of coming to your own conclusions.

Huh?

As shown in the article's conclusions, the real-world performance of a RAID 1 and a RAID 10 array on these motherboard-based RAID chipsets are going to be very similar.

 

[4dm]

???

Every single page from 7 onward shows that raid 10 offers significant performance advantages over raid 1. Sometimes as little as +24% (Partition to partition copy - ISO), other times as much as +100%. A transaction rate increased of +73% for the file server and workstation benchmarks alone make it a no-brainer for me.

Read speeds are near equivalent, both raid arrays have 2x drives for 1x the data, write speeds are significantly improved when you stripe versus when you mirror.

 

[dBTelos]

Thanks dBTelos. That is exactly what I was looking for.

Np.

 

[BDawg]

Originally posted by: 4dm
???

Every single page from 7 onward shows that raid 10 offers significant performance advantages over raid 1. Sometimes as little as +24% (Partition to partition copy - ISO), other times as much as +100%. A transaction rate increased of +73% for the file server and workstation benchmarks alone make it a no-brainer for me.

Read speeds are near equivalent, both raid arrays have 2x drives for 1x the data, write speeds are significantly improved when you stripe versus when you mirror.

You will not notice a real world difference. You're just wasting your money.

 

[erwos]

RAID 5 is a better bet, IMHO. You don't lose everything to a single drive failure, and you don't lose nearly as much space as you do with RAID 10 (1/n vs 1/2).

 

[BDawg]

Originally posted by: erwos
RAID 5 is a better bet, IMHO. You don't lose everything to a single drive failure, and you don't lose nearly as much space as you do with RAID 10 (1/n vs 1/2).

5 isn't a better bet unless you have a dedicated hardware controller.

 

[erwos]

Originally posted by: BDawg

Originally posted by: erwos
RAID 5 is a better bet, IMHO. You don't lose everything to a single drive failure, and you don't lose nearly as much space as you do with RAID 10 (1/n vs 1/2).

5 isn't a better bet unless you have a dedicated hardware controller.

That's what a dual-core CPU is for. Remember, a lot of these dedicated controllers are running on <500mhz XScale CPUs. Your own CPU is definitely up to the task.

 

[Madwand1]

Originally posted by: erwos

Originally posted by: BDawg
5 isn't a better bet unless you have a dedicated hardware controller.

That's what a dual-core CPU is for. Remember, a lot of these dedicated controllers are running on <500mhz XScale CPUs. Your own CPU is definitely up to the task.

Often this has nothing to do with the CPU, and everything to do with the RAID 5 implementation. MB RAID 5 performs poorly. Windows software RAID 5 (up to 2003 at least) performs poorly. Linux sometimes performs well. For many cases here, the broad solution to good RAID 5 performance is a decent add-on RAID controller.

Of course if you're dealing with 100 Mb/s networking or very old hardware, etc., then these differences in performance are moot.

 

[d3n]

RAID5 to me is what I woud use for archive only. Raid 10 is the prefered method for busy databases becuase of the lack of slowdown wtih the parity checking compared to RAID 5 and it also lets you better utilize your HD space as compared to RAID 1.

I use RAID 5 with a highpoint contoller at home. Its not a cheapo contoller but its not a hardware one ethier. To me is just a 'read only' solution. Any type of writes I can expect a slowdown on my reads because of the parity calculations. I write files to a diffrent raid set (diffrent controller) when my RAID 5 is being read from (streaming movies, music, backups) and then it gets moved to the archived during downtime.