[aklug] Re: Raid speed

From: Shane R. Spencer <shane@bogomip.com>
Date: Fri Apr 01 2011 - 13:27:50 AKDT

If it weren't fast enough for a single user I'd be worried. I use it all the time. I've
got a few arrays that seem happy with 60MB/s using software raid which is more than I
need. You can get a LOT higher with good drives and a zippy drive controller.

http://www.linux.com/var/uploads/Image/articles/140734-1.jpg

The benefits you see with hardware raid are typically lower latency and potentially a
having a higher I/O transactions per second. However if there is only a 25% discrepancy
in the IOPS it's not worth it to go full hardware.

For instance.. if you give yourself a 15 second window to snag 20 files off of a raid
array at maximum speed and the hardware array completes the request in 10 seconds did you
gain anything over a software array that completed the request in 12 seconds? That's a
really simple and crude example, but you have to remember that your CPU is just sitting
there most of the time pondering its own existence which is mostly futile. Why offload
storage control and processing to a hardware controller if you don't need the bells and
whistles and have the horsepower to do it on the CPU? On top of that, what sort of drive
analysis are you getting from the raid controller if it doesn't support that level of awesome.

- Shane

On 04/01/2011 12:46 PM, Christopher Howard wrote:
> I'm building my own tiny server for net storage, pxeboot, and a few small services. Prob mount nfs shares from it, or maybe iscsi. Wanted to build 4 or 5 disk software raid array. Do u think that software raid will be fast enough for practical use on my small in-home LAN, or am i going to have to buy a raid controller card of some kind? Raid1 with 2 drives is all i've ever tried before.
> frigidcode.com
> Sent from my HTC
> ---------
> To unsubscribe, send email to<aklug-request@aklug.org>
> with 'unsubscribe' in the message body.
>

---------
To unsubscribe, send email to <aklug-request@aklug.org>
with 'unsubscribe' in the message body.
Received on Fri Apr 1 13:27:39 2011

This archive was generated by hypermail 2.1.8 : Fri Apr 01 2011 - 13:27:39 AKDT