<html><body bgcolor="#FFFFFF"><div>On Apr 15, 2011, at 9:17 AM, Rudi Ahlers <<a href="mailto:Rudi@SoftDux.com">Rudi@SoftDux.com</a>> wrote:<br><br></div><div></div><blockquote type="cite"><div><br><br><div class="gmail_quote">On Fri, Apr 15, 2011 at 3:05 PM, Christopher Chan <span dir="ltr"><<a href="mailto:christopher.chan@bradbury.edu.hk"><a href="mailto:christopher.chan@bradbury.edu.hk">christopher.chan@bradbury.edu.hk</a></a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div><div></div><div class="h5">On Friday, April 15, 2011 07:24 PM, Benjamin Franz wrote:<br>
> On 04/14/2011 09:00 PM, Christopher Chan wrote:<br>
>><br>
>> Wanna try that again with 64MB of cache only and tell us whether there<br>
>> is a difference in performance?<br>
>><br>
>> There is a reason why 3ware 85xx cards were complete rubbish when used<br>
>> for raid5 and which led to the 95xx/96xx series.<br>
>> _<br>
><br>
> I don't happen to have any systems I can test with the 1.5TB drives<br>
> without controller cache right now, but I have a system with some old<br>
> 500GB drives (which are about half as fast as the 1.5TB drives in<br>
> individual sustained I/O throughput) attached directly to onboard SATA<br>
> ports in a 8 x RAID6 with *no* controller cache at all. The machine has<br>
> 16GB of RAM and bonnie++ therefore used 32GB of data for the test.<br>
><br>
> Version 1.96 ------Sequential Output------ --Sequential Input-<br>
> --Random-<br>
> Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block--<br>
> --Seeks--<br>
> Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP<br>
> /sec %CP<br>
> pbox3 32160M 389 98 76709 22 91071 26 2209 95 264892 26<br>
> 590.5 11<br>
> Latency 24190us 1244ms 1580ms 60411us 69901us<br>
> 42586us<br>
> Version 1.96 ------Sequential Create------ --------Random<br>
> Create--------<br>
> pbox3 -Create-- --Read--- -Delete-- -Create-- --Read---<br>
> -Delete--<br>
> files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP<br>
> /sec %CP<br>
> 16 10910 31 +++++ +++ +++++ +++ 29293 80 +++++ +++<br>
> +++++ +++<br>
> Latency 775us 610us 979us 740us 370us<br>
> 380us<br>
><br>
> Given that the underlaying drives are effectively something like half as<br>
> fast as the drives in the other test, the results are quite comparable.<br>
<br>
</div></div>Woohoo, next we will be seeing md raid6 also giving comparable results<br>
if that is the case. I am not the only person on this list that thinks<br>
cache is king for raid5/6 on hardware raid boards and the using hardware<br>
raid + bbu cache for better performance one of the two reasons why we<br>
don't do md raid5/6.<br>
<div class="im"><br>
<br>
><br>
> Cache doesn't make a lot of difference when you quickly write a lot more<br>
> data than the cache can hold. The limiting factor becomes the slowest<br>
> component - usually the drives themselves. Cache isn't magic performance<br>
> pixie dust. It helps in certain use cases and is nearly irrelevant in<br>
> others.<br>
><br>
<br>
</div>Yeah, you are right - but cache is primarily to buffer the writes for<br>
performance. Why else go through the expense of getting bbu cache? So<br>
what happens when you tweak bonnie a bit?<br>
<div><div></div><div class="h5">_______________________________________________<br><br></div></div></blockquote><div><br></div><div><br></div><div>As matter of interest, does anyone know how to use an SSD drive for cach purposes on Linux software RAID drives? ZFS has this feature and it makes a helluva difference to a storage server's performance. </div></div></div></blockquote><br><div>Put the file system's log device on it.</div><div><br></div><div>-Ross</div><div><br></div></body></html>