gvinum raid5 performance seems slow

[Greg 'groggy' Lehey]

On Sunday, 29 October 2006 at 23:05:32 -0800, R. B. Riddick wrote:
> --- Greg 'groggy' Lehey <grog at FreeBSD.org> wrote:
>> "Sufficiently large data blocks" equates to several megabytes.
>> Currently MAXPHYS, the largest transfer request that would get to the
>> bio layer, is 131072 bytes.  This would imply a stripe size of not
>> more than 32 kB for a five disk array, which is unrealistically small.
>>
>> I did consider this optimization, but it would only work if Vinum were
>> first to buffer multiple requests, and there are all sorts of
>> reliability issues there.  For example, you'd have to lie about the
>> first few requests that were only buffered and not actually sent out
>> to disk.  Possibly it should be done anyway.
>>
> I did it that way in my graid5 class:
> http://home.tiscali.de/cmdr_faako/geom_raid5.tbz

I would have taken a look at it if the sources had been directly web
viewable.

> The source might look somehow incomprehensible,

Heh.  I know the feeling :-)

> but it seems to be quite stable (on amd64 and i386)...
>
>
> I tried to benchmark with a 3 disk array:
> Single disk: 30MB/sec write speed
> graid5 device: 20MB/sec write speed

Can you give more details?  Stripe size, request size, number of
concurrent accessors, etc.?

> I would be interested in other benchmarks about it.

Try rawio.

Greg
--
See complete headers for address and phone numbers.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-performance/attachments/20061030/ec8d63cb/attachment.pgp