Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 14 Feb 2013 14:00:30 +0100
From:      Sven Brandenburg <sven@crashme.org>
To:        freebsd-scsi@freebsd.org
Subject:   Re: Increase mps sequential read performance with ZFS/zvol
Message-ID:  <511CDFEE.10307@crashme.org>
In-Reply-To: <20130205023609.GA99100@FreeBSD.org>
References:  <510E987C.4090509@oxit.fi> <510E9FD1.5070907@freebsd.org> <20130205023609.GA99100@FreeBSD.org>

next in thread | previous in thread | raw e-mail | index | archive | help



On 02/05/2013 03:36 AM, John wrote:
[..]
>     Gmultipath is used to bind the disks such that each disk can be
> addressed by either controller and the I/O balanced.
>
>     The zfs pool consists of 24 mirrors, each pair one from each shelf.
> The multipaths are rotated such that I/O is balanced between shelves
> and controllers.
[..]
Not sure if this applicable to your problem but in a similar setup like 
yours, I found gmultipath in round-robin mode to be terrible for 
performance.

Afaik the D2700 have their own caching controllers and may shield you 
from this, but if HBAs are directly connected to the disk, concurrent IO 
requests from two 'sides' via gmultipath seem to terrorize them, cutting 
effective read performance down by at least 75%.

Try to switch gmultipath to active/passive. I bet its faster :)

regards,
Sven



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?511CDFEE.10307>