AF:
NF:0
PS:10
SRH:1
SFN:
DSR:
MID:<20090402184819.02590660@ripper.onstor.net>
CFG:
PT:0
S:andy.sharp@onstor.com
RQ:
SSV:mail.onstor.net
NSV:
SSH:
R:<patrick.haverty@onstor.com>,<dl-Leopard@onstor.com>
MAID:1
X-Sylpheed-Privacy-System:
X-Sylpheed-Sign:0
SCF:#mh/Mailbox/sent
RMID:#imap/andys@onstor.net@exch1.onstor.net/INBOX	68296	102AB4F33EBBDB4C91915B145C8E9FB3127E51A2D2@exch1.onstor.net
X-Sylpheed-End-Special-Headers: 1
Date: Thu, 2 Apr 2009 18:48:43 -0700
From: Andrew Sharp <andy.sharp@onstor.com>
To: Patrick Haverty <patrick.haverty@onstor.com>
Cc: dl-Leopard <dl-Leopard@onstor.com>
Subject: Re: Spinning Disk Intent Log
Message-ID: <20090402184843.231bbb33@ripper.onstor.net>
In-Reply-To: <102AB4F33EBBDB4C91915B145C8E9FB3127E51A2D2@exch1.onstor.net>
References: <102AB4F33EBBDB4C91915B145C8E9FB3127E51A2D2@exch1.onstor.net>
Organization: Onstor
X-Mailer: Sylpheed-Claws 2.6.0 (GTK+ 2.8.20; x86_64-pc-linux-gnu)
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit

On Thu, 2 Apr 2009 14:24:53 -0700 Patrick Haverty
<patrick.haverty@onstor.com> wrote:

> Andy,
> 
> Your suggestion brought up some very interesting results.  Write performance with the spinning disk intent log device is less than half of what it is with the SSD intent log, or about 20-25% of the write performance without a separate intent log device.
> 
> Although the SSD itself seems to perform very fast, I don't think there's much chance that, by itself as an intent log, it will speed up pool throughput.
> 
> The chart near the bottom of this blog, linked below, appears to confirm what I have been seeing during testing with separate intent logs.
> 
> http://blogs.sun.com/perrin/entry/slog_blog_or_blogging_on
> 
> So that bears the question (open to everyone), "why have a separate intent log?"

I think I got this one figured out.  Maybe.  Non-sychronous mounted
volumes doing streaming writes with a separate intent-log that is
faster than your regular pool storage will benefit, but for sync
mounted NFS (or all of CIFS), it will actually be slower, because the
data first goes to the intent log device, and then to the pool
(oversimplification alert).

I would think that using the SSD as a separate journal/log device, most
likely mirrored, would increase performance for large scale random I/O
workload profiles.  But wouldn't effect streaming/sequential at all.
