From mboxrd@z Thu Jan  1 00:00:00 1970
From: Ian Campbell <ian.campbell@citrix.com>
Subject: Re: Shared image files and block script performance
Date: Wed, 30 Sep 2015 11:35:20 +0100
Message-ID: <1443609320.16718.177.camel@citrix.com>
References: <7906712.vmDJM0hI90@mlatimer1.dnsdhcp.provo.novell.com>
	<1443518732.16718.16.camel@citrix.com>
	<1931582.zTGj9PHWO8@mlatimer1.dnsdhcp.provo.novell.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
Return-path: <xen-devel-bounces@lists.xen.org>
In-Reply-To: <1931582.zTGj9PHWO8@mlatimer1.dnsdhcp.provo.novell.com>
List-Unsubscribe: <http://lists.xen.org/cgi-bin/mailman/options/xen-devel>,
	<mailto:xen-devel-request@lists.xen.org?subject=unsubscribe>
List-Post: <mailto:xen-devel@lists.xen.org>
List-Help: <mailto:xen-devel-request@lists.xen.org?subject=help>
List-Subscribe: <http://lists.xen.org/cgi-bin/mailman/listinfo/xen-devel>,
	<mailto:xen-devel-request@lists.xen.org?subject=subscribe>
Sender: xen-devel-bounces@lists.xen.org
Errors-To: xen-devel-bounces@lists.xen.org
To: Mike Latimer <mlatimer@suse.com>
Cc: xen-devel@lists.xen.org
List-Id: xen-devel@lists.xenproject.org

On Tue, 2015-09-29 at 15:18 -0600, Mike Latimer wrote:
> Hi Ian,
> 
> On Tuesday, September 29, 2015 10:25:32 AM Ian Campbell wrote:
> > On Mon, 2015-09-28 at 17:14 -0600, Mike Latimer wrote:
> > > Any better options or ideas?
> > 
> > Is part of the problem that shell is a terrible choice for this kind of
> > check?
> 
> There is some truth to that...  ;)
> 
> > Would shelling out to a helper utility allow this to be written in
> > something better?
> 
> A helper utility would be useful, however, I'm seeing a huge amount of gain 
> with nothing more than a little code motion. Specifically, if shared_list is 
> generated within the check_sharing function, the (potentially) large list of 
> devices is not too painful to work with.
> 
> For example, the attached patch works well in my environment, and removes the 
> exponential slowdown. The main change is that $devmm becomes a comma delimited 
> list of devices (major:minor) to check against the vbd's found in xenstore. A 
> few minor changes are required along the way, but nothing significant. The 
> comma delimited list might become problematic at very large numbers (hundreds) 
> of a single shared device, but I don't think it will be a problem in practice. 
> Even if it has limitations, this approach offers significant improvements in 
> performance.
> 
> I'll continue to test this patch here, but I'm interested in your opinion.

If you have a patch which improves things then I think you should just go
ahead formally submit it (http://wiki.xen.org/wiki/Submitting_Xen_Patches).

Ian.