From mboxrd@z Thu Jan 1 00:00:00 1970 From: Zoltan Kiss Subject: Re: xen-netback in linux 3.12 Date: Mon, 20 Jan 2014 10:28:29 +0000 Message-ID: <52DCFA4D.4070903@citrix.com> References: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; Format="flowed" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: xu cong , xen-devel@lists.xensource.com Cc: wei.liu2@citrix.com, ian.campbell@citrix.com List-Id: xen-devel@lists.xenproject.org On 17/01/14 21:30, xu cong wrote: > Hi all, > > I found xen-netback changed a lot in latest kernel. The shared > xen-netback threads in driver domain is replaced by per-vif kernel > thread. In my platform, I found the I/O throughput and scalability of > per-vif netback is better than previous implementation. Another > advantage is that I can use cgroups to control the CPU fair share among > all VMs in driver domain (I group netback thread and blkback thread for > each VM). Is there any other motivation for this modification? Thanks. as far as I remember scalability was the only motivation. Previously you could end in a situation where one thread (pinned to a vCPU) did a lot of work while others nothing. In this thread-per-VIF model the kernel has the ability to schedule the workloads as it want, and users can also poke around with settings. Zoli