From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: Re: [PATCH v3 net-next 0/4] net: low latency Ethernet device polling Date: Tue, 21 May 2013 06:31:39 -0700 Message-ID: <1369143099.3301.216.camel@edumazet-glaptop> References: <20130520101552.14133.45953.stgit@ladj378.jer.intel.com> <519B632F.7040202@mellanox.com> <519B68A4.3010406@linux.intel.com> <519B7355.9090303@mellanox.com> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit Cc: Eliezer Tamir , Dave Miller , linux-kernel@vger.kernel.org, netdev@vger.kernel.org, Jesse Brandeburg , Don Skidmore , e1000-devel@lists.sourceforge.net, Willem de Bruijn , Andi Kleen , HPA , Eliezer Tamir To: Alex Rosenbaum Return-path: In-Reply-To: <519B7355.9090303@mellanox.com> Sender: linux-kernel-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On Tue, 2013-05-21 at 16:15 +0300, Alex Rosenbaum wrote: > On 5/21/2013 3:29 PM, Eliezer Tamir wrote: > > What benchmarks are you using to test poll/select/epoll? > for epoll/select latency tests we are using sockperf as performance > latency tool: https://code.google.com/p/sockperf/ > It is a client-server based tool and it supported ping-pong, throughput, > and under-load test type. > For epoll, you will need to define a 'feedfile' ("-f filepathname") > which has a list of TCP and/or UDP socket and defined your IO mux type > ("-F epoll"). I totally agree, most modern applications use poll/select/epoll, and a fair amount of sockets per task, sendfile()/vmsplice()/... and netperf is not using same paths. Thanks Alex !