From mboxrd@z Thu Jan 1 00:00:00 1970 From: Nivedita Singhvi Subject: Re: [RFC] TCP Vegas for 2.6 Date: Tue, 09 Mar 2004 11:36:34 -0800 Sender: netdev-bounce@oss.sgi.com Message-ID: <404E1CC2.1010300@us.ibm.com> References: <20040309180331.GC11604@wotan.suse.de> <20040309102206.33c1720c@dell_ss3.pdx.osdl.net> <404E1374.5050600@us.ibm.com> <20040309110323.1295cf92@dell_ss3.pdx.osdl.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Cc: John Heffner , Andi Kleen , linux-net , netdev@oss.sgi.com Return-path: To: Stephen Hemminger In-Reply-To: <20040309110323.1295cf92@dell_ss3.pdx.osdl.net> Errors-to: netdev-bounce@oss.sgi.com List-Id: netdev.vger.kernel.org Stephen Hemminger wrote: > I used ttcp and netperf, but the tests aren't as interesting as setting up the > toplogies, like 1G Ethernet; Ethernet -> PPP -> Ethernet and Ethernet -> IRDA -> Ethernet > and DSL connections through SSH. Plus, LTP (and are local automated test machines) > don't seem to be setup for multi-machine tests. I agree the environments are more interesting in this case - but not sure about just using ttcp and netperf in this case. Single stream, I presume? Both are blasting out data without disk I/O occurring. I'm concerned about enterprise server situations seeing a hit due to the memory/cacheline impact - i.e. catch the environments where this is not going to be of help. We might be able to get some other performance benchmark run data (Specweb* ?). I'm trying to solve two problems - testing of the current mainline kernel for the new features that have gone in and also helping the testing guys with the automated testing of kernel.org kernels for networking stuff (solving that multi- machine problem if they haven't already).. thanks, Nivedita