From owner-freebsd-hackers Sat Aug 16 15:35:19 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.5/8.8.5) id PAA13679 for hackers-outgoing; Sat, 16 Aug 1997 15:35:19 -0700 (PDT) Received: from avon-gw.uk1.vbc.net (jdd@avon-gw.uk1.vbc.net [194.207.2.20]) by hub.freebsd.org (8.8.5/8.8.5) with ESMTP id PAA13673 for ; Sat, 16 Aug 1997 15:35:11 -0700 (PDT) Received: (from jdd@localhost) by avon-gw.uk1.vbc.net (8.8.2/8.7.3) id XAA22641; Sat, 16 Aug 1997 23:35:02 +0100 (BST) Date: Sat, 16 Aug 1997 23:35:01 +0100 (BST) From: Jim Dixon X-Sender: jdd@avon-gw.uk1.vbc.net To: Tom Samplonius cc: freebsd-hackers@FreeBSD.ORG Subject: Re: Throughput on kernel routing rules? In-Reply-To: Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-hackers@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk On Sat, 16 Aug 1997, Tom Samplonius wrote: > > > > > Anyone ever done any tests to see how many packets per second FreeBSD > > > > > can route? > > > > > > > > We have seen something like 10-15Kpps. There is a group at MIT working > > > > > > On what kind of CPU? Since I guess such things are CPU bound... > > > > P133. It's I/O bound at these speeds. > > Depending on the number and speed of the interfaces, and the average > packet size. I am reporting observations. I am not theorizing. The CPU was nowhere near being saturated. > It is easy to get 10Kpps per second on a single full-duplex 10BT > interface. The machine I was talking about has three 100 Mbps ethernet interfaces. The fastest I have seen it running is a something more than 60 Mbps across all three interfaces. Let's say that's 8 MB/s; at packet sizes of 250 B that's 32 Kpps in and out, so it was running in the region of 16 Kpps forwarded. We were a little busy at the time and I wasn't that interested in pps rates; what I noted down was the bytes through the interface, converted to bps. > It would be best to test with minimum sized packets, to figure out how > many pps FreeBSD can really handle. This was real traffic. MIT -- the reference you clipped out -- has done a lot of testing. They are running a heavily modified version of FreeBSD and report much higher pps rates; they think that they can do 500 Kpps through the kernel. -- Jim Dixon VBCnet GB Ltd http://www.vbc.net tel +44 117 929 1316 fax +44 117 927 2015