Date: Fri, 5 Sep 2003 14:42:45 -0700 From: Yuchung Cheng <ycheng@cs.ucsd.edu> To: Luigi Rizzo <rizzo@icir.org> Cc: ipfw@freebsd.org Subject: Re: dummynet and modem pipes Message-ID: <20030905144245.A472@cs.ucsd.edu> In-Reply-To: <20030905112459.A78583@xorpc.icir.org>; from rizzo@icir.org on Fri, Sep 05, 2003 at 11:24:59AM -0700 References: <20030905095038.D28924@cs.ucsd.edu> <20030905112459.A78583@xorpc.icir.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On 09-05-2003, Luigi Rizzo wrote: > basically there is no way to accurately simulate modem compression. > As a rule of thumb, for bulk traffic you can think that the main > effect is the bandwidth changes (the exact ratio depends on the > type of traffic), whereas for interactive apps (e.g. rpc and the > like) more than compression what affects you is the additional > delay that the modem causes to decide that it is time to build&send > a new packet. > > also you'd need to take care of ip compression, not just what > happens in the modem itself > i assume you mean ip header compression over ppp. the simulation traffic is mainly short web flows (~ 10-20K), so i think header compression does not matter much (40 bytes vs payload 500 or 1500 bytes) compared to modem compression. for html, the compress rate can be as high as 3 based on real measurements. it that true? thanks. -yuchung
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030905144245.A472>