Date: Tue, 28 Jul 2009 07:15:32 +0100 From: Mark Sandford <j.m.sandford@hotmail.co.uk> To: <freebsd-ipfw@freebsd.org> Subject: RE: Using dummynet to restrict bandwidth with more than 2 active pipes / queues Message-ID: <COL110-W1117070BA101EAA0AD722492150@phx.gbl> In-Reply-To: <COL110-W10D341161C0B631FE5239892140@phx.gbl> References: <COL110-W10D341161C0B631FE5239892140@phx.gbl>
next in thread | previous in thread | raw e-mail | index | archive | help
Sorry if anyone's wasted time looking at this. The problem appears to be w= ith the traffic generator. Once we get above two generation processes we t= hink that the data is being sent in bursts so although it appears to be rig= ht averaged over a second at a finer granularity the burstiness is meaning = it's either exceeding the bandwidth or idle at each point. ~~~~~~~~~~~~~~~~~~~~~~~~ Mark Sandford email: j.m.sandford@hotmail.co.uk mob: 07990 565976 ~~~~~~~~~~~~~~~~~~~~~~~~ > From: j.m.sandford@hotmail.co.uk > To: freebsd-ipfw@freebsd.org > Date: Mon=2C 27 Jul 2009 14:08:22 +0100 > Subject: Using dummynet to restrict bandwidth with more than 2 active pip= es / queues >=20 >=20 > Hi all=2C >=20 > =20 >=20 > I've been using dummynet for a while to perform degraded network testing = which has been really useful. >=20 > =20 >=20 > Recently=2C we wanted to measure the performance limits of it on our > hardware. To do this we setup a machine with 8 interfaces paired into 4 > ethernet bridges. >=20 > =20 >=20 > We are having throughput issues when more than 2 pipes are being used > simultaneously. These issues appear to be independent of the bandwidths > specified. >=20 > For example: > We set two traffic generators transmitting at 30Mbps across two of the br= idges (pipes)=2C sending 1000 byte UDP packets (1042 bytes on the wire) for= a 20 second period. >=20 > These are passed through dummynet pipes set up to restrict the bandwidth = to 20Mbps at the bridge and we can see from the ipfw counters that all the = packets hit the right rules and only the right rules. >=20 > We the capture on the far end and can see that bandwidth has been restric= ted to 20Mbps as specified. All good! :o) >=20 > The problem comes when we add any extra flows. >=20 > The above example is repeated but with two extra traffic generators trans= mitting at just one packet per second each across a further two pipes. >=20 > Again we can see from the counters that the packets all arrive at ipfw=2C= however we only get 10Mbps at the receiving end (and we get a number of pa= cket_drops logged at dummynet). >=20 > We feel we must have missed something obvious but after over a week of re= ading / testing we're running out of ideas. >=20 > Is anyone able / willing to help? >=20 > ~~~~~~~~~~~~~~~~~~~~~~~~ > Mark Sandford >=20 > email: j.m.sandford@hotmail.co.uk > mob: 07990 565976 >=20 > ~~~~~~~~~~~~~~~~~~~~~~~~ >=20 >=20 >=20 >=20 >=20 > _________________________________________________________________ > Celebrate a decade of Messenger with free winks=2C emoticons=2C display p= ics=2C and more. > http://clk.atdmt.com/UKM/go/157562755/direct/01/_________________________= ______________________ > freebsd-ipfw@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-ipfw > To unsubscribe=2C send any mail to "freebsd-ipfw-unsubscribe@freebsd.org" _________________________________________________________________ Windows Live Messenger: Celebrate 10 amazing years with free winks and emot= icons. http://clk.atdmt.com/UKM/go/157562755/direct/01/=
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?COL110-W1117070BA101EAA0AD722492150>