Date: Tue, 21 Jun 2005 10:29:54 -0700 From: Luigi Rizzo <rizzo@icir.org> To: Pieter de Boer <pieter@thedarkside.nl> Cc: freebsd-net@freebsd.org Subject: Re: Issues with a Large Fat pipe Network simulation Message-ID: <20050621102954.A66904@xorpc.icir.org> In-Reply-To: <42B84AC8.7050802@thedarkside.nl>; from pieter@thedarkside.nl on Tue, Jun 21, 2005 at 07:13:44PM %2B0200 References: <42B722EF.2090203@thedarkside.nl> <20050620135044.B35720@xorpc.icir.org> <42B731CD.1040104@thedarkside.nl> <20050621075247.D63359@xorpc.icir.org> <42B84AC8.7050802@thedarkside.nl>
next in thread | previous in thread | raw e-mail | index | archive | help
oh yes one thing... you are using 'via foo0' in your rule, which means the packet is intercepted both in the input and output path, which causes further contention on the queues. try 'pipe 1 in recv foo0 ...' which should only intercept traffic in the input path. also you can set the queue size in kbytes, and can probably go as high as 1000kb. maybe this helps too. I am pretty sure there is some issue there, also related to some timing issues and tcp window opening mode (slow start vs. linear) cheers luigi On Tue, Jun 21, 2005 at 07:13:44PM +0200, Pieter de Boer wrote: > Luigi Rizzo wrote: > > >>However.. when I deleted the pipe rules on 'network', the speed suddenly > >>went up to around 800mbit/s too! I remade them, and voila, 200mbit/s. > > network emulation is a tricky job :) > It sure is, so I'm happy you're trying to help out :) > > > in any case i believe what happens is the following. > > > > The pipe has a default size of 50 slots, which at 1500 bytes is > > little above 64k. If the sender is bursting a large number of packets, > > it may well overflow the pipe's queue causing a backoff (which > > may simply be immediate, or delayed, depending on how you configure > > various things). > > > > I believe setting the queue size in the pipe to a value larger than > > the window should fix things. > I had the same thought, so I already fiddled with it a bit. Because you > brought it up I tested the following this evening: > send/recv spaces at 128KB > > 00001: unlimited 0 ms 50 sl. 1 queues (1 buckets) droptail > 00002: unlimited 0 ms 50 sl. 1 queues (1 buckets) droptail > > I'm getting 300-400mbit/s (which is higher than yesterday; it seems the > speed creeps up a bit after a while). > > > 00001: unlimited 0 ms 100 sl. 1 queues (1 buckets) droptail > 00002: unlimited 0 ms 100 sl. 1 queues (1 buckets) droptail > > I'm getting 300-400mbit/s. > > There doesn't seem to be a direct relation between the pipe's queuing > slots and the throughput. Setting the send/recvspaces to 65535 again > does give me an immediate throughput of >800mbit/s, though. > > > Hope you still have some other ideas, since I'm a bit puzzled here.. > > -- > Pieter > _______________________________________________ > freebsd-net@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-net > To unsubscribe, send any mail to "freebsd-net-unsubscribe@freebsd.org"
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050621102954.A66904>