Date: Sun, 21 Jun 1998 12:07:00 -0400 (EDT) From: Jamie Bowden <jamie@itribe.net> To: Open Systems Networking <opsys@mail.webspan.net> Cc: hackers@FreeBSD.ORG Subject: Re: TweakDUN Message-ID: <Pine.SGI.3.96.980621120030.18254A-100000@animaniacs.itribe.net> In-Reply-To: <Pine.BSF.3.95.980620013654.20127D-100000@orion.webspan.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 20 Jun 1998, Open Systems Networking wrote: > On Fri, 19 Jun 1998, spork wrote: > > > I can't think of anywhere this is true. I'll use our dialup pools as an > > example: > > > > modem-> dialup PPP 1500 -> term server -> ethernet 1500 -> router -> T1(s) > > HDLC 1500 -> core router -> fast ethernet 1500 -> upstream's border > > router -> FDDI 40?? -> upstream core router -> ATM/SONET/whatever ? > > > > Generally, one avoids small MTUs on big links, I beleive. ATM's small > > cell size makes *every* packet get fragmented at layer 2, but I'm not sure > > that's even relevant. > > > > Anyone else? I've never heard of the oft quoted "Internet standard MTU of > > 576"... > > Me either, and theres a reason. Its braind dead to do so. Im REALLY tired > right now so im not gonna crack open stevens and quote it. Anyone using an MTU > of 576 destroys performance on bulk transfers such as FTP and and the > like. The ONLY good an MTU of 576 is good for is interactive traffic, like > telnet. At least thats how I have always understood it. For an excercise > the reader can change his MTU from 1500 to 576 and test FTP, and telnet > over it, then switch back to 1500 and do the same. If you do the > calculations of an MTU of 576 and figure in latency. On bulk transfers of > an MTU of 1500 winds over 576. Because for each 1500 byte packet sent over > the MTU link with say a latency of 533ms for dialup, it would take 533ms > for a 1500 byte packet. Now if your using an MTU of 576 and transfer the > same 1500 byte packet with latency of 533ms, you have to send 3 packets of > 536 bytes @ 533ms latency each. It isnt rocket science to see why for > anything BUT interactive traffic where all your traffic will fit inside > your MTU your gonna loose and loose big. So why in gods name people think > an MTU of 576 is the "internet standard" AND their actually believing it > astonishes me. My math above maybe be incorrect but you get the idea. > Basically I chalk it up to uninformed admins on the loose, who need to be > put back in their cages. :) > Not to mention by fragmenting all your non-interactive traffic your > creating insane ammounts of traffic. 3 packets at an MTU of 576 or 1 > packet at 1500. tripple the traffic. Well you can tell im tired cause im > ranting and probably making no sense. So im going to bed. > Ill probably kick myself for what I just wrote when I read it in the > morning but I am sure some of its right. The reason you lower MTU and MRU on modem links is the problem with your media. It's unreliable to transmit large packets cleanly. 576 is probably too low, but keeps your retransmit levels to a bare minimum. Line noise on modems is an unfortunate reality. Large packet sizes are more likely to be hit by this and have to be retransmitted. I use an MRU of 1024 at home, and find it to be about the best for overall throughput. The best method for this is changing the MRU/MTU on your links, and running tests. The cleaner your pnone line, the higher you can reliably go. I have friends who run their MRU at 768 because that's what works best for them. For an ISP to force everyone to the LCD is unacceptable. Get a new ISP. Jamie Bowden Systems Administrator, iTRiBE.net To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.SGI.3.96.980621120030.18254A-100000>