Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!zaphod.mps.ohio-state.edu!think.com!yale.edu!ox.com!math.fu-berlin.de!fauern!faui43.informatik.uni-erlangen.de!eckert From: eckert@immd4.informatik.uni-erlangen.de (Toerless Eckert) Newsgroups: comp.dcom.sys.cisco Subject: Re: Aggregate thru-put of an AGS+ Message-ID: <1991Jun14.181125.18693@informatik.uni-erlangen.de> Date: 14 Jun 91 18:11:25 GMT References: <13670001@hprnd.rose.hp.com> Organization: CSD., University of Erlangen, Germany Lines: 55 From article <13670001@hprnd.rose.hp.com>, by pugh@hprnd.rose.hp.com (Rex Pugh): > __________________________________ > | AGS+ | > |_________________________________| > | _____ | > | / Y \ | > | ^ v | > | Slot0 =0===1===2===3===4===5= | > | v ^ | > | \__R__/ | > |_________________________________| > | | > | Ethernet | > | Stream Input Output Drop% | > | ----------------------------- | > | Red 14531 9023 38% | > | Yellow 14454 9015 38% | > | ------------------- | > | 28985 18038 38% | > | | > | 802.3 | > | Stream Input Output Drop% | > | ----------------------------- | > | Red 14531 749 95% | > | Yellow 14447 742 95% | > | ------------------- | > | 28978 1491 95% | > | | > | SNAP | > | Stream Input Output Drop% | > | ----------------------------- | > | Red 14531 749 95% | > | Yellow 14447 742 95% | > | ------------------- | > | 28978 1491 95% | > |_________________________________| What i don't understand is what the relevance of the numbers is. Is this meant to measure how fast the box can drop packets ? Why are there no numbers for which the box had a maximum throughput WITHOUT dropping packets, you only had this information for the first two or so diagrams ? The number of packets on the outputs will most surely be higher for the maximum number of input packets that the box can forward without dropping packets. From my understanding (and measurements on different equipment) you can have quite a sharp negative response onto packet overload on input, so the output rate can drop rapidly if you overload the box. I don't see this to be a big problem, because upper layers have to cope with overload anyway, so a negative characteristic in the router might even speed that process up ? --- Toerless.Eckert@informatik.uni-erlangen.de /C=de/A=dbp/P=uni-erlangen/OU=informatik/S=Eckert/G=Toerless bandwidth - the final frontier.