sob@harvisr.harvard.edu (Scott Bradner) (02/16/89)
We did some simple throughput testing of a cisco router that some of you might find interesting. We were going to run the same tests on other vendor equipment but things conspired to keep us from doing so. We had gotten the loan of the backup router from one of the area universities but the day before we were to run the tests one of their production units went south ... We did run the tests on one other vendor but it does not seem fair to just post that as the only thing other than the cisco so the numbers here are for the cisco AGS/2-4EM running V7.0 of the roms. We hope to be able to get back to this and do some more tests on the cisco and other routers. We don't know the impact of adding the rejection of selected ip addresses or protocols etc. We would like to be able to test the performance in various DECnet or bridging modes, maybe someday. Scott Bradner Harvard University test setup: TX--------X--------XT TX--------------XT ------ | | | ------------ | | ------ | hp1|=== | ====| test box |====== ====| hp2| ------ --------- ------------ ------ | ex | --------- hp1 - HP Lan Protocol analyzer 4972A monitoring the LANalizer output hp2 - HP Lan Protocol analyzer 4972A monitoring the test box output ex - Exelan type 5000 LANalizer on Compact 386 T - terminator X - tranceiver == - tranceiver cable | - tranceiver cable -- - thinnet The test packet was an IP packet. The LANbridge 100 was in bridge mode (the only mode it has ) The cisco router was in ip router mode. ( The LANbridge was included as a control, it should be able to pass all packets received. ) size ipi kb/sec % pps LB cwin cbet 64 900 633 6.38 1040 1040 1040 1040 600 920 9.26 1513 1513 1513 1513 300 1682 16.85 2767 2767 2767 2767 100 3127 31.51 5142 5142 5142 5142 "0" 6277 63.44 10323 10323 10323 5914 128 900 1106 11.15 988 988 988 988 600 1572 15.84 1404 1404 1404 1404 300 2715 27.36 2424 2424 2424 2424 100 4558 45.94 4069 4069 4069 4069 "0" 7598 76.58 6784 6784 6784 4343 256 900 1923 19.40 897 897 897 897 600 2630 26.52 1227 1227 1227 1227 300 4163 41.96 1942 1942 1942 1942 100 6158 62.07 2873 2873 2873 2873 "0" 8584 86.52 4004 4004 4004 2930 512 900 3278 31.99 758 758 758 758 600 4110 41.45 981 981 981 981 300 5824 58.68 1389 1389 1389 1389 100 7581 76.42 1809 1809 1809 1777 "0" 9222 92.92 2199 2199 2199 1773 1024 900 4790 48.30 579 579 579 578 600 5800 58.42 700 700 700 700 300 7340 73.99 885 885 885 885 100 8608 86.78 1040 1040 1040 992 "0" 9590 96.65 1157 1157 1157 991 1514 900 5760 58.05 471 471 471 471 600 6713 67.53 549 549 549 549 300 8025 80.86 657 657 657 657 100 9012 90.84 738 738 738 697 "0" 9710 97.86 797 797 797 695 size packet size in bytes as told to the Exelan We think that the size does not include the preamble or CRC. ipi Inter Packet time in micro sec as told to the Exelan "0" was as fast as it could go. kb/sec data rate in bytes per second as reported by the HP % load on ethernet as reported by the HP pps packets per second as reported by the HP LB DEC LANbridge 100 cwin cisco gateway with "quad" board, software v7.0 - within a single board cbet cisco gateway with "quad" board, software v7.0 - between two boards (NOTE: I've checked over the transcription of these numbers more than once but there are an awful lot of them and I make no absolute guarantee. )