fred@mot.UUCP (Fred Christiansen) (11/25/85)
to date, all Ethernet LAN's at this location have been modest: 2 or 3 systems connected for some development project. we're about to connect these up to make a 15 node network. several folk (don't laugh, now) are scared this will already be saturation. they're comparing it to a disk drive with a 10 Mbits/sec bandwidth and how many users it will support. i've argued (from gut feel) that the traffic patterns are totally different. what i haven't said yet in these arguments is that 10 Mbits/sec is never really reached in an Ethernet LAN. real throughput seems to flatten out out a much lower rate, like 3 or 4 Mbits/sec (my recall on what i've read is a big foggy). if i say that, it'll really frighten 'em. could someone with genuine experience with TCP/IP or XNS LAN's tell me what the practical number of nodes is on a LAN in an engineering organization? also, can anyone point to papers/studies which discuss the experienced throughput limitations of Ethernet? thanks! -- << Generic disclaimer >> Fred Christiansen ("Canajun, eh?") @ Motorola Microsystems, Tempe, AZ UUCP: {seismo!terak, trwrb!flkvax, utzoo!mnetor, ihnp4, attunix}!mot!fred ARPA: oakhill!mot!fred@ut-sally.ARPA "Families are Forever"
darrelj@sdcrdcf.UUCP (Darrel VanBuer) (11/28/85)
In my experience with a variety of workstations and VAXen is that most systems become CPU bound by the time they are using from 100,000 to 300,000 bits per second (and I mean for fairly simple, but real, tasks like file transfer). What this means that it takes more than 50 machines to saturate an Ethernet doing real work. A recent note from ISI in ACM SIGCOMM reported results of monitoring one of their Ethernet's traffic. A net with some 70 assorted workstations and mainframes had some interesting statistics. The day-long average traffic was under 3000 bits per second and even in the two hour peak of midafternoon, the average traffic was around 25,000 bits per second. Ethernet runs at 10 million bits per second not because anyone believes you can use that much bandwidth, it's so that there is so much excess capacity that it can provide service for almost any transient need. In most systems disk bandwidth is not the limiting factor on transfers from disk either, but seek and rotational latency are. -- Darrel J. Van Buer, PhD System Development Corp. 2525 Colorado Ave Santa Monica, CA 90406 (213)820-4111 x5449 ...{allegra,burdvax,cbosgd,hplabs,ihnp4,orstcs,sdcsvax,ucla-cs,akgua} !sdcrdcf!darrelj VANBUER@USC-ECL.ARPA