Cell Relay Archive

[vjs@calcite.rhyolite.com (Vernon Schryver)]

In article <3lr0nt$5s5@hpindda.cup.hp.com> raj@cup.hp.com writes:
>Vernon Schryver (vjs@calcite.rhyolite.com) wrote:
>
>[It's entirely too easy to attack marketing blurbs, just go to a web
> server with .com in its name to find examples:) so I've deleted that
> part - raj]

Web server marketing blurbs do not arrive unasked for.  Advertising
blurbs in netnews and (unsolicited) email are exactly like postage-due
junk mail, while web blurbs are like brochures you pick up at trade shows.
Postage-due junk mail does not sound like a good way to build market share.

Yes, with the advent of AOL, CompuServ, even if you ignore the small
number of people like me who buy our own news feeds, most people
pay money to read netnews.  Thust, netnews arrives postage due.


>: How fast does it go, as measured by `ttcp` or `netperf`?
>
>A fair question - after I see the product announcement of anyone's new
>card, I know that I, individually, as the "shameless promoter of
>netperf and the netperf database," :) will be encouraging the folks
>developing/selling that card to submit netperf numbers to the database
>at http://www.cup.hp.com/netperf/NetperfPage.html. That encouragement
>is for *both* throughput (TCP_STREAM) *and* latency (TCP_RR) tests.
>
>I have had some success in the area of ATM - some numbers from DEC,
>and from HP (the EISA ATM Adaptor), but most other folks in .com
>(nudge, nudge :) have been avoiding it.
>
>Even today in the increasingly enlightened '90's, precious few folks
>submit latency numbers - they seem to prefer to stick with throughput
>numbers :(

Most of us still have applications that care more about throughput
than latency, and do not yet have many if any applications that care
about latency.  Besides, it takes time to port and run benchmarks.


Vernon Schryver    vjs@rhyolite.com