home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.unix.wizards:5227 comp.unix.admin:6643
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!darwin.sura.net!sgiblab!sti!sti!not-for-mail
- From: westes@smurf.sti.com (Will Estes)
- Newsgroups: comp.unix.wizards,comp.unix.admin
- Subject: Client-server vs. Host-based computing: performance comparisons sought
- Date: 14 Dec 1992 19:58:18 -0800
- Organization: Mail Group
- Lines: 27
- Message-ID: <1gjl4qINN9b1@smurf.sti.com>
- NNTP-Posting-Host: smurf.sti.com
-
- I'm comparing the effects on system load of two different user
- population profiles:
-
- a) N users connected to SunOS via SLIP, and using POP to read mail,
- SMTP to send mail, and NNTP to read and post news. All user
- agents for this case are running on the client PC/Mac/UNIX, and
- the SunOS server is only used via POP, SMTP, and NNTP.
-
- b) N users connected to SunOS via a telnet connection, using elm to
- read mail, vi/emacs to edit mail, SMTP to send mail, and
- trn/nn/rn to read and post news. In this case, both clients and
- servers reside on the SunOS server.
-
- For the sake of argument, let's assume that in case a), all users are
- directly connected to a router via 9600 bps modems that is then
- ethernet-connected to the UNIX box. In case b), all users are dialing
- into 9600 bps terminal servers, which are then ethernet-connected to the
- UNIX box.
-
- Which case - a) or b) - puts a greater load on the SunOS server host,
- and what are the reasons? Can someone quantify the difference in the
- user loads in the two cases, assuming 50% of the traffic is mail and 50%
- of the traffic is news, and that there are no other loads on the SunOS
- server?
-
- --
- Will Estes Internet: westes@netcom.com
-