home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.unix.admin:5021 comp.windows.x:16624
- Newsgroups: comp.unix.admin,comp.windows.x
- Path: sparky!uunet!sci34hub!tybrin4!tybse1!swhite
- From: swhite@tybse1.uucp (William C. "Spike" White)
- Subject: Re: Xterminal-Server ratio wanted
- Organization: Tybrin Corporation, Shalimar, FL
- Date: Sat, 12 Sep 1992 20:13:12 GMT
- Message-ID: <1992Sep12.201312.3019@tybse1.uucp>
- References: <1992Sep1.200609.15078@progress.com> <1836euINNaqs@early-bird.think.com>
- Lines: 65
-
- In article <1836euINNaqs@early-bird.think.com> barmar@think.com (Barry Margolin) writes:
- >In article <1992Sep1.200609.15078@progress.com> tucker@bedford.progress.COM writes:
- >> We are currently performing an initial investigation into upgrading
- >>our present dumb-terminal based user environment to a X windows based one.
- >>We are weighing all the dis/advantages of PCs, Xterminals and workstations
- >>and look favorably at Xterminals from a maintenance standpoint. I use one
- >>myself and see no lack of functionality. We would like any opinions on this
- >>decision itself. And if we do go with Xterminals, then the question of how
- >>many Xterminals running off of how many Sparcs, Sparc2s, Solbournes, etc. We
- >>would like to hear about any success/horror stories about implementing such
- >>a configuration and performance issues regarding Xterm/Server ratios which
- >>seem to work well. These users will mainly be running Progress applications
- >>and some word-processing, news readers and email UAs.
- >
- >What are "Progress applications"?
-
- Applications built using the Progress database?
-
- >
- >My impression is that except for heavily graphics-oriented applications, an
- >X terminal isn't much worse than a dumb terminal. There's more overhead
- >for keyboard input, but since that only happens at human typing rates, it's
- >not really significant. Output has some extra overhead, but unless you're
- >doing it continuously (e.g. for animation), it's generally not enough to
- >cause problems. As far as the network is concerned, I'd guestimate about
- >20-50% overhead for using X terminals over using dumb terminals (assuming
- >the dumb terminals were on the network via a terminal concentrator). I've
- >never noticed any network congestion problems that I could attribute to X.
- >
- >As for server overhead, that also seems to be pretty low. We some SS2's
- >and multiprocessor SPARCservers with connections to two dozen X servers (a
- >mix of X terminals, Sun workstations, and Macs running MacX) as well as
- >some dumb terminals and Macs running Telnet, and at least one 4-processor
- >690MP with connections to over 50 X terminals. The main reason that X
- >terminals tend to put more load on a server is that users tend to run more
- >simultaneous applications; they'll have an xclock and xbiff running all the
- >time, and they'll have several windows open rather than exit one
- >application in order to start another. You may need more swap space to
- >hold all these active processes.
- >
- >If you're mainly going to be using text-based applications as you
- >described, I think X terminals are definitely the way to go.
-
- Here's a dissenting opinion.
-
- We ran an X11R2-based X-windows/Open Look system on a heavily-loaded database
- server. Even though we had a single X-terminal, we had to pull it off
- because the database performance would be terrible whenever any user used
- the X-terminal. The users weren't doing anything fancy, just two xterms and
- 1 graphics application. Things would have been different if:
- 1. We had a modern version of X-windows (>= X11R4).
- 2. The server wasn't so overworked.
- 3. The server was faster (it was about a 10 MIPS machine).
-
- In my opinion, X-terminals place a significant load on the server's resources
- and should be figured in. NCD has an excellent white paper called "Host
- Loading in an X-terminal environment". It has great rules of thumb for
- figuring the load on the server and the network.
-
-
- --
- Spike White Tybrin Corporation, Shalimar, FL | Moderation in all things --
- spike.white@tybrin.com | and abstinence in none!
- Disclaimer: The guys down the hall disagree with everything I say! Guess
- who speaks for the company.
-