Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watnot!watmath!clyde!cbatt!ucbvax!LBL.ARPA!nagy%43198.hepnet
From: nagy%43198.hepnet@LBL.ARPA.UUCP
Newsgroups: mod.computers.vax
Subject: RE:DECServer Load Balancing Formula
Message-ID: <8704110525.AA03369@ucbvax.Berkeley.EDU>
Date: Fri, 10-Apr-87 16:13:06 EST
Article-I.D.: ucbvax.8704110525.AA03369
Posted: Fri Apr 10 16:13:06 1987
Date-Received: Sun, 12-Apr-87 00:55:41 EST
Sender: daemon@ucbvax.BERKELEY.EDU
Distribution: world
Organization: The ARPA Internet
Lines: 16
Approved: info-vax@sri-kl.arpa

Well, I don't know what the formula is, but it works pretty well.  The
one exception seems to be with systems that are 100% busy because
batch jobs are ALWAYS running in the background and eating up that
idle time.  

We have a 2 node LAVC (both MicroVAXes) and get the users fairly
well split between the two systems.

One point to beware of when you try it: if your systems are lightly
loaded AND you a trying "load balancing" by logging in another
session from your terminal, you will find that the DECServer's
bias to add a new terminal to an existing virtual circuit will
cause your 2nd session to go to the same VAXCluster node (this is
done to more efficiently use the communications channel - the
controller and CPU on the host - by doubling up characters for
two terminals in the same Enet packets).