Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!bbn!inmet!ishmael!cball From: cball@ishmael Newsgroups: comp.unix.wizards Subject: Re: rwhod Message-ID: <143300010@ishmael> Date: 14 Feb 89 13:06:00 GMT References: <89@ Lines: 31 Nf-ID: #R:karl@triceratops.cis.ohio-state.edu.U writes: >>roy@phri.UUCP (Roy Smith) writes: >> The N^2 effect isn't so bad when you've got >> 15 or 20 machines, but it kills you when you've got hundreds. My >> idea was to make rwhod write into /usr/lib/rwho instead of >> /usr/spool/rwho. >Better, run an rwhod which doesn't broadcast at all, but rather [a] >NFS-mount /usr/spool/rwho from a single server, and then [b] just >scribbles /usr/spool/rwho/its-own-name. Poof, no more broadcasts at >all. I have an rwhod(called lwhod) that does exactly this and works with Suns and Vaxes. I didn't use licensed sources so it's available, if there's any interest. While rwhod is a pig, I don't think it is really N**2. More like W**N where W is the number of workstations/server. With 200 workstations and a maximum of 10 diskless nodes/server the number should be less than 12 interrupts/sec(significant, but not enough to swamp a system). I suspect that there is a quantum syncronization effect when lots of machines boot simultaneously. Synronization probably even increases over time due to scheduling anomalies since rwhod wakes up on both sleep and remote packet events. As written, I think rwhod would work better if it slept some fixed time plus a random constant. Charles Ball Intermetrics, Inc. cball@inmet.inmet.com uunet!inmet!cball