home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!dtix!darwin.sura.net!uvaarpa!murdoch!fermi.clas.Virginia.EDU!gl8f
- From: gl8f@fermi.clas.Virginia.EDU (Greg Lindahl)
- Newsgroups: alt.irc
- Subject: Re: IRC STATS ?
- Message-ID: <1992Jul24.192506.12927@murdoch.acc.Virginia.EDU>
- Date: 24 Jul 92 19:25:06 GMT
- References: <1992Jul24.180953.4073@ctr.columbia.edu>
- Sender: usenet@murdoch.acc.Virginia.EDU
- Organization: Department of Astronomy, University of Virginia
- Lines: 18
-
- In article <1992Jul24.180953.4073@ctr.columbia.edu> dougmc@ccwf.cc.utexas.edu (Doug McLaren) writes:
-
- >It seems to me that the best way to measure a server's usage would be to just
- >report it's average number of connections, to both servers and clients. If
- >we want to be fair to hubs, we could say that a server connection is worth
- >twice (an arbitrary number ...) as much as a client connection.
-
- But we want to be fairest to hubs that serve leaves that have lots of
- users -- remember, the point here is to serve users. I'd suggest
- counting the number of users on a server as the first measure, and as
- a second, counting all the users on links to the server other than the
- biggest number. Report both numbers separately. This will give hubs
- that are "serving" parts of the tree with essentially no users a low
- number, and hubs that serve parts of the tree with lots of users a very
- large number.
-
- The hubs with few secondary users and large amounts of outgoing bytes
- are a burden, not a benefit.
-