home *** CD-ROM | disk | FTP | other *** search
- Organization: Chemical Engineering, Carnegie Mellon, Pittsburgh, PA
- Path: sparky!uunet!cis.ohio-state.edu!news.sei.cmu.edu!fs7.ece.cmu.edu!crabapple.srv.cs.cmu.edu!andrew.cmu.edu!dl2p+
- Newsgroups: alt.irc
- Message-ID: <UePpo3q00XQGAEO35f@andrew.cmu.edu>
- Date: Thu, 23 Jul 1992 21:29:39 -0400
- From: Douglas Allen Luce <dl2p+@andrew.cmu.edu>
- Subject: Re: Call for submission on layout of stats.
- In-Reply-To: <1992Jul23.100457.9157@alf.uib.no>
- References: <9220518.26456@mulga.cs.mu.OZ.AU>
- <1992Jul23.100457.9157@alf.uib.no>
- Lines: 46
-
- buboo@alf.uib.no (Ove Ruben R Olsen) writes:
- > the cure writes:
- >
- > >
- > > Server Name [ Outgoing/s] [ Incoming/s]
- > > [Byte] [Pckt] [Byte] [Pckt]
- > >
- > > CSULX.Weber.EDU [0283] [0007] [0089] [0002]
- >
- > In such stat you must have the following:
- >
- > Servername Days+hours:min:sec-open outgoingBytes incomingBytes
-
- If the point is to determine how many servers on the network are
- "necessary" for IRC operation, or how many of the servers on the
- net are "unnecessary," I don't think even this will be enough.
-
- It could be that a single "vanity" server has convinced two uplinks to
- not L-line it, and through some trick of the net, is positioned
- between two equally loaded network halves. This could show up in your
- statistics at the top of the list. And a highly-client-burdened, but
- leafed server (like a server acting as the link to a public client)
- may show up fairly far down the list.
-
- Perhaps what is needed is a better definition of what makes a server
- "necessary," and how one might generate statistics that are meaningful
- in this way.
-
- You might find that keeping accurate statistics difficult with servers
- going up and down all the time. Real statistics should be done by
- each server. Number of links, number of users, number of channels the
- server is attuned to, how much of the traffic is human generated, how
- much is server generated, server downtimes, and traffic generated by
- "netsplits" instead of by "normal operation" might all be dominant
- factors in deriving a useful "server importance" equation (not even
- counting physical network location and connectivity!).
-
- Straight counting of users seems to be the current tradition in
- gauging the importance of any particular server. While this might
- show just how much IRC is used in various locations (or how well
- clients are installed at the particular location), it's not much help
- in deciding routing.
-
- Douglas Luce
- CMU IRC
-
-