Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!sdd.hp.com!elroy.jpl.nasa.gov!ncar!asuvax!hrc!gtephx!wilsonj From: wilsonj@gtephx.UUCP (Jay Wilson) Newsgroups: comp.sys.apollo Subject: Re: The Hardy Few Summary: NFS vs NS Message-ID: <335@texas.gtephx.UUCP> Date: 17 Apr 91 15:03:48 GMT References: <1991Apr6.072832.4012@cs.ubc.ca> Distribution: comp,world Organization: gte Lines: 52 I have been watching this little discussion for awhile now and up until now I have found little to disagree with. Well I have a problem with the comments that follow: > Are Suns really that bad? Try getting "gdb" and the X11R4 release > from MIT. If you find bugs, report them. As for NFS, it has at least > two advantages over the Apollo file system: > > a) It is stateless. A curse and a blessing, I must admit, but at least > diskless machines don't crash if the server does. On the other hand, > you do have to live with "NFS server not responding..." occasionally. > > b) It seems to be portable. A lot of systems have some NFS compatibility, > probably because of the public specification of the protocol and its > relative independence on a particular underlying file system. 1) Suns are really that bad. Over the last year we have lost 3 sys_admins to a new company here in town building a Sun network of 300 workstations. I keep in contact with these guys and they wish they were back supporting Apollos. 2) NFS has more problems than "NFS server not responding". Five years ago when we decided to go with Apollo workstations instead of Sun workstations, the Apollo won out because of ease of connectivity. Originally we were going to have a network of 500 workstations and everyone had to have the ability to talk to everyone else. Well we tested a small network of Suns and a small network of Apollos. Based on the amount of time needed to across mount all of the Sun systems, it would take 500 workstations 1 1/2 months to get everyone talking to everyone else if we took a power hit. The Apollos on the other hand would take about 10 mins. Additionally the management aspect of NS is much easier than NFS. With NFS you have to update each workstation's tables from that workstation. With NS you update using ctnode (from any workstation) and all of the other nodes will be able to find the node (if the address is not hashed already). We have a cron job on every node that refreshes the hash table twice aday. This is still better than going to every workstation and removing a mount point. As a side note we at one point had 900+ workstations and 8 sys_admins (112+:1). Over the last year we have scaled back to 583 workstations and 6 sys_admins (97:1). Lets see Sun sys_admins support that many workstations. ---------- Jay Wilson (wilsonj@gtephx) UUCP: {ncar!noao!asuvax | uunet!zardoz!hrc | att}!gtephx!wilsonj AG Communication Systems, Phoenix, AZ voice (602) 581-4496 fax (602) 581-4967 As usual, the comments above are solely mine and not necessarily that of AG Communication Systems.