Return to BSD News archive
Xref: sserve comp.os.linux.networking:10541 comp.os.ms-windows.networking.tcp-ip:13151 comp.unix.bsd.freebsd.misc:3847 Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!simtel!vtc.tacom.army.mil!agis!gatech!news.sprintlink.net!tank.news.pipex.net!pipex!dispatch.news.demon.net!demon!minnow.render.com!news From: dfr@render.com (Doug Rabson) Newsgroups: comp.os.linux.networking,comp.os.ms-windows.networking.tcp-ip,comp.unix.bsd.freebsd.misc Subject: Re: Internet service providing-which OS? Date: 24 Jul 1995 15:16:45 GMT Organization: RenderMorphics Ltd. Lines: 48 Message-ID: <DFR.95Jul24161646@trout.render.com> References: <3ue5qa$ain@panix.com> <id.VAPL1.0SA@nmti.com> <3ugeup$44k@news.unicomp.net> <3uk3b5$35a@legend.txdirect.net> <DC3sEM.1Jr@saturn.tlug.org> <tgmDC46x0.A4M@netcom.com> NNTP-Posting-Host: trout.render.com X-NNTP-Posting-Host: render.demon.co.uk In-reply-to: tgm@netcom.com's message of Sat, 22 Jul 1995 11:07:48 GMT In article <tgmDC46x0.A4M@netcom.com> tgm@netcom.com (Thomas G. McWilliams) writes: > Mike Frisch (mfrisch@saturn.tlug.org) wrote: > : Let's not forget ftp.cdrom.com which is a very busy (probably one > : of the busies on the 'net) ftp/www server running FreeBSD on a very large > : Pentium system. > > FreeBSD ftp.cdrom.com crashed earlier this year and was down for over a > week. The fallout was catastrophic. Anyone who would be so foolish as > to use FreeBSD as a mission critical server should ask the OS/2 > community which was schwacked hard by the loss of a major FTP site. > > Without a doubt, the crash of '95 was ftp.cdrom.com and FreeBSD. What > was the cost to the Internet community? What would be the cost to your > business to lose connectivity for a week? How "Free" is FreeBSD when > you factor in over a week of downtime? Face it, if you need 24 hour per > day, 365 day per year reliability, FreeBSD will *NOT* cut it. > The "crash of '95", as you put it, was caused by a massive hardware failure which took out some of its disks, I think. The choice of operating system had little to do with it, other than the possibility that other operating systems might not have driven the hardware so hard. Now if you are arguing that ftp.freebsd.org should have been running a fault tolerant RAID system which can survive major component failures without crashing, then you are welcome to contribute the code to support it to the project :-) I used to work on fault tolerant systems where a design goal was that no single component failure can crash the system and that all failed components are hot replaceable. These systems had four power supplies, each with its own battery backup, two I/O busses, twin disk controllers (one on each bus), and multiple mirrored disks. They could have up to 3 cpu cards (allowing a single bad cpu to be 'outvoted' and killed by the healthy ones), each of which had 2 68040s running Unix. You could walk up to one of these things when it was running and pull out one of its disks/cpus/cards etc. and all that would happen is that a red light would come on. I don't think that any PC based server can reach this level of reliability, whatever the OS. On the other hand, you can't buy them in rack-mount form factor either :-) -- Doug Rabson, Microsoft RenderMorphics Ltd. Mail: dfr@render.com Phone: +44 171 251 4411 FAX: +44 171 251 0939