Date: Fri, 24 May 1996 22:48:11 -0400 From: dennis@etinc.com (Dennis) To: Andrew McRae <amcrae@cisco.com> Cc: hackers@freebsd.org Subject: Re: The view from here (was Re: ISDN Compression Load on CPU) Message-ID: <199605250248.WAA02360@etinc.com>
next in thread | raw e-mail | index | archive | help
andrew McRae writes.... > >I have also looked at the FreeBSD code, and written some of it. >The biggest problem with FreeBSD is not the software, but the >crappy hardware - and I don't just mean badly designed or >badly made, but the architecture is somewhat lacking. >It seems to be getting better with PCI, and PCI >is also a technology that cisco is using for their own I/O >adaptors. But even then, we use a different connector, and have >extra pins so that hot swap can be supported. And even then >we are pushing the technology to its limit. > >Saying that cisco engineers write better code is a specious >argument, and totally ignores the hardware area. Again, *I* >know, because I work with cisco hardware engineers every day, >and I know the effort that goes into *both* hardware and >software to make products viable. I was being cynical. I think most of ciscos stuff is mediocre at best, particular their wan protocol implementation. > >Sure enough, at the low end it is a price war, and perhaps >cisco doesn't own 75% of the market; the access market will >*always* price sensitive, and not as performance and function >driven as mid and high end. But on the other hand, if your >access routers fall over occasionally, or you don't mind >powering it off to replace or install new line cards, then >it's not the end of the world to have downtime. PC's can be used in >this environment, and FreeBSD is a really good fit here. perhaps true, but you get a whole lot more bang for your buck. The big joke is that most equipment is obsolete in a year or less, so the hardware issue is really almost moot. Cisco doesnt have a low-end product that can compete with a pentium performance wise, and price-wise the pentium is a much higher performance choice until you get into high port densities. > >For mid range and certainly high end, I can guarantee you >will not be routing 100,000 packets per second on a PC, >however much memory and disc you put in it :-) >And for the core routers of the Internet or in a large >organisation, I suspect that you *would* be fired if >you tried to use anything except a serious dedicated router. you can come pretty close, but not with standard O/Ss. But i've never claimed that PCs were good candidates for backbone routers. Its just the 2500s and the 4000s that can be replaced. > >C'mon guys, use the right tool for the job. Don't tell me >you can replace routers with PC's. I would like to see the >PC that can sustain routing of over a million packets per second >like a fully loaded 7513 can. On the other hand, I have yet >to see a router run Doom... no, but 30 PCs can, for about the same cost as a 7513. Can a 2500 route a million packets per second? Why not just scrap it then, since its clearly inadaquate by your own standards? and lets be real. Cisco's performance numbers have always been theoretical. They can't do anything close to what the specs say. Funny how the numbers are always just about dead on the theoretical bus maximum. Perhaps you should check with your own company...Ive heard they've acquired a few PC card companies.... Dennis ---------------------------------------------------------------------------- Emerging Technologies, Inc. http://www.etinc.com Synchronous Communications Cards and Routers For Discriminating Tastes. 56k to T1 and beyond. Frame Relay, PPP, HDLC, and X.25 for BSD/OS, FreeBSD and LINUX
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199605250248.WAA02360>