Scalability? - Feedback from C Matthew Curtin

Fri, 7 Aug 1998 08:08:19 -0400 (EDT)


great article.

i found your comments about not being able to reach the tesco site funny. microsoft had the same problem on june 20 of last year:

for the graphs for internet/intranet networking features, reliability and scalability, and system management (that came from dec?), is there any description of the methodology used to rate them? the ratings seem potentially arbitrary, and might be used to critique your argument, potentially as being biased and willing to include anything that will help your case. publishing (or providing a link for) what an os would need to do to quality for "excellent" vs "ok" would be helpful and overcome any possible objection, except one with the methodology itself.

i'm actually a bit surprised that you've never seen a kernel panic. i've seen a few, but i'm a little more unix heavy than the average administrator or user. (i developed firewalls at bell labs back before there were commercial firewalls, and have done system development, and have 8 unix boxen at home.) you might be interested to know that in every case of kernel panic in production that i've seen it has been because of hardware problems. the most common is a memory parity error. but even the "most common" isn't that common. i've got a sparc2 (not an ultrasparc2) that's acting as my home internet gateway and firewall bastion host, running openbsd and just happily chugging along with 235 days of uptime. 235 days ago, i installed openbsd and booted the system. since that time, i have made many changes in the system's kernel, including turning routing on and off, changing access control lists, bringing interfaces up and down on various networks, all without a reboot or the least bit of instability.

i hope some of this turns out to be useful (or at least worth reading) for you.

Matt Curtin

 <= Back to the Feedback Index

 <= Back to the article