Karl Denninger writes:
> You don't want Bay and you certainly don't want 3Com. If your network
> is fairly slow (ethernets and T1s only) you can use PCs running a
> reasonable BSD and GateD. Otherwise, the only commercial choice is
> Cisco.
FreeBSD and decent networking cards (ie: Intel PRO100Bs) can route a couple
of 100Mbps switched fast ethernets (yes, full duplex too)
Yeah, but you'll have trouble getting decent T3 cards for it. Its true
that 100Mbps ethernet should be fine.
BTW, NetBSD with the recent flow cache mods can handle at least
150,000 packets per second. We haven't seen what the actual upper
limit is, but that number doesn't seem to be eating a lot of CPU.
Contrary to popular belief, national networks are not simple to set up in a
way which will insure that they have maximum survivability and performance.
Amen.
.pm
perry@piermont.com (Perry E. Metzger) writes:
Yeah, but you'll have trouble getting decent T3 cards for it. Its true
that 100Mbps ethernet should be fine.
SDL makes decent (tho not perfect) T3 cards for PCs.
BTW, NetBSD with the recent flow cache mods can handle at least
150,000 packets per second. We haven't seen what the actual upper
limit is, but that number doesn't seem to be eating a lot of CPU.
Interesting. Have you done any scalability testing? Per-flow state has
been shown to scale poorly in Internet backbones.
Tony
> Yeah, but you'll have trouble getting decent T3 cards for it. Its true
> that 100Mbps ethernet should be fine.
SDL makes decent (tho not perfect) T3 cards for PCs.
So does <URL:http://www.lanmedia.com/products.htm>.
> BTW, NetBSD with the recent flow cache mods can handle at least
> 150,000 packets per second. We haven't seen what the actual upper
> limit is, but that number doesn't seem to be eating a lot of CPU.
Interesting. Have you done any scalability testing? Per-flow state has
been shown to scale poorly in Internet backbones.
That was with one flow. The idea of "there basically aren't flows" is not
obvious on first approach to these problems, wouldn't you say, Tony :-)?
paul@vix.com (Paul Vixie) writes:
> > BTW, NetBSD with the recent flow cache mods can handle at least
> > 150,000 packets per second. We haven't seen what the actual upper
> > limit is, but that number doesn't seem to be eating a lot of CPU.
>
> Interesting. Have you done any scalability testing? Per-flow state has
> been shown to scale poorly in Internet backbones.
That was with one flow. The idea of "there basically aren't flows" is not
obvious on first approach to these problems, wouldn't you say, Tony :-)?
Not obvious to some, but we sat and watched a significant router vendor
make that mistake (despite good advice to the contrary) anyhow. I'd hate
to see someone else repeat that mistake.
Tony