I compiled and installed a new kernel for HEIWA to bump up the amount of memory dedicated to networking.

Periodically we'd been getting "Mbuf cluster allocation failed" kernel messages and networking would hang for a couple of minutes. HEIWA is running FreeBSD 4.4, prior to the introduction of dynamically sized kernel parameters. I increased maxusers to 256 and set NMBCLUSTERS to 32768. We should see better performance out of this now.

Periodically run netstat -m to see what the network memory statistics look like. Previously we'd maxed out on mbufs and mbuf clusters and had gotten a large number of requests denied and a smaller number delayed. We should see no requests denied or delayed and we shouldn't ever have a peak buffer or cluster usage that hits the max.