Jordi Espasa Clofent wrote: > Yes Karl, you've the reason. > > I've searched in the archives of the present list and I've found a lot > of useful info about the initial question. > > At present moment I consider the next tips: > > * sk(4)-based NICs have the best performance > * one dedicated data-bus per NIC (I think a PCI-E x8/x16 will be > enough in my environment) > * OpenBSD 4.2 PF's (it seems the performance has been improved > until 50%!!! in relation to previous releases) > > So, now I've one additional question > > ¿Can the hard-disk performance to mean a bootle-neck in the PF-based > FW box? I understand the major part of PF operations are made in RAM > memory, but maybe I'm not right.
pf performance is essentially independent of disk performance, with the possible exception of loading rules. Russell
