> Brad Brad wrote: > > Scp's between the guest and host only manage about 5KB/s so I tried > > going back to le which worked great. I configured a new > kernel with > > "disable pcn*" but on next boot I had no nics at all, so i > tried again > > "disable pci*" also since I think le is isa, but it still > didn't work. > > That's strange that pcn is giving you worse performance than > le. I've > had performance problems under VMware (ESX 2.x) where the > host's NIC is > hard-set speed and duplex, and the guest winds up with a ton of > collisions, and setting the duplex in the guest doesn't make a > difference. Maybe that's the case here? > > You could always try using vmxnet/vic. >
I'll second the use of vic(4). I had the same issue with a pcn(4) showing large amounts of collisions, and terrible throughput. With the addition of the vic(4) driver, the collisions have gone away, and throughput is much better. Just remember to change the NIC vmware emulates to vmxnet, as Steve pointed out. -- Craig
