Everything I did for Server 1.0 still works for 2.0
Here's my setup for Debian Etch:
General tuning. Add the following files to /etc/sysctl.conf:
vm.dirty_background_ratio = 5
vm.dirty_ratio = 100
Tune network performance. Add the following files to /etc/sysctl.conf:
net.core.wmem_max = 16777216
net.core.rmem_max = 16777216
net.ipv4.tcp_timestamps = 0
net.ipv4.tcp_rmem = 4096 262144 16777216
net.ipv4.tcp_wmem = 4096 262144 16777216
net.core.optmem_max = 524288
net.core.netdev_max_backlog = 200000
Tune HD performance. Add the following lines to /etc/rc.local (change sda to your hd):
echo "64" > /sys/block/sda/queue/max_sectors_kb # This depends on the block size of your RAID controller
echo "8192" > /sys/block/sda/queue/nr_requests
blockdev --setra 16384 /dev/sda
Add the following line to /etc/fstab (create the directory /tmp/vmware first
):
tmpfs /tmp/vmware tmpfs defaults,size=100% 0 0
VMware setup (add the following lines to /etc/vmware/config):
tmpDirectory = "/tmp/vmware"
mainMem.useNamedFile = "FALSE"
sched.mem.pshare.enable = "FALSE"
MemTrimRate = "0"
MemAllowAutoScaleDown = "FALSE"
prefvmx.useRecommendedLockedMemSize = "TRUE"
prefvmx.minVmMemPct = "100"
After all that VMware Server is really fast (at least on my hosts).
It has only one disadvantage. After "redirecting" the mmap file to the RAM the sum of all guest RAM is limited to the real RAM minus the VMware overhead. That's because the mmap file always uses always the full RAM specified for the guest, but I can live with that.
BTW: This setup doesn't double the RAM usage.