What the optimum value of vfs.read_max and kern.ipc.nmbclusters for my hardware

abdurrahman

I have been using pfSense 2.0.3-RELEASE (amd64) on sunfire x4150 machine.
my hardware specifications are listed below
2 x Intel Core Quad CPU
16 GB RAM
5x146 GB 7200 RPM disk (RAID5)

I'm not using squid in pfSense. firewall property of pfSense is enoguh for my network…

In system Tunables, vfs.read_max = 32... According to my hardware, what is the optimum value for vfs.read_max?
In /boot/loader.conf, if I change kern.ipc.nmbclusters value from 0 to 32767 , will it increase performance...note that I'm not using squid
only firewall of pfsense...
In addition to these, what do you advise to me to increase the performance of pfSense?

Thanks in Advance

wallabybob

Unless there is something you haven't told us, it is unlikely you can do anything to the system to observably increase performance.

Is there some particular reason you are asking about increasing performance?

abdurrahman

i have 1000-1550 clients on LAN interface…because of the fact that the machine on which pfSense installed is very powerful,
ı want to take benefit ot he power of the machine...for that reason i want to increase performance....
for example what is the optimum value of vfs.read_max for my hardware? i wonder it

dhatz

Your best bet to increase overall performance would be to switch to pfsense 2.1-BETA (based on FreeBSD 8.3, with new drivers for most hardware). As long as you won't be doing any disk-intensive tasks (e.g. squid) the most important factor affecting performance would be your NIC(s) e.g. Intel or Broadcom etc.

Since you have plenty of RAM, you could try
kern.ipc.nmbclusters=262144

abdurrahman

thanks for your reply..
@dhatz:

Your best bet to increase overall performance would be to switch to pfsense 2.1-BETA (based on FreeBSD 8.3, with new drivers for most hardware). As long as you won't be doing any disk-intensive tasks (e.g. squid) the most important factor affecting performance would be your NIC(s) e.g. Intel or Broadcom etc.

Since you have plenty of RAM, you could try
kern.ipc.nmbclusters=262144