I am using squid like Internet proxy server on RHEL 4 update 6 & 8 with quite heavy load i.e. 8k established connections during peak hour. Without depending much on application provider's expertise I want to achieve maximum o/p from linux. W.r.t. that I have certain questions as following:

  1. How to find out if there is scope for further tcp fine-tuning (without exhausting available resources) as the benchmark values given by vendor looks poor! Is there any parameter value that is available from OS / network stack that will show me the results.
  2. If at all there is scope, how shall I identify & configure OS tcp stack parameters i.e. using sysctl or any specific parameter
  3. Post tuning how shall I clearly measure performance enhancement / degradation ?

The sysctl values listed in the page below are similar to other places I remember reading that cover solving C10k or C100k problems:


(search for "Tuning the Linux Kernel" on the page)

  • Thanks for your reference. I wanted to know if there is exact method to measure results after tuning /proc or sysctl parameters? Besides I am not sure how to find out if there is any scope available to fine-tune system? – Vijay Gharge Jun 6 '10 at 7:35
  • Depends on what you're looking to test. You mainly seemed to point out 8k active connections at peak, so you want to increase any resource limits that would make 8k active connections safe. These include sysctl and ulimit values (ulimit -n for max file descriptors) – Jordan Sissel Jun 6 '10 at 9:44
  • Yeah i did that but I am still trying to understand how shall I find out if there is still any scope for server fine-tuning? How exactly I shall use IO, nic utilization, cpu utilization etc to find out max possible output from my server? – Vijay Gharge Jun 12 '10 at 19:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.