Hi,
I am quite fuzzle about how epoll get enabled and how can i disable it or enable it at will. Maybe someone have some insight about epoll could lend a hand.
I first install squid 2.6.STABLE4-20061009 on a box with redhat linux 3. The cache.log shows it use poll for IO loop. Then I get another box with redhat linux enterprise AS 4 update 3. I did same compile and installation of squid. When i start the squid, the cache.log show Using epoll for IO loop. So my questions are ..
1. Does my squid actually use epoll when it says 'Using epoll for IO loop'? If not then ignore the subsequence questions.
2. How the epoll is enable on my second box, given that i have the same code, installation steps. Does squid detect something in the OS and activate it by default. The only main difference of two server in term of kernel OS is the file discriptor. The first box has 1024 file descriptors, the second box has 32768 file descriptors.
3. How can I disable epoll? and reenable it later on. I want to see how much gain I get with epoll so I need some performance data when epoll is disable.
Thanks,
Khanh
Received on Wed Oct 18 2006 - 12:34:12 MDT
This archive was generated by hypermail pre-2.1.9 : Wed Nov 01 2006 - 12:00:04 MST