I got confirmation that the reason of above issue is memory allocation limit in 32-bit kernel (as expected).
I got confirmation that the reason of above issue is memory allocation limit in 32-bit kernel (as expected).
I wrote bofh of my MSc and PhD theis in LaTeX (as well some scientific paper) and many presentation, however, it's not so popular among chemists.
I also have my CV written in LaTeX. It's nice because gives special and fresh look and feel.
Thanks for great response. As each good response it lead to further question.
Do data cached by kernel have to fit into 1GB of kernel memory?
Is there different in handling this memory between 32bit and 64bit version of x86 arch?
I'm asking because on a busy 32-bit server (apache serving static content - from 2 or 3 week non-stop) I see:
~ > free
total used free shared buffers cached
Mem: 16633136 3370264 13262872 0 520812 2391332
-/+ buffers/cache: 458120 16175016
Swap: 1052248 88 1052160
My estimation is that if the system kept all served files in the cache the memory would filled in few hours. So there are some software limit of the memory usage.
Could you please confirm or denied that system call sendfile(), opposite to read()+write() operations, does not use memory cache at all.
What means that i.e. in an Apache serving many times the same static content turning EnableSendfile Off might bust the performance, because files would be read from memory rather then from disk.
Please check the link under the following article on my blog:
http://larryn.blogspot.com/2009/08/stone-redskin-comparision-of-apache2.html
BTW. In the article there are some test of Apache on HDD and SSD. Might be interested for some.
The Linux Foundation is a non-profit consortium dedicated to the growth of Linux.
Join / Linux Training / Board