In my test.tcl when hitting nsd 5000 times i get 720 req/sec with zippy,
640 req/sec with vmalloc and 520 req/sec with ptmalloc. for some reason
nedmalloc crashes when i LD_PRELOAD it on my machine.
The nice thing is this:
after the test when nsd has 30 running threads:
vlad 17932 105 20.8 738768 215756 ? Ssl 18:46 0:47
/usr/local/ns/bin/nsd -i -
I have threadtimeout set to 300 to see the size when conn threads age
gone, below is the size without conn threads.
oss2:nscp 1> ns_server threads
{min 10} {max 100} {current 10} {idle 10} {stopping 0}
vlad 19912 0.0 5.4 241844 56548 ? Ssl 00:19 0:02
/usr/local/ns/bin/nsd -i -u
Zoran Vasiljevic wrote:
On 20.12.2006, at 01:04, Vlad Seryakov wrote:
I put together another thing to test, it runs fine with my Naviserver
and worked well with my old Tcl/Adp tests. Speed is less than zippy
but
better than Tcl or ptmalloc. But it does return memory back on thread
exit, this made my naviserver nsd process always run around 170M
instead
of 800M with zippy.
Very nice. I will have to integrate this in tclThreadAlloc.c.
The trick with LD_PRELOAD is nice but it doesn't work on MacOSX.
If I can get those sizes as you and if it is on-pair (it can be
slower but not 2 times more then zippy or nedmalloc) then it will
be the malloc of choice for us.
-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
naviserver-devel mailing list
naviserver-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/naviserver-devel
--
Vlad Seryakov
571 262-8608 office
[EMAIL PROTECTED]
http://www.crystalballinc.com/vlad/
