Re: Kernel 2.6.24 branch and cpulimit [message #38070 is a reply to message #37987] |
Fri, 13 November 2009 14:15 ![Go to previous message Go to previous message](/theme/ovz3/images/up.png) |
maratrus
Messages: 1495 Registered: August 2007 Location: Moscow
|
Senior Member |
|
|
I'd like to share some thoughts about CPULIMITS.
What purpose do we pursue restricting VEs CPULIMIT? Not allow a particular VE to use a free CPU? So, in case of intensive CPU consumption on the part of VEs, CPUUNITS are the main restriction that balances the system load. In case of low CPU consumption we would be likely restricted by CPULIMITS (if they don't equal to zero) while actually CPU would be idle. Thus, it's not clear what kind of benefit we'll obtain by setting up CPULIMIT parameters. Moreover, the situation may be occur when a particular VE hold some kind of locks (and other VEs wait these locks too) but cannot release them because of the CPULIMIT restriction. But it worth saying that in case of CPULIMITS=0 this situation may occur too (hence, with less probability). So, if somebody wants to set up CPULIMIT parameters it's of course not prohibited but it worth bearing in mind the above arguments.
|
|
|