OpenVZ Forum: Devel » [RFC 0/7] Initial proposal for faster res

Home » Mailing lists » Devel » [RFC 0/7] Initial proposal for faster res_counter updates

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

Re: [RFC 5/7] use percpu_counters for res_counter usage [message #45853 is a reply to message #45725]

Mon, 09 April 2012 01:48

KAMEZAWA Hiroyuki
Messages: 463
Registered: September 2006

Senior Member

(2012/03/30 22:53), Glauber Costa wrote:

> On 03/30/2012 11:58 AM, KAMEZAWA Hiroyuki wrote:
>> ==
>>
>> Now, we do consume 'reserved' usage, we can avoid css_get(), an heavy atomic
>> ops. You may need to move this code as
>>
>> rcu_read_lock()
>> ....
>> res_counter_charge()
>> if (failure) {
>> css_tryget()
>> rcu_read_unlock()
>> } else {
>> rcu_read_unlock()
>> return success;
>> }
>>
>> to compare performance. This css_get() affects performance very very much.
>
> thanks for the tip.
>
> But one thing:
>
> To be sure: it effectively mean that we are drawing from a dead memcg
> (because we pre-allocated, right?

Cached stock is consumed by the current task. It blocks removal of memcg.
It's not dead.

Thanks,
-Kame

Report message to a moderator

[Message index]

		[RFC 0/7] Initial proposal for faster res_counter updates By: Glauber Costa on Fri, 30 March 2012 08:04
		[RFC 5/7] use percpu_counters for res_counter usage By: Glauber Costa on Fri, 30 March 2012 08:04
		Re: [RFC 5/7] use percpu_counters for res_counter usage By: KAMEZAWA Hiroyuki on Fri, 30 March 2012 09:33
		Re: [RFC 5/7] use percpu_counters for res_counter usage By: KAMEZAWA Hiroyuki on Fri, 30 March 2012 09:58
		Re: [RFC 5/7] use percpu_counters for res_counter usage By: Glauber Costa on Fri, 30 March 2012 13:53
		Re: [RFC 5/7] use percpu_counters for res_counter usage By: KAMEZAWA Hiroyuki on Mon, 09 April 2012 01:48
		Re: [RFC 5/7] use percpu_counters for res_counter usage By: Glauber Costa on Fri, 30 March 2012 12:59
		[RFC 6/7] Add min and max statistics to percpu_counter By: Glauber Costa on Fri, 30 March 2012 08:04
		[RFC 7/7] Global optimization By: Glauber Costa on Fri, 30 March 2012 08:04
		[RFC 4/7] move res_counter_set limit to res_counter.c By: Glauber Costa on Fri, 30 March 2012 08:04
		[RFC 3/7] bundle a percpu counter into res_counters and use its lock By: Glauber Costa on Fri, 30 March 2012 08:04
		[RFC 2/7] consolidate all res_counter manipulation By: Glauber Costa on Fri, 30 March 2012 08:04
		Re: [RFC 0/7] Initial proposal for faster res_counter updates By: KAMEZAWA Hiroyuki on Fri, 30 March 2012 08:32
		Re: [RFC 0/7] Initial proposal for faster res_counter updates By: Glauber Costa on Fri, 30 March 2012 10:46

Previous Topic:	[PATCH] memcg: Do not open code accesses to res_counter members
Next Topic:	[PATCH v2 0/5] per-cgroup /proc/stat statistics

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Wed Aug 07 11:19:43 GMT 2024

Total time taken to generate the page: 0.04849 seconds