Re: [patch 2.6.12-rc1-mm4] fork_connector: add a fork connector

[Evgeniy Polyakov <johnpol@2ka.mipt.ru>]

[-- Attachment #1: Type: text/plain, Size: 1524 bytes --]

On Thu, 2005-03-31 at 16:10 -0800, Jay Lan wrote:
> Andrew Morton wrote:
> 
> >Guillaume Thouvenin <guillaume.thouvenin@bull.net> wrote:
> >  
> >
> >>  This patch adds a fork connector in the do_fork() routine.
> >>...
> >>
> >>  The fork connector is used by the Enhanced Linux System Accounting
> >>project http://elsa.sourceforge.net
> >>    
> >>
> >
> >Does it also meet all the in-kernel requirements for other accounting
> >projects?
> >
> >If not, what else do those other projects need?
> >  
> >
> Hi Andrew,
> 
> As the discussion in this thread of the past few days showed, this
> patch intends to take care of process grouping, but not the
> accounting data collection. Besides my concern of possibility of
> data loss, this patch also provides CSA information to handle process
> grouping as it intends to do. I plan to run some testing to see percentage
> of data loss when system is under stress test, but improvement at
> CBUS as Evgeniy indicated should help!
> 
> Please be advised that i still need an do_exit handling to save accounting
> data. But, it is a separate issue.

My five copecks [or two cents]: 
fork connector with CBUS [with theirs upto 2.5 % degradation
with huge disk writes per fork] are still much faster than any existing
accounting models.

But it is purely accounting project author to think about 
accounting design though...

> Thanks,
>  - jay
-- 
        Evgeniy Polyakov

Crash is better than data corruption -- Arthur Grabowski

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]