BearItAll wrote:
> Roy Schestowitz wrote:
>
>> Linux: The Completely Fair Scheduler
>>
>> ,----[ Quote ]
>> | He explained, "this project is a complete rewrite of the Linux
>> | task scheduler. My goal is to address various feature requests
>> | and to fix deficiencies in the vanilla scheduler that were
>> | suggested/found in the past few years, both for desktop
>> | scheduling and for server scheduling workloads."
>> `----
>>
>> http://kerneltrap.org/node/8059
>>
>
> I must say that I didn't know we had a problem with the scheduler. It is
> true that the needs of the server are different to the needs of a desktop
> in this particular area. Also that a more dynamic scheduler would very
> likely be of use, though it has to be said that there is some dynamic
> properties of the scheduler already, maybe there is room for improvement,
> but except for in the more extreme situations load tables don't suggest a
> need.
>
> I had my doubts whether this scheme would be the direction though, no more
> heuristic dynamic allocation, instead a single, well sorry but my
> subconcious did this of it's own account, it's a quicker reader than I am
> so read ahead and replaced the line with 'a single bottle neck'.
>
> Sometimes you know a thing is wrong, a product, an idea, because you know
> a better way, but other times you don't know a better way but the thing
> has the wrong feel to it. This particular idea on the scheduler doesn't
> feel right. I think it is the wrong direction.
I have seen some problems with the scheduler in some of the later kernels.
It is not yet as bad as the Windows scheduler for allowing rogue apps to
kill the OS, but I have seen a few cases where it has come close.
Ian
|
|
