Re: [RFC][PATCH 5/8] RSS controller task migration support

[Balbir Singh <balbir@in.ibm.com>]

Patrick.Le-Dot wrote:
> On Fri, 17 Nov 2006 22:04:08 +0530
>> ...
>> I am not against guarantees, but
>>
>> Consider the following scenario, let's say we implement guarantees
>>
>> 1. If we account for kernel resources, how do you provide guarantees
>>    when you have non-reclaimable resources?
> 
> First, the current patch is based only on pages available in the
> struct mm.
> I doubt that these pages are "non-reclaimable"...

I am speaking of a scenario when we start supporting kernel accounting
and of-course the swapless case.

> 
> And guarantee should be ignored just because some kernel resources
> are marked "non-reclaimable" ?
> 

Ok.. but can you have a consistent guarantee definition with un-reclaimable
kernel resources? How do you define a guarantee in a consistent manner?
In my discussions earlier on lkml, I had suggested that we define guarantee
only for reclaimable resources and provide support only for them.

> 
>> 2. If a customer runs a system with swap turned off (which is quite
>>    common),
> 
> quite common, really ?

Yep, I was listening to a talk from a customer service expert and he
mentioned that it's used to boost performance.

> 
>>             then anonymous memory becomes irreclaimable. If a group
>>    takes more than it's fair share (exceeds its guarantee), you
>>    have scenario similar to 1 above.
> 
> That seems to be just a subset of the "guarantee+limit" model : if
> guarantee is not useful for you, don't use it.
> 
> I'm not saying that guarantee should be a magic piece of code working
> for everybody.
> 
> But we have to propose something for the customers who ask for a
> guarantee (ie using a system with swap turned on like me and this is
> quite common:-)
> 

Like I said I am not against guarantees, but do we have to implement
them in our first iteration?


> Patrick
>


-- 

	Balbir Singh,
	Linux Technology Center,
	IBM Software Labs
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/