Negative caching

[Hal Burgiss]

On Fri, Jun 19, 2009 at 02:09:42PM -0400, Rashkae wrote:
> Hal Burgiss wrote:
> 
> > 
> > But here is what I don't understand, and why I think there is a fundamental
> > problem with the way swap is managed with this particular kernel: this issue
> > gets progressively worse over time. It will probably take at least a week,
> > maybe two, before I starting thinking evil thoughts of restarting X or the
> > system. Its like the physical memory at some point is just not utilized well.
> > The stuff I use frequently and interact with, winds up on disk, and "something
> > else" has physical memory bottlenecked. 
> > 
> 
> I would counter that running desktop with over 3GB of allocated memory
> on a system with only 1GB of physical ram for weeks before it becomes
> unbearable is a testament to how finely tuned the memory system really
> is.  I notice that not only do you shoot way past my estimate of 2GB,
> but you are flirting with your Commit limit, at which point the kernel
> might start to even refuse further allocation. (Sorry, I forget which
> tunables control this behaviour, but either way, the situation is dire).
>  You'll have to either reduce your memory footprint or add more ram.

Thank you sir. I feel better understanding. I'll dig in and find the pig(s). 

-- 
Hal