Thread: Linux memory handling improvement

Heads up: it appears that the Linux stable kernel series is about to get a
sane VM system from Andrea Arcangeli, which has proper page accounting and
does not have an Out Of Memory killer at all.

This will probably take a while to make its way into vendor kernels, and
even then we'll need to keep the warnings in the docs for people running
older kernels. I am not sure at this stage what its status is for the 2.6
kernel series.

I will be keeping an eye on this.

cheers

andrew

On Mon, 2003-09-15 at 07:32, Andrew Dunstan wrote:
> This will probably take a while to make its way into vendor kernels, and
> even then we'll need to keep the warnings in the docs for people running
> older kernels. I am not sure at this stage what its status is for the 2.6
> kernel series.

The 2.6 kernel series uses a VM written by someone else (Rik van Riel),
so I don't think that 2.4 VM improvements are relevant to it. But it's
definitely a good thing that the -AA VM improvements are finally being
merged into the mainline 2.4 kernel.

-Neil

Neil Conway wrote:

>On Mon, 2003-09-15 at 07:32, Andrew Dunstan wrote:
>  
>
>>This will probably take a while to make its way into vendor kernels, and
>>even then we'll need to keep the warnings in the docs for people running
>>older kernels. I am not sure at this stage what its status is for the 2.6
>>kernel series.
>>    
>>
>
>The 2.6 kernel series uses a VM written by someone else (Rik van Riel),
>so I don't think that 2.4 VM improvements are relevant to it. But it's
>definitely a good thing that the -AA VM improvements are finally being
>merged into the mainline 2.4 kernel.
>  
>

Really? I haven't been following that closely. I thought it was 
basically the Arcangeli VM with the van Riel rmap stuff. Anyway, Joseph 
Pranovich's "Wonderful World of Linux 2.6 says this:

There is one further stability issue that has been resolved with Linux 
2.6: it is no longer possible to allocate more than the maximum amount 
of RAM (plus swap) you have on a system. Previously, Linux would allow 
the malloc() ("memory allocation") system call to succeed in some cases, 
even when memory is exhausted. The overcommitment logic has been revised 
and this case should now be impossible. (Of course, if you run out of 
RAM on the system-- even without exceeding the maximum-- you have worse 
problems to worry about.)

(see http://kniggit.net/wwol26.html )

So from a PostgreSQL p.o.v. we should be in good shape with any luck.

However, doing a little more reading I'm not quite as confident as I was 
that we are getting an improvement in 2.4. See 
http://kerneltrap.org/node/view/805

cheers

andrew