Re: Memory overcommitting (was Re: http://www.redhat.com/redhat/)

bofh@snoopy.virtual.net.au
Sun, 23 Feb 97 20:09:30 +1000


>> I'd like to see this as a per process option. Should this option
>> be inherited across fork/clone/exec?

>Overcommit control is an all or nothing thing, otherwise one process can
>overcommit anothers space..

I was under the impression that what they were talking about in this thread was the idea of having a set of swap space reserved for a process and not allowing memory allocation of any form if there was no swap space to back it up. This means that the process in question wouldn't be able to exceed physical storage (if you have 16meg RAM and 32meg swap then a 50meg malloc() is guaranteed to fail), and if another process starts committing large amounts of memory that it had allocated then our process won't be interfered with, so it wouldn't have to be "an all or nothing thing".

However I could have misunderstood some of what they were saying, I wasn't paying a lot of attention to this thread. In any case I believe that the method I describe in the previous paragraph will work.

Russell Coker