Messages in this thread | | | From | James Sutherland <> | Subject | Re: Overcommitable memory?? | Date | Tue, 21 Mar 2000 23:44:49 +0000 |
| |
On Tue, 21 Mar 2000 09:04:04 -0800 (PST), you wrote:
> >[CC list trimmed again, I doubt Stephen Tweedie or Rik van Riel are >interested in this discussion.] > >On Tue, 21 Mar 2000, James Sutherland wrote: > >>>Preventing system OOM using resource limits is equivalent to disabling >>>overcommit. You have to restrict each of N users to 1/N of the total >>>system memory. >> >>No. That is NOT overcommit. Overcommit, in this context, is when a >>process calls malloc() and is given unpopulated address space, which >>will be populated on use. > >In the quota case, in order to prevent a system-wide OOM you must give >each of N users an average of 1/N of the total system memory (ignoring >kernel overhead). The side effect is that overcommittment is now >impossible, because the system can only be overcommitted if a user has >exceeded their quota, which is not allowed... > >Unless you don't count COW pages against a user's quota?
No; overcommit involves something totally different. All the users could have malloc()ed 2Gb blocks of address space, provided they didn't USE it. (A sparse matrix, for example. A classic example of overcommit making code MUCH simpler.)
Overcommit doesn't involve quotas. It just involves demand-allocation of memory.
James.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |