On Thu, May 15, 2014 at 11:24 PM, Bruce Momjian <bruce@momjian.us> wrote:
> On Thu, May 15, 2014 at 10:23:19PM +0900, Amit Langote wrote:
>> On Thu, May 15, 2014 at 9:06 PM, Bruce Momjian <bruce@momjian.us> wrote:
>> >
>> > This is the same problem we had with auto-tuning work_mem, in that we
>> > didn't know what other concurrent activity was happening. Seems we need
>> > concurrent activity detection before auto-tuning work_mem and
>> > effective_cache_size.
>> >
>>
>> Perhaps I am missing something obvious here, but would mmgr have any
>> useful numbers on this? Like any book-keeping info maintained by
>> mcxt.c/aset.c? Would extending that interface help?
>
> No, all memory allocat is per-process, except for shared memory. We
> probably need a way to record our large local memory allocations in
> PGPROC that other backends can see; same for effective cache size
> assumptions we make.
>
I see. I thought there would be some centralised way to traverse, say,
a linked list of contexts that individual backends create or something
like that. But, I suppose it would not be straightforward to make any
of that work for what we are after here.
--
Amit