Re: Effect of changing the value for PARALLEL_TUPLE_QUEUE_SIZE

On Fri, Jun 2, 2017 at 9:15 AM, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
> On Fri, Jun 2, 2017 at 6:38 PM, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
>> On Fri, Jun 2, 2017 at 9:01 AM, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
>>> Your reasoning sounds sensible to me. I think the other way to attack
>>> this problem is that we can maintain some local queue in each of the
>>> workers when the shared memory queue becomes full. Basically, we can
>>> extend your "Faster processing at Gather node" patch [1] such that
>>> instead of fixed sized local queue, we can extend it when the shm
>>> queue become full. I think that way we can handle both the problems
>>> (worker won't stall if shm queues are full and workers can do batched
>>> writes in shm queue to avoid the shm queue communication overhead) in
>>> a similar way.
>>
>> We still have to bound the amount of memory that we use for queueing
>> data in some way.
>
> Yeah, probably till work_mem (or some percentage of work_mem). If we
> want to have some extendable solution then we might want to back it up
> with some file, however, we might not need to go that far. I think we
> can do some experiments to see how much additional memory is
> sufficient to give us maximum benefit.

Yes, I think that's important. Also, I think we still need a better
understanding of in which cases the benefit is there.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company