Greg Smith <greg@2ndquadrant.com> writes:
> You might note that only one of these sources--a backend allocating a
> buffer--is connected to the process you want to limit. If you think of
> the problem from that side, it actually becomes possible to do something
> useful here. The most practical way to throttle something down without
> a complete database redesign is to attack the problem via allocation.
> If you limited the rate of how many buffers a backend was allowed to
> allocate and dirty in the first place, that would be extremely effective
> in limiting its potential damage to I/O too, albeit indirectly.
This is in fact exactly what the vacuum_cost_delay logic does.
It might be interesting to investigate generalizing that logic
so that it could throttle all of a backend's I/O not just vacuum.
In principle I think it ought to work all right for any I/O-bound
query.
But, as noted upthread, this is not high on the priority list
of any of the major developers.
regards, tom lane