On 10/30/2009 12:43 PM, Merlin Moncure wrote:
> On Thu, Oct 29, 2009 at 11:40 AM, Steve Crawford
> <scrawford(at)pinpointresearch(dot)com> wrote:
>> Use a parent table and 20 child tables. Create a new child every day and
>> drop the 20-day-old table. Table drops are far faster and lower-impact than
>> delete-from a 120-million row table. Index-bloat is limited to one-day of
>> inserts and will be eliminated in 20-days.
>> Read up on it here:
> From a performance point of view, this is going to be the best option.
> It might push some complexity though into his queries to invoke
> constraint exclusion or deal directly with the child partitions.
Seeking to understand.... is the use of partitions and constraint-exclusion
pretty much a hack to get around poor performance, which really ought
to be done invisibly and automatically by a DBMS?
Much as indexes per se are, in the SQL/Codd worldview?
Or, is there more to it?
I appreciate the "Simple Matter Of Programming" problem.
In response to
pgsql-performance by date
|Next:||From: Jesper Krogh||Date: 2009-10-30 19:46:37|
|Subject: Re: Queryplan within FTS/GIN index -search.|
|Previous:||From: Robert Haas||Date: 2009-10-30 15:03:47|
|Subject: Re: sub-select in IN clause results in sequential scan|