It seems to me one solution is to alter your table topology by
partitioning your table by the keys you need to query on, and then
using simple aggregates.
You;d have to set up ON INSERT DO INSTEAD rules, and you might get a
performance hit.
Another solution might be to break up the query int
On Tue, Jul 5, 2011 at 9:49 AM, Alban Hertroys
wrote:
> On 5 Jul 2011, at 9:13, Daniel Farina wrote:
>
>>> Setup a materialized view.
>>
>> This rather defeats the point AFAIK, because keeping the materialized
>> view up to date (being more than thirty seconds out of date is not
>> desirable) will
On 5 Jul 2011, at 9:13, Daniel Farina wrote:
>> Setup a materialized view.
>
> This rather defeats the point AFAIK, because keeping the materialized
> view up to date (being more than thirty seconds out of date is not
> desirable) will be expensive. Maintaining the index on the (key,
> recency)
On 05/07/11 11:48, Daniel Farina wrote:
This is basically exactly the same as
http://archives.postgresql.org/pgsql-sql/2008-10/msg9.php; I'm
just asking again, to see if thinking on the problem has changed:
The basic problem, restated, is one has a relation with tuples like this:
(key, rece
On Tue, Jul 5, 2011 at 12:32 AM, Simon Riggs wrote:
> I think its a pretty common requirement and we should be looking to
> optimize it if it isn't handled well.
I agree; although I wanted to be sure that it is not in fact handled
well by some mechanism I haven't seen yet.
> The only problem is
On Tue, Jul 5, 2011 at 12:48 AM, Daniel Farina wrote:
> This is basically exactly the same as
> http://archives.postgresql.org/pgsql-sql/2008-10/msg9.php; I'm
> just asking again, to see if thinking on the problem has changed:
>
> The basic problem, restated, is one has a relation with tuples
On Mon, Jul 4, 2011 at 11:55 PM, Alban Hertroys
wrote:
> On 5 Jul 2011, at 3:23, David Johnston wrote:
>
>>> Does anyone have fresh thoughts or suggestions for dealing with
>>> INSERT-mostly tables conceived in this manner?
>
> You're struggling with read-performance in an INSERT-mostly table? Whe
On 5 Jul 2011, at 3:23, David Johnston wrote:
>> Does anyone have fresh thoughts or suggestions for dealing with
>> INSERT-mostly tables conceived in this manner?
You're struggling with read-performance in an INSERT-mostly table? Where are
your performance priorities, on INSERT or on SELECT?
>
On Jul 4, 2011, at 19:48, Daniel Farina wrote:
> This is basically exactly the same as
> http://archives.postgresql.org/pgsql-sql/2008-10/msg9.php; I'm
> just asking again, to see if thinking on the problem has changed:
>
> The basic problem, restated, is one has a relation with tuples lik
