On 13/11/2011 3:45 a.m., Igor Tandetnik wrote:
Josh Gibbs wrote:
To reiterate from my original question, if we don't add the order by then
the results come back at a very acceptable speed. But of course we'd
then have to iterate through the results ourselves to siphon off the
top item set that
On 13/11/2011 3:37 a.m., Igor Tandetnik wrote:
Josh Gibbs wrote:
The timestamps are already integers. We stumbled across that CAST operation
optimization purely by accident. I don't remember what led to it, but
we found
that it gave a measurable performance boost casting the integer as an
int
On 12 Nov 2011, at 2:37pm, Igor Tandetnik wrote:
> ANALYZE
Yeah. Do an ANALYZE. Then test timing with and without the CAST. See if it
helps.
I don't understand why your query should be faster when you suppress a useful
index. Oh wait. Your entire WHERE clause is about the joined table.
Josh Gibbs wrote:
> To reiterate from my original question, if we don't add the order by then
> the results come back at a very acceptable speed. But of course we'd
> then have to iterate through the results ourselves to siphon off the
> top item set that we're after. I'd really like the DB to d
Josh Gibbs wrote:
> The timestamps are already integers. We stumbled across that CAST operation
> optimization purely by accident. I don't remember what led to it, but
> we found
> that it gave a measurable performance boost casting the integer as an
> integer.
This works by *suppressing* the i
On 12/11/2011 5:02 p.m., Simon Slavin wrote:
On 12 Nov 2011, at 3:43am, Josh Gibbs wrote:
We are struggling to find a way to rework this query in a way
that performs efficiently for large data sets.
It's all about the indexes.
The problem with this query seems to be related to the aggregatio
On 12 Nov 2011, at 3:43am, Josh Gibbs wrote:
> We are struggling to find a way to rework this query in a way
> that performs efficiently for large data sets.
It's all about the indexes.
> The goal is to find the top x most active senders of e-mails
> within a date range.
Do you frequently do
Hi all,
We are struggling to find a way to rework this query in a way
that performs efficiently for large data sets.
The goal is to find the top x most active senders of e-mails
within a date range.
The killer of this query is the ORDER BY clause. Without it
the results are quick and snappy.
8 matches
Mail list logo