Hi Yes, your query basically calls for a linear scan with PCRE pattern matching. To understand where time goes within a query, you can prepend it with TRACE. regards, Martin On 15/10/15 20:44, Paul Ramsey wrote:

Hi all,

I’ve loaded a relatively small table (I think) of 30M records, 7 columns, and I’m doing some test queries summarizing and filtering the table, like this:

select vendor_id, count(*) from taxi where geohash like '90efbc6a6fc%’ group by vendor_id;

On my 4-core iMac, the CPU doesn’t get very hot (20% on top, so it’s not even lighting up one of the four CPUs) , but it’s hard to tell since the return time is only 2.5 second.

Is that as fast as things will get? I doubled the table size and query times more-or-less doubled. I was just hoping to get query times in the < 1s range.

Thanks,

P

-- *Paul Ramsey* Solutions Engineer

*CartoDB—Map Your World's Data* 143 Roebling St 2A, New York, 11211, USA T: +1.917.463.3232 tel://19174633232 Skype: pwramsey3 W: www.cartodb.com http://www.cartodb.com/

Live sessions to learn CartoDB http://cartodb.com/webinars/

_______________________________________________ users-list mailing list users-list@monetdb.org https://www.monetdb.org/mailman/listinfo/users-list