Re: [GENERAL] Performance options for CPU bound multi-SUM query

2016-01-27 Thread David Rowley
On 28 January 2016 at 08:41, Matt wrote: > Moving from NUMERIC to FLOAT(8) did indeed lower query times by about 20%. > > I will try fixeddecimal and agg() as time permits. That's surprisingly little gain. Please note that you'll not gain any further improvements from the fixeddecimal type than y

Re: [GENERAL] Performance options for CPU bound multi-SUM query

2016-01-27 Thread Matt
Moving from NUMERIC to FLOAT(8) did indeed lower query times by about 20%. I will try fixeddecimal and agg() as time permits. On 25 Jan 2016, at 4:44, David Rowley wrote: On 25 January 2016 at 15:45, Matt wrote: I have a warehousing case where data is bucketed by a key of an hourly timesta

Re: [GENERAL] Performance options for CPU bound multi-SUM query

2016-01-25 Thread David Rowley
On 25 January 2016 at 15:45, Matt wrote: > I have a warehousing case where data is bucketed by a key of an hourly > timestamp and 3 other columns. In addition there are 32 numeric columns. The > tables are partitioned on regular date ranges, and aggregated to the lowest > resolution usable. > > Th

Re: [GENERAL] Performance options for CPU bound multi-SUM query

2016-01-25 Thread Andreas Kretschmer
Matt wrote: > I have a warehousing case where data is bucketed by a key of an hourly > timestamp and 3 other columns. In addition there are 32 numeric columns. The > tables are partitioned on regular date ranges, and aggregated to the lowest > resolution usable. > > The principal use case is to