Re: [GENERAL] very large tables

2009-05-25 Thread Alban Hertroys
On May 25, 2009, at 10:58 AM, Ramiro Diaz Trepat wrote: The table with the atmosphere pixels, currently has about 140MM records, and the one the values about 1000MM records. They should both grow to about twice this size. Did you tune postgres to use the available resources? By default it

Re: [GENERAL] very large tables

2009-05-25 Thread Laurent Wandrebeck
Hello, I'd try to use postgis ( http://postgis.refractions.net ), that allows you to geolocalize your data. We use such an extention here for the same kind of data you use. It will lighten up table size, and queries. Otherwise, partitioning and cluster are good things to do. I would increase defau

Re: [GENERAL] very large tables

2009-05-25 Thread Matthew Brand
You might want to try CREATE CLUSTER. I had a 40M row table that was taking ages to access, I tried partitioning it into 12000 sub-tables, and obtained a modest speed up. Using CREATE CLUSTER on an un-partitioned table resulted in an enormous speed up though. You will need to choose the axis you w

Re: [GENERAL] very large tables

2009-05-25 Thread Otandeka Simon Peter
Try partitioning. It should sort you out. -Peter- On 5/25/09, Ramiro Diaz Trepat wrote: > > Hello list,I will try to make this as brief as possible. > I have a brother who is a scientist studding atmospheric problems. He was > trying to handle all of his data with flat files and MatLab, when I

Re: [GENERAL] Very large tables

2008-11-28 Thread Ron Mayer
Grzegorz Jaśkiewicz wrote: true, if you don't want to search on values too much ,or at all - use float[]. But otherwise, keep stuff in a tables as such. It might be humongous in size, but at the end of the day - prime thing when designing a db is speed of queries. If he's worried about speed

Re: [GENERAL] Very large tables

2008-11-28 Thread William Temperley
On Fri, Nov 28, 2008 at 5:46 PM, Simon Riggs <[EMAIL PROTECTED]> wrote: > > I would look carefully at the number of bits required for each float > value. 4 bytes is the default, but you may be able to use less bits than > that rather than rely upon the default compression scheme working in > your f

Re: [GENERAL] Very large tables

2008-11-28 Thread Simon Riggs
On Fri, 2008-11-28 at 15:40 +, William Temperley wrote: > Hi all > > Has anyone any experience with very large tables? > > I've been asked to store a grid of 1.5 million geographical locations, > fine. However, associated with each point are 288 months, and > associated with each month are 5

Re: [GENERAL] Very large tables

2008-11-28 Thread Grzegorz Jaśkiewicz
2008/11/28 William Temperley <[EMAIL PROTECTED]> > > Any more normalized and I'd have 216 billion rows! Add an index and > I'd have - well, a far bigger table than 432 million rows each > containing a float array - I think? > > Really I'm worried about reducing storage space and network overhead >

Re: [GENERAL] Very large tables

2008-11-28 Thread Alvaro Herrera
William Temperley escribió: > Really I'm worried about reducing storage space and network overhead > - therefore a nicely compressed chunk of binary would be perfect for > the 500 values - wouldn't it? An array that large would likely be compressed by Postgres internally for storage; see http://w

Re: [GENERAL] Very large tables

2008-11-28 Thread Scara Maccai
Really I'm worried about reducing storage space and network overhead - therefore a nicely compressed chunk of binary would be perfect for the 500 values - wouldn't it? For storage space you might want to look at ZFS with compression on in case you are using FreeBSD or Solaris. That would s

Re: [GENERAL] Very large tables

2008-11-28 Thread William Temperley
On Fri, Nov 28, 2008 at 3:48 PM, Alvaro Herrera <[EMAIL PROTECTED]> wrote: > William Temperley escribió: >> So a 216 billion row table is probably out of the question. I was >> considering storing the 500 floats as bytea. > > What about a float array, float[]? I guess that would be the obvious cho

Re: [GENERAL] Very large tables

2008-11-28 Thread Grzegorz Jaśkiewicz
On Fri, Nov 28, 2008 at 3:48 PM, Alvaro Herrera <[EMAIL PROTECTED]>wrote: > William Temperley escribió: > > > I've been asked to store a grid of 1.5 million geographical locations, > > fine. However, associated with each point are 288 months, and > > associated with each month are 500 float values

Re: [GENERAL] Very large tables

2008-11-28 Thread Alvaro Herrera
William Temperley escribió: > I've been asked to store a grid of 1.5 million geographical locations, > fine. However, associated with each point are 288 months, and > associated with each month are 500 float values (a distribution > curve), i.e. 1,500,000 * 288 * 500 = 216 billion values :). > >