Houman, Partition by date and revise your processes to create and load a new child table every day. Since you already know the date append it to the table base name and go straight to the data you need. Also, the index on T.c won't help for this query, you're looking at a full table scan every time. Bob Sent from my iPhone On Oct 15, 2012, at 3:59 PM, houmanb <houman@xxxxxx> wrote: > Dear all, > We have a DB containing transactional data. > There are about *50* to *100 x 10^6* rows in one *huge* table. > We are using postgres 9.1.6 on linux with a *SSD card on PCIex* providing us > a constant seeking time. > > A typical select (see below) takes about 200 secs. As the database is the > backend for a web-based reporting facility 200 to 500 or even more secs > response times are not acceptable for the customer. > > Is there any way to speed up select statements like this: > > SELECT > SUM(T.x), > SUM(T.y), > SUM(T.z), > AVG(T.a), > AVG(T.b) > FROM T > GROUP BY > T.c > WHERE > T.creation_date=$SOME_DATE; > > There is an Index on T.c. But would it help to partition the table by T.c? > It should be mentioned, that T.c is actually a foreign key to a Table > containing a > tiny number of rows (15 rows representing different companies). > my postgres.conf is actually the default one, despite the fact that we > increased the value for work_mem=128MB > > Thanks in advance > Houman > > > > > > > -- > View this message in context: http://postgresql.1045698.n5.nabble.com/SELECT-AND-AGG-huge-tables-tp5728306.html > Sent from the PostgreSQL - performance mailing list archive at Nabble.com. > > > -- > Sent via pgsql-performance mailing list (pgsql-performance@xxxxxxxxxxxxxx) > To make changes to your subscription: > http://www.postgresql.org/mailpref/pgsql-performance -- Sent via pgsql-performance mailing list (pgsql-performance@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
