Thread: Paritioning vs. caching
Hello, I am evaluating a materialized view implemented as partitioned table. At the moment the table is partitioned yearly and contains 5 numeric/timestamp columns. One of the columns is ID (but it's not what the table is partitioned on). Partition for one year occupies about 1200 MB. Each of the columns is indexed, with each index weighing about 160 MB. I am trying to avoid RAM/disk thrashing. Now I have the following questions: 1. When I query the table by ID, it performs index scan on each partition. The result is only found in one partition, but I understand why it needs to look in all of them. How much disk reading does it involve? Is only the "head" of indexes for partitions that do not include the row scanned, or are always whole indexes read? I would like to know the general rule for index scans. 2. Is it possible to tell which PG objects are read from disk (because they were not found in RAM)? Thank you. -- Konrad Garus
If the partitioned column in your where clause does not use hardcoded values ...e.g datecolumn between 'year1' and 'year2' ..the query planner will check all partitions ..this is a known issue with the optimizer On Mon, Mar 8, 2010 at 10:28 AM, Konrad Garus <konrad.garus@gmail.com> wrote: > Hello, > > I am evaluating a materialized view implemented as partitioned table. > At the moment the table is partitioned yearly and contains 5 > numeric/timestamp columns. One of the columns is ID (but it's not what > the table is partitioned on). > > Partition for one year occupies about 1200 MB. Each of the columns is > indexed, with each index weighing about 160 MB. I am trying to avoid > RAM/disk thrashing. Now I have the following questions: > > 1. When I query the table by ID, it performs index scan on each > partition. The result is only found in one partition, but I understand > why it needs to look in all of them. How much disk reading does it > involve? Is only the "head" of indexes for partitions that do not > include the row scanned, or are always whole indexes read? I would > like to know the general rule for index scans. > > 2. Is it possible to tell which PG objects are read from disk (because > they were not found in RAM)? > > Thank you. > > -- > Konrad Garus > > -- > Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) > To make changes to your subscription: > http://www.postgresql.org/mailpref/pgsql-performance >
> 1. When I query the table by ID, it performs index scan on each > partition. The result is only found in one partition, but I understand > why it needs to look in all of them. How much disk reading does it > involve? Is only the "head" of indexes for partitions that do not > include the row scanned, or are always whole indexes read? I would > like to know the general rule for index scans. If you're not including the partition criterion in most of your queries, you're probably partitioning on the wrong value. --Josh Berkus