Hemant Pandey wrote:
Operation A:            6 Fetch + 1 Update
Operation B:            9 Fetch
Operation C:            5 Fetch + 3 Update  ( Tables has 140 records)

I have run these operations while Auto Vacumm is running and observed the
time taken in thse operations. I found that Operation C is taking highest
time and A is the lowest.
So i inferrred that, UPDATE takes more time.

Now i run these operations again, without running Auto Vacuum. I observed
that, time taken for operation A & B is almost same but time for Operation
C is increasing.

I am not able to analyze, why only for operation C, time is increasing??
Does auto vacuum affects more on UPDATE.

Depends on what is happening. Without vacuuming (automatic or manual) a table will tend to have "holes" since an update with MVCC is basically a delete and an insert.

Since you say the table has only 140 records, almost any operation will tend to scan rather than use an index. However, if you issue lots of updates you will end up with many "holes" which have to be scanned past. PG won't know they are there because its statistics will be out of date unless you have analysed that table recently. So - everything will start to get slower.

So - for a small, rapidly updated table make sure you vacuum a lot (perhaps as often as once a minute). Or, run autovacuum and let it cope.

  Richard Huxton
  Archonet Ltd

---------------------------(end of broadcast)---------------------------
TIP 3: Have you checked our extensive FAQ?


Reply via email to