Hi,
I'm looking for people who have got experience at splitting a table with
heavy records into two relations.
In my case there exists one table with the mentioned heavy records. These
are processed by a statemachine. Thus a bunch of columns gets changed
several times in the livetime of one record. The text and blob colmns remain
unchanged. However due to Postgres# MVCC one changed record gets duplicated,
even if just one bit is changed. This makes us IO-performance problems.
I intend now to split the table in transport data (which stays constant) and
status data (which is changed several times).
I have then several options: Change all stored procedures to mathc the new
relations and keep the interface equal.
Or create a view with a bunch of RULES to show the procedures the 'original
table'.
What is best?
Who has done such bevore?
Will this help us in IO-terms?
thanks