Partition

Question

I have a table that has telephone number and subscriber details.
Subscriber_table
subscriber_id
tel_no (e.g 416 888 224)
address
pin
This is a massive table having millions of records.The table is indexed on tel_no ,yet when a search is made on telephone number the sql takes long time to finish.
Wondering if I partition it say subscriber whose telephone area code start with 416 should go on partition 1 ,area code 905 partition 2...
does this partition make sense considering I already have index on tel_no.Will it make my search faster.
thanks in advance

Answer 1

Hmmm my $0.02 worth
I work with Telco billing systems, and deal with similar problems all day. 10 Million rows is not over the top, you should easily get get sub-second returns.
In my experience, partitioning will be of marginal value. As phone numbers are unique, (or near unique) therefore an index should be highly selective and return a single row. it's rare to do any sort of range scan or filtering on phone numbers, or perform any sort of group or aggregate function either.
Of course it depends upon how you are using your data, Also depending on how many users are hitting the same table and how. (is it an application? or a script?) Are you also searching on other fields which are not indexed.?
Is your query perfoming any function on the indexed field? I have seen queries using trim() refuse to use an index.
If you have an application like an ASP page executing dynamic sql, you would be much better off using a stored proc with bind variables.
I suggest you update your statistics, rebuild your indexes and start tuning from there.
I'd be very interested in seeing how you get on.
Good luck :)
Edited by: peetmoore on 23-Jun-2009 19:10
Edited by: peetmoore on 23-Jun-2009 19:13