Partioning based on string column

[yogesh chawla]

Hello All,
We currently have an application where almost all searches are done against one table and every search requires two characters of a last name be entered. As this table has grown to around 10 million records the searches have been bogging down.

I was thinking a solution to this problem would be replication. However the database gets considerably more writes than reads and it seemed like replication worked better for high read, low write databases. Regardless I have it set up and working.

I would like to partition the table based on this last name field. From what I see so far, partitioning based on a string is not available. It looks like I could do a key partition, but I don't see how I would get a performance gain there since queries might still have to work over multiple partitions.

Can I partition by Last name using a range like a-c, d-f, etc? If I am required to have a numeric field for the partition, should I add a column called where last names that start with A get assigned the number 1, b->2, c->3 and do that?

Any suggestions that anyone has would be helpful. I am not a DBA and mostly work writing web apps. Unfortunately no DBAs work on the team so any assistance you can provide would be great.

Thanks,
Yogesh