Chaps & Chappettes,

We have done some work on Jira's database performance, and have come up
with some performance enhancements. (Very significant if you are using SQL
Server 2000+)

First of all, if you are using SQL Server 2000, there is a huge perf gain
to be had by setting up string database columns as unicode (varchar is
ascii).
If your jdbc driver is sending strings as unicode (you can force some to
send ascii), and the columns are ascii, then SQL server wont use any
indexes that exist for those ascii columns.
This means most of the indexing wont be used.

We have added a new database "dialect" mssql2000 that sets up string
columns as nvarchar (ie unicode) for all strings. The existing mssql one
uses varchar (ascii)
NB (text stays as text, rather than ntext - as we dont index those fields,
so we dont care).
Here is the mssql2000 dialect:
(See attached file: fieldtype-mssql2000.xml)

We have also added some indexes and modified one existing one.
Here is the modified entitymodel.xml with the new/modified indexes (if you
are installing a new Jira instance)
Here is also a "delta" script that you can execute on your existing Jira
3.2 database:
(See attached file: optimised-jira-3.2-entitymodel.xml)
(See attached file: optimised-jira-3.2-index-delta.sql)

Many thanks to Michael Thorpe, who did much of the work.