Alright, as you have stats showing that there is latency from the disk system I would try and prove it.

You're right when you initially said that figures showing latency are hard to completely confirm that the disk system is the bottleneck. You don't have a baseline so it is impossible to determine that that is the problem (the latency from the disk could have always been at those figures and something else is causing the problem).

I would try and prove it by running a query that would retrieve a large number of records from the database.

Initially (if the table has not been queried recently), SQL Server would have to retrieve the data from disk and copy into memory. Once that has been done, subsequent executions of that query would return quickly. This should prove that as the initial execution was longer, the disk system is the bottleneck.

Should be quite a good way of demonstrating the problem by showing the run times of the first query against subsequent queries.

If your system has many different queries executed against it, this would be a good demonstration of where the problem is. However if you have multiple executions of the same query, the problem lies elsewhere.

I would caution against running this on your live system as it would affect performance.

What do you think?

Andrew