views:

31

answers:

2

While running an application load test, I am observing some weird behavior. Lock requests/sec counter is increasing in a linear fashion throughout the whole test (duration 12 hours, load levels off to a constant level within first 10 minutes). The value reached 6 million at 12 hours. There was no apparent impact to the response time of the application. There was also no impact to lock wait time (200ms average). Database CPU slowly increased from 20% to about 30% at 12 hours.

What could be causing such behaviour?

+1  A: 

You are going to need to start profiling the database to see what items are requesting locks, and from there you will be able to see what is happening with the lock requests. Is the amount of data growing in your application? If so, that could be a source of the increased lock numbers.

Mitchel Sellers
The amount of data is unlikely to be the reason here. The amount of data is more or less constant. The data throughput is not growing either.
Evguenia
A: 

We were able to get rid of the escalating request locks/sec issue by setting the default transaction isolation level to "READ_COMMITTED_SNAPSHOT". However, there is still no explanation why this was happening in the first place. Any ideas are welcome.

Evguenia