Why am I Getting so Many Deadlocks From Windows Azure SQL Database?

The short answer is that you’re probably not using transactions.

I came across this issue a few time and decided to create the ReliableModel. When I modify the state of my database, it automatically wraps my code with a TransactionScope. By default, TransactionScopes are set to timeout after 30 minutes. This configuration is set in the machine.config and it cannot be modified on Windows Azure Roles. Therefor, if your statements need more than 30 minutes to execute, I strongly recommend using Stored Procedures. Be sure to use Transactions in your Stored Procedures.

Don’t Forget About Index Maintenance on Windows Azure SQL Database

Granted that Azure SQL Database takes care of a lot of maintenance tasks usually done by DBAs, you should definitely keep an eye on those Indexes. Indexes get fragmented and eat up lots of precious space. Furthermore, fragmented indexes make for slower queries.

I’m currently using the following technique to keep my Indexes healthy so that I may offer a constant end user experience. Furthermore, giving my Indexes regular doses of love prevents my database from growing too fast, which results in drastic cost savings. Throughout my test period, I noticed that fragmented indexes for large amounts of records would eat up close to 4gb of extra data. Once the Indexes are rebuilt, that extra space is recuperated and I am able to push back the dreaded moment where I have to pay for more SQL Database storage.