When designing a data application, traditionally the bottleneck lies within the database. A traditional on-premises database does not scale very well. It is difficult to split a database between a cluster of machines so usually the database runs on the most powerful machine available which tend to be expensive.
To solve the problem some data processing is moved form its natural environment (i.e. DB) to the business tier which scale much better. The business tier is easy to replicate so it can help reduce the performance bottleneck of the database.
In Sql Azure things are different. Sql Azure is NOT running on simple Sql servers 2008 machines which happen to be on the cloud. Sql Azure has a different architecture. It runs on a grid of servers split into service and platform layers. The platform layer runs the sql engine and the service layer is responsible for the connection and provisioning.
"SQL Azure subscribers access the actual databases, which are stored on multiple machines in the data center, through the logical server. The SQL Azure Gateway service acts as a proxy, forwarding the Tabular Data Stream (TDS) requests to the logical server. It also acts as a security boundary providing login validation, enforcing your firewall and protecting the instances of SQL Server behind the gateway against denial-of-service attacks. The Gateway is composed of multiple computers, each of which accepts connections from clients, validates the connection information and then passes on the TDS to the appropriate physical server, based on the database name specified in the connection"
An important aspect about Sql Azure is the fact that this is a shared infrastructure. You do not own a specific server. This is why Sql Azure billing is calculated according to the amount of storage you use only and does not take the cpu consumption of the database into consideration.
Another thing to know about Sql Azure is the fact that your databases are replicated across 3 different servers. Large databases can be split horizontally using Sql Azure Sharding.
"Sharding is an application pattern for improving the scalability and throughput of large-scale data solutions. To “shard” an application is the process of breaking an application’s logical database into smaller chunks of data, and distributing the chunks of data across multiple physical databases to achieve application scalability"
Sql Azure in contrast to on-premises databases scales very well.
SQL Azure combined with database sharding techniques provides for virtually unlimited scalability of data for an application.
And a final word about billing: The only compute power customers are charged for in windows azure is on the running roles. Sql Azure CPU is free !!!
So the conclusion is that in cloud applications you might want to consider to move data processing back to the database, use Sql Azure Sharding and save money.