Now that the prices are clear, the big question is how we can predict the costs. We do not want to be surprised with a high bill at the end of the month, so it is important for us to be able to estimate the costs per billing period. As described in the previous sections, a few costs are hard to estimate. These are bandwidth, transactions, and the amount of data stored. Predicting the costs of compute power is fairly simple, at least, if your deployment is a flat deployment with no auto scaling implemented (see Chapter 8, Windows Azure Patterns, for auto scaling using the Enterprise Library).
Bandwidth is only charged when data is accessed from outside the datacenter. Hosted services deployed in the same datacenter, using storage, will not cause any bandwidth charges.
Every REST request to any of the storage concepts (not SQL Database) will cause a storage transaction (which is billable).
Everything that is stored in your storage accounts will be added up to measure storage costs...