I talked to a couple of developers about their experience of working with Windows Azure.
http://vistasquad.co.uk/blogs/nondestructive/default.aspx
Ray Booysen has developed two Windows Azure applications. The first is a proof-of-concept/demonstration project to visually display how articles are linked in Wikipedia (www.dotnetsolutions.co.uk/evidence/wikiexplorer/, shown in Figure 16-21), and the second is an application for providing financial risk assessment.
Ray had the following tips:
When working with Azure storage, stop thinking normalization—this is not a normal database.
Carefully consider your partition and row keys because cross-partition searching is very slow (in his Wikipedia Explorer example, the first letter of the article was used as the partition key).
Consider duplicating data for performance reasons.
Ray felt that Azure could be improved by the following:
Better logging and analytics
Performance counters and some kind of query analyzer for Azure table storage
SharpCloud is currently developing a Silverlight/Azure project risk assessment application (discussed in Chapter 15 as well). The application carries out many complex calculations (using Azure worker roles) and suggests the optimal project selection.
Rusty and Andy had the following to say:
The move to Azure was very easy (previously they were using Amazon's cloud services).
Azure table storage can take a bit of getting used to but is quite intuitive once you start using it.
Azure is about scalability, not performance—this is an important difference, and you will have to do additional performance work.
18.118.0.42