mca blog

spent teh afternoon in sessions on Azure Storage. interesting stuff.

Azure Storage provides three main classes of storage:

• BLOB
• Table
• Queue

BLOB

BLOB storage is all about files - big ones, too. you can add/update/delete BLOBS. even large ones (currently 50MB) are supported via a continuation upload pattern to allow restarting failed long uploads. you can also attach custom metadata to your BLOBs.

Table

Table storage is a simple data model for storing documents that are searchable. you create a partion and a key, then add any metadata you wish. not as powerful as SQL Data Services (SDS), but easier to deal with.

Queue

Queues allows you to create and destroy your own queues and push and pop messages there. they have a nice model for handling simple failures - if you pop a message off and fail to tell the service you completed it successfully, the service will 're-surface' that message for processing again. looks straight-forward and easy to work with.

the examples shown use ADO.NET Data Services classes and support HTTP/REST. but the URI patterns are decidedly *not* REST-ful since some operation information is included in the URI as arguments. pretty RPC right now. but at least you can get work done w/o SOAP.

authentication is done through a token service and custom header. no concrete plans to support HTTP Auth (Basic/Digest). HTTPS is available as an option, but not as a 'forced requirement.' finally, you can mark your containers as public or private. private requires owner auth. public requires no auth at all.