This project is read-only.

WindowsAzureTableSink - why are payload fields duplicated? Partition key resolution?

Topics: General, Using
May 19, 2015 at 11:32 PM
Looking for some explanation, costs & benefits of why event properties are persisted the way they are in Azure table. In particular, the event payload is persisted twice -- once as a JSON blob, also as separate properties. Why the duplication? Could this be made configurable?

Why both ProviderId and ProviderName? Can't one typically be determined from the other?

Trying to figure out whether I really shouldn't worry about these and use the service out-of-the-box or whether it would be better to customise it for our needs. It looks like we're going to need to write some querying/viewing tools anyway, so why not tweak the schema ourselves?

How was the one-minute resolution of the partition key determined? Is this a "generally good" value for everyone regardless of whether the frequency of events is 1000s per second or just a few per hour? What are some guidelines that we could use to determine optimum partition resolution for our usage?