[Statistics] Reduce DB size by removing unused data - edge server IP

[skofman1]

Impact on database

A database size reduction could be achieved totalling 46,284.26 MB (or 45.2 GB)!

Removing these indices will also improve performance of insert/delete operations against the [dbo].[Fact_Download] and [dbo].[Fact_Dist_Download] tables.

In addition, we generate less churn on the data by avoiding clean-up operations.

The following database objects would become obsolete:

• Table [dbo].[Fact_IpAddress] (size = 3.22 MB)
• Table column [dbo].[Fact_Download].[Fact_EdgeServer_IpAddress_Id] (size = 2,585.75 MB)
• Table column [dbo].[Fact_Dist_Download].[Fact_EdgeServer_IpAddress_Id] (size = 23.63 MB)
• Stored Procedure [dbo].[CleanupFactIpAddress]
• Stored Procedure [dbo].[EnsureIpAddressFactsExist]
• SQL Type [dbo].[IpAddressFactTableType]
• Non-clustered index [dbo].[Fact_Download].[Fact_Download_NCI_EdgeServer_IpAddress] (size = 43,278.21 MB)
• Non-clustered index [dbo].[Fact_Dist_Download].[Fact_Dist_Download_NCI_EdgeServer_IpAddress] (size = 392.80 MB)
• Non-clustered index [dbo].[Fact_IpAddress].[IX_Fact_IpAddress_UniqueIndex] (size = 672 KB)

Impact on jobs

The Stats.ImportAzureCdnStatistics job would no longer need to deal with edgeserver IP-address data.

We could:

• avoid the need to check whether an EdgeServer IP-address already exists, and if not, create it during import;
• avoid the need to retain already imported EdgeServer IP-addresses into in-memory cache;
• reduce complexity of import by no more linking EdgeServer IP-address identifiers to the download facts.

[skofman1]

We decided to go in a different direction to reduce DB size. Will reopen if needed.