Please let me know if you have implemented large scale data warehouse on SQL Azure. We are looking at a size of 15-20 TB data. A significant part of our data originates from files (semi-structured, unstructured) for which we are planning to use HDInsight to mine/process the data and store the extracted + aggregated (for example minute-level data to be aggregated to day level) data on the DWH.

Has SQL Azure been used successfully as a DWH for high volumes? Please share your thoughts.

just curious that azure db will have a 150gb limit per db, with your data size needs you will need a large number of databases, 

have you explored option of on premises db?

SQL Azure is not focussed on a Datawarehouse environment, particularly high volumes.  For a start, the maximum database size is 150GB, so you'd need a lot of them, and then for reporting you'd need to run fan-out queries then collate and aggregate the data in some way.  At present no reporting tools offer this functionality, so you'd probably have to write a custom datasource to do it, or pull the data into a central area first (thus negating the use of SQL Azure in the first place). 

Even a persistent VM would probably not support your requirements, as at present you are limited to 16TB of storage - that is disk space, so you need to take everything into account such as transaction logs, tempdb etc.

Thanks for the reply. I was thinking in terms of federations - however, I understand that managing/accessing the DWH may be a challenge in that case. Yes, on premise DB, PDW, HIVE (on HDInsight) are the other solutions we are thinking of.