PaaS (Platform-as-a-Service) is best chosen when majority of your business needs relies upon web, not having necessity of storing data on-premise. A popular choice within cloud computing services.Combing with Infrastructure as a Services that provides virtual machines and/or storage from a provider on demand, the kind of PaaS is handy to use service when it is needed.
There are certain best practices to be followed on DWH and BI perspective, being data is primary objective for business growth. When it comes to Big Data Apache Hadoop comes first on the list and most commonly used associated with MapReduce, Hive, HBase and Pig (list goes on…). There are specific use case scenario associated with these technologies within BI, DW and DA (Data Analytics), but traditional RDBMS are better placed within Core-Relational (OLTP), BI & DW arena. So the tools are very specific to usage such as reporting tools, analytics tools and consumption tools. So in my opinion Hadoop will only complement existing BI.DW data platform in your organisation, as it is harder to implement an on-premise Hadoop environment that will impact TCO and understand how best you can make use of these new technology & tools, maxium business value with less cost is primer.
In any case Microsoft’s Azure platform has put up strong foot hold within data platform and specific to cloud computing arena. In the recent times few new developments extended the new capabilities with Hadoop and Hortonworks data platform combining Microsoft technology capabilities.In order to understand what is offered within Azure, you need to understand loading patterns & strategies that can help to design affordable solution with preferable methos on fast load methods (data ingest methods), by using SQL Server capabilities as well (suggestion to refer What is Azure SQL Data Warehouse?). To understand much deeper into architecture, how Polybase is useful with SQL Data warehouse loads and other best practices refer to this Azure SQL Data Warehouse loading patterns and strategies guide from SQLCAT team.