SQL Data Warehouse
The promise of "An elastic data warehouse-as-a-service with enterprise-class features" really got my interest when it came up on a LinkedIn post recently. How could it not, really!?!?!
What Does SQL Data Warehouse Offer?
Let's take each of the seven bullet points in turn...
1) "Petabyte Scale with massive parallel processing"
Now we're not going to diss the notion of "Petabyte Scale" - that's the new norm in many sectors. Terabytes are so last decade dahling. However, we would be interested to see just how many in Microsoft land will be thinking "phew, we're saved, Microsoft has finally got us a solution to our petabyte sized problem". Unlikely to be a stampede, but you never know.
Now, the notion of scalability, rather than scale, is likely to be of far more interest, which is probably where the "with massive parallel processing" comes in. The shorthand for this capability is MPP, which is not a concept that many of those in Microsoft land are likely to be overly familiar with, let alone intimate with the benefits MPP brings. Then there are the MPP challenges that don't exist in normal non-MPP SQL Server database land.
2) "Independently scale compute and storage in minutes"
Now you're talking. Depending on the requirements, this should allow a CPU heavy stack, a balanced stack or a storage heavy stack. All good stuff, especially if it can be achieved quickly and easily.
What's not said is perhaps more interesting. It may be able to scale the individual nodes by adding more CPU and/or storage, but the Fat Node or Scale Up approach is diametrically opposite the 'scale out' notion that underpins the MPP architecture. The whole point of MPP is that we don't scale up - we scale the cluster out by adding more nodes, not by making the existing nodes bigger/faster.
3) "Query across relational and non-relational data with T-SQL"
Not sure how this will resonate in Microsoft land. Once again, you never know.
4) "Full enterprise-class SQL Server experience"
OK, it's still SQL Server. No surprises there then.
5) "Self-managed for zero maintenance"
That's the OS and DBMS covered, by the sound of it. But does that include a database administrator (DBA)? Highly unlikely.
Try running a "petabyte scale" (or even terabyte scale) MPP data warehouse without a skilled DBA and let me know how you get on. There's a reason those folks are well paid. And you'll need more than one.
6) "Compatible with a wide ecosystem of SQL Server partners"
Few of whom will have have any "petabyte scale" or MPP experience. The high-end world of clustered MPP data warehousing is ever so slightly different to your run-of-the-mill SQL Server system, trust me on that.
7) "Integrated with Power BI, Machine Learning, HD Insight and Data Factory"
So it works with other Microsoft products, as expected.
The marketing bullet points above may not get us all excited, but we'll still give Microsoft's SQL Data Warehouse a thorough inspection in due course. How could we not, we've been waiting long enough!<History Lesson>Microsoft's first foray into the world of MPP data warehousing stretches back to the acquisition of Datallegro back in 2008. Why Microsoft thought a Linux/Ingres based stack was what they needed is anyone's guess.
At the same time, yours truly was working with Netezza co-founder Foster Hinshaw at his next venture, Dataupia. Here's the press release for a 25 billion row MPP-powered SQL Server system we built at Dataupia in 2009: Dataupia ITIS SQL Server</History Lesson>