Redmond Review

Data Platform Needs Foundation, Not Just Planks

As Microsoft continues to offer newer, more innovative data platforms in the cloud, the SQL Server flagship is languishing. Here's what Microsoft needs to do to ensure it moves forward.

• By Andrew Brust
• 04/07/2015

Microsoft has been bulking up its data platform fiercely of late, especially its cloud offerings over the last year. Take Azure DocumentDB, Microsoft's NoSQL database, which enters general availability on April 8. There's also Azure Stream Analytics, something that resonates especially well in this age of the Internet of Everything (IoT). Azure Machine Learning is getting traction, and respect across the industry. Azure Data Factory represents a new take on ETL. There have been improvements to Azure SQL Database, Microsoft's platform-as-a-service relational offering, too. And Power BI is being revamped in a serious way, even being extended to the iPhone and iPad.

But what of Redmond's database flagship, SQL Server?  The truth is that it's been languishing.

Think about it. SQL Server 2008 brought real advances to the core relational engine, including spatial data types, filestream access, full text search and more. But SQL Server 2008 R2 was essentially an update to the BI stack, 2012 added little more than columnstore indexes and high availability features and SQL Server 2014 was really just the ship vehicle for the new "Hekaton" in-memory OLTP engine. All of these are important features, but they don't enhance the mainstream OLTP developer experience.

So here Microsoft is, in 2015, with a growing data stack, but with a flagship relational engine that hasn't been substantially updated in seven years and a reporting engine that hasn't seen much improvement for five. There's no JSON or key-value data support, no support for INTERVAL data types, clustering is relatively difficult and the reporting platform has little support for modern HTML and virtually none for mobile devices.

Worse yet, successive years of pushing most of the new features that have been developed exclusively into the Enterprise edition of the product have left the Standard edition especially neglected. That's not just inconvenient for the budget-conscious; it's actually a huge problem for the platform overall. In a day when so many database products are being open-sourced, Microsoft's insistence on shoving all of its RDBMS innovation into a premium version of the product runs counter to industry expectations.

Ignore ISVs at Your Own Peril
Independent Software Vendors (ISVs) do not want to require customers to buy into Microsoft's premium product just to qualify for entry-level editions of their own offerings. And while ISVs may not be Microsoft's prime target, they are nonetheless a significant constituency, and the ones best able to influence adoption of important new features, like in-memory OLTP. If new features go only into the "up-level SKUs," those features may never garner critical levels of adoption. That, in turn, may cause the features to be deprecated in the future, eroding value delivered to the enterprises who do license the higher-priced editions.

Fellow SQL Server MVP Greg Low told me recently, "Whenever I get into a discussion with the [SQL Server] product group about pricing, the response is that 'it's still cheaper than Oracle.' That seems to be the only test."  Perhaps it's no surprise, then, that Microsoft has seemingly been blindsided by more robust competition from data platforms like PostgreSQL and reporting platforms like Jaspersoft (now owned by TIBCO). Not surprisingly, both of these are open source, and popular among ISVs.

And even back on the Oracle front, hubris can be a dangerous thing. I spoke to a consultant recently who told me his firm had to work carefully to prevent a client from leaving the SQL Server platform because, for its use case, Oracle was actually cheaper. In the end, the customer stuck with Microsoft, but only because of its investment in the T-SQL skill set. That self-perpetuating phenomenon can only work for so long.

TCO, RIP?
Microsoft likes to assume its data platform has the lowest total cost of ownership and that it's the most developer-friendly platform. Something else that Low shared with me recently was his opinion that "a big cost of ownership is the cost of coding and maintenance of code. There are many ways that PostrgreSQL allows for simpler, cleaner code."  Take a look at Postgres projects like PostGIS and Npgsql, or its hstore module, and you'll begin to appreciate Low's point.

Microsoft likes to invest in advanced analytics and cloud. That's a smart investment but, ultimately, it's only as sound as the stack's foundation. Back in the 1970s, New York City neglected its infrastructure under a program it euphemistically called "deferred maintenance," to disastrous results. Such a program is one Microsoft cannot afford to implement any longer. It's time to make SQL Server great again. Modernize its data types, make it easier to scale out, make the Developer edition an easily available, free download, make the Standard edition feature complete, revamp the reporting platform and add long-requested features like support for JSON, UTF8, table inheritance, arrays and enums.

Microsoft needs to do this soon, before deferred maintenance turns into deferred license renewals and accelerated re-platforming by longtime customers.