Archive

Archive for the ‘Database Engine’ Category

Microsoft Changes Feature Support Between Editions in SQL Server 2016 Service Pack 1

In an unprecedented announcement at CONNECT(); //2016 in New York City today, Microsoft has announced that a number of Enterprise Edition only features will now be available in all editions of SQL Server. That’s right, all editions of SQL Server!

The delivery vehicle for these changes will be SQL Server 2016 Service Pack 1, which is available immediately!

This is a first for Microsoft, who has to date only ever moved features from Enterprise Edition to Standard Edition between different versions of SQL Server.

Specifically the following features will be available in all SQL Server editions:

  • Always Encrypted (AE)
  • Change Data Capture (CDC)
  • Columnstore indexes
  • Data Compression
  • Database Snapshots
  • Dynamic Data Masking (DDM)
  • Fine Grained Auditing
  • In-Memory OLTP (IMOLTP)
  • Multiple FILESTREAM Containers
  • Partitioning
  • Polybase
  • Row Level Security (RLS)

That is quite an impressive list of functionality which will now be available in all editions. And by all editions I mean all editions. So from LocalDB all the way through to Enterprise Edition a developer will have the same, consistent programmability surface area. Because of the differences in how some of these editions work there will be some limitations. For example Change Data Capture will not work on LocalDB and Express editions due to a various dependencies. In-Memory OLTP will not be available with LocalDB due to technical limitations. With Columnstore Indexes Microsoft will be limiting the degree of parallelism for queries used by the Database Engine. With Standard Edition queries will be limited to 2 cores per query. Web Edition and Express Edition will be limited to 1 core per query. Polybase will not be available with LocalDB. Furthermore Polybase will only be able to be installed as a Compute Node on the lower editions. For scale-out solutions the Head Node will need to be installed on an Enterprise Edition instance. This could potentially dramatically reduce the licensing cost for enterprise customers.

In most cases these “limitations” will not dramatically affect most customers as I expect them to take advantage of this new functionality using Standard Edition.

I fully expect the feature set to potentially evolve as Microsoft embraces this “brave new world” and gets feedback on these changes from the industry.

You might recall the SQL Server Standard Edition is currently limited to 128GB of memory, with Express Edition supporting 1GB of memory. So how will features such as In Memory OLTP tables and Columnstore Indexes, which are memory hungry technologies, play within these memory limits? Instead of allocating a subset of these hard memory limits to these features, which would potentially lead Buffer Pool pressure and lack of out-of-memory problems, Microsoft have decided that these “hero” features will be able to use up to 25% of the hard memory limit for both In-Memory OLTP features and Columnstore Indexes. That means that on an Express Edition instance both In-Memory OLTP tables and Columnstore Indexes could each consume 250MB of memory above the Buffer Pool allocation. On a Standard Edition instance both technologies would be able to take advantage of 32GB of memory each. I expect the primary reason for this approach has been to reduce the potential impact on existing database solutions.

So why the change in “policy”?

I see the primary driver for Microsoft being to reduce the friction that developers have when potentially having to develop multiple versions of what is fundamentally the same solution. To have a consistent programmability surface area, or feature parity for Developers if your like, between all editions of SQL Server. As Enterprise Edition is not deployed everywhere, developers tend to code for the “lowest common denominator” with respect to features.

On the flip side of the same “driver” coin it will also help drive adoption of the “hero features” by both developers and in particular Independent Software Vendors (ISVs), who have for years complained about the lack of certain features in lower editions of SQL Server. I believe Microsoft’s promise/vision of being able to deploy a database solution on LocalDB with all of these programmability features and then being able to run the same solution on all higher editions represents a “game changer” in the Developer space.

This can be evidenced from the CONNECT(); //2016 fact sheet:

SQL Server 2016 SP1 makes leading innovation available to any developer.
Microsoft is making it easier for developers to benefit from the industry-leading innovations in SQL Server for more of their applications. With SQL Server 2016 SP1 is making key innovations more accessible to customers across editions. Developers and partners can now build to a single application programming surface to create or upgrade new intelligent applications and use the edition which scales to the application’s needs. SQL Server Enterprise continues to offer the highest levels of scale, performance and availability for enterprise workloads.

So where does that leave Enterprise Edition?

As the fact sheet hints at, Microsoft still sees SQL Server Enterprise Edition as the edition that will provide unlimited scale, unlimited availability and unlimited virtualization. So being able to scale out your database solution using Readable Secondaries in Availability Groups, for example, will remain in the realm of Enterprise Edition. Likewise there are no changes to Basic Availability Groups in Standard Edition.

In that context Online Index Rebuilds will remain in Enterprise Edition. For now…

Is Online Index Rebuilds a “programmability” feature? Or an “availability” feature? I am sure there have been a number of interesting and heated discussions in Building 16 for the last 6 months on what constituted a “programmability” feature in preparation for the release and announcement of SQL Server 2016 Service Pack 1.

And before I forget, and before you ask, these changes will not be backported to SQL Server 2014.

Both tactically and strategically this is an interesting “play” by Microsoft.

In the Enterprise space Microsoft has proven to be very competitive against Oracle, with a substantially “cheaper price”, an extremely compelling / comparative feature set and a number of Migration Programs where you can get SQL Server Enterprise licenses for free. So I don’t think too much will change in this space.

Tactically this should encourage existing Microsoft customers who only have Standard Edition to upgrade to SQL Server 2016 Service Pack 1. In this use case being able to take advantage of Data Compression transparently will potentially have huge benefits, including substantially improved performance, and substantially reduced backup times and sizes. Likewise, Dynamic Data Masking (another transparent feature) might prove to be very compelling to customers who do not want production database to be so easily exposed in DEV and UAT environment.

More importantly SQL Server 2016 Service Pack 1 will allow SQL Server’s entire customer base to take advantage of the “hero features” that Microsoft have invested in for the last 5+ years, such as In-Memory OLTP and Columnstore Indexes

As I have said this is completely unprecedented in the history of SQL Server releases. With Microsoft’s new philosophy of ensuring a “consistent programming surface area”, or feature parity if you like, developers will be able to build the best possible solutions that they can on the SQL Server platform. That is a huge change of mindset by Microsoft, and highlights their maturity and confidence in their Data Platform.

For me the more interesting “play” is how this is going to strategically impact the Open Source database ecosystem and community in particular. A quick search on the web will reveal how popular the MySQL and in particular PostgreSQL platforms are. We have a new generation of developers in the IT industry that are developing on these Open Source platforms. Yes, as has been well documented, Microsoft will be releasing SQL Server on Linux in 2017. But that is not enough to encourage Open Source Developers to adopt / embrace the SQL Server platform. Microsoft also needs to have a compelling platform to use in its own right, in terms of functionality and features. For the “right price”.

Think back to the mid-nineties in the relational database ecosystem, with SQL Server, Oracle, Access, Clipper, DBase adoption. There were no Open Source database platforms like there is today. Both corporations and Developers / IT Professionals embraced Microsoft SQL Server over other platforms for a number of reasons, with functionality being a primary consideration. Now think to today and the kind of database-centric solutions that can now be easily and affordably built in the cloud. Times have changed. And it’s important to recognise that the Open Source movement is not in competition to Microsoft. Microsoft has embraced Open Source in their cloud like no one could have anticipated or even imagined.

With the release of SQL Server 2016 Service Pack 1’s “consistent programming surface area” / feature parity across all editions, in conjunction with a phenomenal feature set and an extremely mature relational database engine, Microsoft has effectively invited the Open Source community to develop and deploy database-centric solutions on their Enterprise-grade database platform.

It’s certainly interesting to see how this will play out in the next couple of years.

Advertisements