Inside the In-Memory Database Wars

There’s general agreement that the rise of in-memory databases has created a massive, multibillion-dollar opportunity for the channel. Beyond that, there’s next to no agreement about how that opportunity will actually manifest itself.

While in-memory database as a technology has been around for years, it wasn’t until SAP (SAP) launched the SAP HANA in-memory computing platform that interest in the technology took off. For the last few years SAP has been making the case for replacing traditional SQL databases with an SAP HANA database platform that runs applications orders of magnitude faster. For the most part, SAP customers have responded by augmenting their analytics and reporting applications with SAP HANA deployed in what is known as a “sidecar” to the data warehouse.

Naturally, SAP would prefer to have not only the data warehouse running on SAP HANA, but every transaction-processing application as well. Better still, SAP would like to see every one of those applications running in a cloud managed by SAP.

Oracle (ORCL), IBM (IBM) and Teradata (TDC) have responded differently to the rise of in-memory computing. Oracle has created an in-memory computing option that it sees being mainly used to augment analytics. IBM is focusing its efforts on a BLU Acceleration platform based on its DB2 database that it plans to update shortly, while at the same time reselling SAP HANA. Teradata is down a similar path with a Teradata Intelligent Memory platform.

In all three cases, Oracle, IBM and Teradata see Flash memory being used to augment in-memory computing with magnetic storage. The basic argument that not all data is “hot” enough to make it economical enough to run in Flash. And truth be told, there are many in SAP that would agree with them, In fact, SAP continues to sell the SAP ASE database it gained with the acquisition of Sybase.

But the biggest driver of in-memory computing adoption just might wind up being Microsoft (MSFT), which claims to have the only in-memory database that is optimized for both transaction processing and analytics applications. Eron Kelly, general manager of Data Platform at Microsoft, said that of all the in-memory databases, Microsoft SQL Server 2014 is the only one optimized to support both transaction processing and analytics. That’s because Microsoft SQL Server 2014 provides support for lock-free, row-versioning data structures and the compiling T-SQL and queries into native code in a way that allow organizations to take advantage of in-memory computing without application rewrites.

In fact, Kelly contended that in-memory databases eliminate the need to “shard” the relational database. Arguably, it was the need to shard the database that drove many organizations to embrace NoSQL databases. If there is no need to shard the database, Kelly noted, the need for a NoSQL database that must be managed alongside existing SQL databases is sharply reduced.

Meanwhile, there are a number of other in-memory database platforms being offered by much smaller companies. Exasol AG, for instance, provides an in-memory database specifically optimized for analytics applications. GridGain, meanwhile, has made its in-memory computing platform available as open source software.

It’s way too early to say to what degree in-memory computing will disrupt the IT world. The one thing that is for certain is the amount of space and energy that servers take up in the data center is about to be greatly reduced. It may take a little while longer for most IT organizations to get around to making that transition. But for solution providers that need to be on board now to help their customers make this transition over the coming year, the in-memory computing train is already starting to leave the station.