SAP announced an evolved big data strategy and vision on Wednesday, and it entails new partnerships, new applications, new integrations with BI and analytics software, and a (partly) new big data science team. The partnerships may not make SAP stand out, but the apps and big data team might become a differentiator.
The headliner amid these many steps is SAP's move to extend existing partnerships with Intel and Hortonworks by reselling their respective Hadoop software distributions. SAP put Intel at the head of the announcement because it's clearly counting on it becoming a special relationship.
"SAP plus Intel can galvanize the major investments that Intel is making [in Hadoop] and make it a preeminent release for the industry," said Irfan Khan, the newly appointed senior VP and general manager of SAP big data, in an interview with InformationWeek. "SAP is going to benefit from deep engineering integration between SAP Hana and Hadoop."
SAP has been working with Intel for years on the technology behind its Hana in-memory database platform, so the closer partnership makes sense. It's an attempt at differentiation in a market in which Oracle and HP are partnered with Cloudera, and Microsoft and Teradata are partnered with Hortonworks for Hadoop software. SAP clearly wasn't prepared to go the IBM route and come up with its own Hadoop distribution.
[ Want the lowdown on SAP's KXEN acquisition? Read SAP Buying KXEN For Predictive Analytics. ]
So why would SAP bother with a second partnership with Hortonworks? For one thing, Intel's Hadoop distribution, which has security and performance features designed to make the most of Intel Xeon processors, was introduced to most of the world early this year (after limited early release in China). Its market presence is unknown but undoubtedly limited at this early stage.
The Hortonworks partnership gives SAP a second Hadoop option differentiated as "100% open source," according to Kahn. Hortonworks developed vendor-specific distributions for both Microsoft and Teradata that are designed to be compatible with their management software (and operating system, in the case of Microsoft), but SAP will resell the standard Hortonworks Data Platform.
The bottom line is that we have yet to see what SAP and Intel jointly deliver to "deeply integrate" SAP Hana and Hadoop, and there is nothing SAP-specific in the Hortonworks distribution. So SAP customers will likely use Intel and Hortonworks management tools to run Hadoop and SAP management tools to run Hana and other SAP databases.
SAP took pains to point out that it will continue to work with and support other Hadoop distributions, including those from Cloudera and IBM. So it makes some sense not to mix management environments -- the tools used for deployment, monitoring and so on.
Where do Hadoop and Hana come together? Through querying, which SAP supports through Hive as well as through a SAP Hana Smart Data Access (SDA) layer introduced in May. Based on Sybase data-integration technologies, SDA lets users employ SQL to query data in the Hadoop Distributed File System. SDA will also support, in the near future, access data in NoSQL databases, Kahn said, so it's a single point of data access, not just a simple connector to Hadoop, he insists.
What about query and analysis tools? Here SAP has multiple options from the SAP BusinessObjects portfolio. A BusinessObjects 4.1 release introduced this week supports direct Hive access to data in Hadoop and Amazon Elastic MapReduce. With this approach, Hive SQL query results can be brought into a BusinessObjects Universe (data model) and analyzed alongside, or in combination with, more conventional data collections. Options include ad hoc query, Lumira data visualizations, and online and mobile reports and dashboards.
Bolstering analytics options, SAP announced Tuesday that it plans to acquire KXEN, a deal that would introduce automated predictive analytics that can be embedded into applications.