Sign in   |   Register

In-Memory Analytics Buyer's Guide: Oracle Big Data/Exalytics Appliances vs. SAP HANA

  |     |   Bookmark
         
Posted February 8, 2012 By Drew Robb     Feedback

Big Data has become a super-heated area of the business intelligence landscape, with companies hoping to extract actionable insights from their growing stores of data. Oracle and SAP are among the vendors introducing solutions designed to help them do so

Big Data has become a super-heated area of the business intelligence (BI) landscape, with companies hoping to extract actionable insights from their growing stores of data. Business intelligence vendors are introducing solutions designed to help them do so. SAP introduced HANA (High-Performance Analytic Appliance) in December of 2010, followed by Oracle’s releases of its Big Data and Exalytics Appliances in October of 2011. Those aren’t the only options, however. IBM, SAS and others also offer Big Data solutions.

This article is the first of a two-part series on these kinds of solutions. The first story focuses on SAP and Oracle products, comparing and contrasting their offerings. Part Two will consider some of the alternative approaches to analytics in a Big Data world.

System Details

Oracle: The Oracle Exalytics Business Intelligence Machine is an in-memory hardware and software system engineered to run analytics faster. The idea is to broaden user adoption of business intelligence through the introduction of interactive visualization capabilities that can be utilized by users who aren’t schooled in business intelligence techniques.

Oracle opted for an integrated approach to hardware, software and middleware to provide users with a box that does one thing really well. Its Exalytics box comes with the Oracle Business Intelligence Foundation, a version of the Oracle TimesTen In-Memory Database, and an Oracle SunFire server which offers 1 terabyte (TB) of RAM and the Intel Xeon E7-4800 processor (40 cores). That’s enough RAM for abundant in-memory business intelligence. According to Oracle, InfiniBand database connectivity eliminates a lot of query latency.

“There are no limits to in-memory analytics,” said Paul Rodwick, Oracle’s vice president of Business Intelligence. ”We have the Oracle BI Foundation pre-engineered into Sun hardware.”

The Oracle Big Data Appliance, on the other hand, is for processing unstructured data using Oracle Database 11g and open source Apache Hadoop technology. It can be hooked up with other Oracle boxes to create hybrid systems that crunch both structured and unstructured data.  

“Exalytics can also sit on top of the Big Data Appliance to analyze unstructured data,” Rodwick said.

SAP: Instead of an engineered system from one vendor, the SAP HANA platform is a combination of in-memory software and partner hardware. This in-memory computing engine enables an integrated database and calculation layer that allows the processing of massive quantities of real-time data in main memory.

The SAP in-memory computing engine incorporates a high-performance calculation engine that embeds procedural language support directly into the database kernel. This approach is designed to eliminate the need to read data from the database, process it and then write data back to the database.

“SAP built its in-memory platform from the ground up and it’s rebuilding its application platform around this technology,” said Jacob Klein, senior vice president of Product Management at SAP. ‘’In-memory will be as transformative to enterprise software as the shift to client/server technology was 25 years ago.”

With the dramatically improving price performance of main memory and the rise of massively parallel multi-core processors, Klein said systems can now support use cases that were impossible before. Examples include the execution of core ERP transactions up to 1,000 times faster.

He agrees with Oracle in one sense – to be able to take advantage of this potential, you have to change and align system architecture. But while Oracle goes for a software/hardware combo, SAP is more focused upon refactoring the application stack itself. 

SAP NetWeaver Business Warehouse (BW) acts as a data warehouse for HANA. The company is rolling out applications that can leverage HANA, including SAP Business Objects Strategic Workforce Planning, SAP Dynamic Invoice Discounting, SAP Trade Promotion Management, sales and operations planning, smart meter analytics, dynamic cash management, profitability analysis accelerator and profitability analysis accelerator for SAP ERP.

Slanging Match

Here's the word from Oracle and SAP: “The other guy’s solution isn’t nearly as good as ours.”

Oracle has invested a lot of blogging hours into differentiating its boxes from HANA. To summarize, Oracle’s position is that since HANA is not just for analytics, it doesn’t perform as well as Exalytics. Also, HANA only works with SAP tools/data whereas Exalytics works with any mainstream database, and HANA doesn’t come with business intelligence tools like dashboards, reports, scorecards and ad-hoc query and analysis.

Amit Sinha, SAP’s vice president for In-Memory Computing and HANA Marketing, counters that: HANA is an integrated appliance with key features at the chip level for in-memory processing, thanks to years of work with Intel. SAP has worked with IBM, HP, Fujitsu, Dell, Cisco and Hitachi to certify their hardware configurations on HANA. You can use business intelligence and reporting clients such as SAP Business Objects BI 4.0, Microsoft Office and other tools on top of HANA.

“While we favor SAP Business Objects BI 4.0, we remain open to how customers want to utilize SAP HANA through a variety of clients,” said Sinha.

While Sinha didn’t mention the amount of memory, it looks like HANA has the edge, as the system is reportedly scalable up to 16 TB. Oracle offers a TB in Exalytics, though it can piggy back that appliance with other Oracle boxes to further boost memory further. The acid test here will be real-world customer successes boasting verifiable amounts of memory in specific use cases.

Despite the competitor bashing that is going on about how different they are from each other, both approaches appear essentially similar. They both tune hardware and software to work together to analyze data. Oracle owns the hardware, while SAP uses partners to provide it. Only time will tell who has the most workable solution. Users are advised to see which functions best fit their own needs as opposed to believing the marketing messaging.

For example, Massimo Pezzini, an analyst at Gartner, advises users to take a cautious approach in adopting HANA.  SAP users, he said, should plan for a migration of their NetWeaver-based applications to SAP's HANA architecture within the next three to five years. In the meantime, he suggested users should adopt HANA technology primarily for high-return/fast ROI projects or for non business-critical applications.

Bottom line: Pezzini feels HANA is still a work in progress. Expect SAP to make major changes moving forward and for this to cause some issues among the user base until it fully matures.

Future Directions

Where are these vendors heading with all of this?

SAP is combining in-memory, cloud and mobility technology to drive what it calls its next-generation architecture of applications. These will be available on-premise and on-demand. HANA is a centerpiece of SAPs envisioned application cloud. Business Objects is already offered in this fashion.

The other element of SAP’s strategy is project "River," a platform-as-a-service (PaaS) environment for cloud developers. The idea is to eliminate complexity and make it easy for users to consume SAP applications.  

Oracle has a similar strategy erected around its Fusion applications, which it has been rolling out in recent months. Oracle is retooling its entire application stack and engineering hardware and middleware to work hand-in-hand with software to do one thing really well.

In the end, users will be the beneficiaries of two heavyweights going head to head to see who can built the fastest, most efficient and most cost-effective in-memory architecture.

“You need at least eight hours currently before data is available in the BI tools associated with ERP,” said SAP's Klein. “In-memory analytics reduces that from hours to a second.”

In part two, we discuss the in-memory approaches of firms including IBM, SAS and Birst.

Drew Robb is a freelance writer specializing in technology and engineering. Currently living in California, he is originally from Scotland, where he received a degree in geology and geography from the University of Strathclyde. He is the author of Server Disk Management in a Windows Environment (CRC Press).

Submit a Comment

Loading Comments...