SAP Talks Up HANA In-Memory Database at TechEd 2011
September 13, 2011 Alex Woodie
SAP has big plans for HANA, the high-performance analytical appliance that will receive lots of attention and scrutiny at SAP’s TechEd 2011 show this week in Las Vegas. And the interesting thing is, not all of SAP’s plans have to do with business intelligence. Sure, the bulk of the focus today is on how quickly HANA’s in-memory design can rip analytical workloads. But SAP executives see a time when HANA–with its slim columnar structure–eventually replaces the relational database that underlies the ERP system.
The relational databases that are popular today were designed 20 years ago and were largely shaped by the hardware limitations that existed then, says Amit Sinha, SAP’s vice president of in-memory computing and HANA solution marketing. “If you were to design a new modern database today, would you use the same application constructs when the constraint was the access between disk and memory? Probably not,” he tells IT Jungle.
So SAP engineers hit the drawing board and came up with a new database architecture that could take advantage of the huge improvements in the price/performance of X86 processors and memory. The result was HANA, an in-memory database that supports a flat two-dimensional structure, and is much simpler than the complex star schema used by most RDBMSes today, SAP managers say.
HANA’s first task is to perform analytical workloads, such as generating reports. SAP partners with system vendors like IBM and HP to create HANA-based appliances that would function as a side-mart of data. Because the side-mart is constantly refreshed with data replicated from the production system, and because HANA’s in-memory engine can perform calculations so quickly, the product was positioned as one of the first of a new breed of “real time” analytic machines.
There is no question that today HANA’s focus is analytics. After all, it has the word “analytic” in its acronymious name. But that doesn’t have to be the case. SAP executives have designs on shaking up the database industry, which by definition would shake up SAP’s archrival Oracle.
In fact, SAP executives see no reason why HANA couldn’t power the next generation of applications. “This artificial boundary between the application layer and the database tier we think is going away,” Sinha continues. “We think a lot of core calculations at the application layer can be moved to the database tier and optimized. When it’s closer to the database tier, the calculations can happen faster in memory, and without the need to move the data up and down between the application tier and the database tier to perform those calculations.”
To that end, SAP is developing a library of about 75 business functions for HANA. These calculations can be complex, such as performing currency conversions based on the current rate. These functions are often programmed at the application level, and moving them into the database would provide a performance boost.
After HANA proves its chops with analytical workloads and the library of business functions, SAP may push harder for HANA to take over everyday transactional workloads. That is absolutely the plan, Sinha says. “HANA is intended to … eventually become the database for the application,” he says. “The long term plan is it becomes the database underneath the [SAP] Business Suite.”
But first, HANA must prove itself with the task at hand, and that’s analytical workloads. At TechEd, SAP unveiled the second and third pre-packaged analytical applications based on HANA. The solutions include Smart Meter Analytics, a system designed to provide electrical companies with real-time insight into different aspects of their customers’ electricity consumption, and the SAP ERP rapid-deployment solution for profitability analysis with SAP HANA.
SAP’s roadmap calls for additional HANA applications, including such corporate favorites as cash management, sales and operations planning, trade promotion management, merchandise and assortment planning, and a demand signal repository.
HANA will also spearhead SAP’s move into the consumer space (yes, you read that right) with a Recall Genie application that tells consumers what recalled products to avoid buying. A personal energy application is also in the works. These apps will likely feature iPhone or Android client front-ends, but the hard core analytical data crunching and sorting will be performed on a HANA system running on the other end of the cloud.
SAP has sold hundreds of licenses for HANA, and has a pipeline of about half a billion Euros, SAP managers say. One of the first HANA implementations occurred at Nongfu Spring Co., the largest distributor of bottled water in China.
Nongfu Spring implemented HANA to speed up its reporting system, which was previously based on an Oracle 11g data mart, according to Ken Tsai, senior director of platform marketing at SAP. Stored procedures and other SQL functions that previously took 24 hours to process in the old system were finished in as little as 37 seconds, SAP says. Overall, the query performance of the HANA system was 200 to 300 times faster than the Oracle system, using the same SQL scripts.
It’s not surprising to see an in-memory database outperforming a database designed to use disk. The large database vendors, including Oracle, IBM, and Microsoft, have all made investments in in-memory databases. But it’s interesting and refreshing to see SAP putting such a big emphasis on the technology.