SAP, IBM team up on in-memory analytics

SAP and IBM tout benchmarks for HANA on Big Blue hardware, underscoring their rivalry with Oracle

SAP announced Friday it has integrated its in-memory HANA (High-Performance Analytic Appliance) with IBM's DB2 database, a move that underscores the companies' increasing alignment against rival Oracle and its Exadata platform.

The integration is enabled by SAP's Sybase Replication Server, which SAP gained through the acquisition of that company last year. DB2 is now "HANA-ready and can efficiently replicate data into SAP HANA in near real time," according to a statement.

[ Explore the current trends and solutions in BI with InfoWorld's interactive Business Intelligence iGuide. | Discover what's new in business applications with InfoWorld's Technology: Applications newsletter. ]

So far, SAP has made no similar announcement regarding Oracle's database, which is a mainstay in many of its customers' ERP (enterprise resource planning) environments.

Also Friday, the companies released benchmark figures derived from a HANA installation on IBM 3850-series servers. The system used an X5 server with 32 cores, 0.5TB of memory and a RAID 5 disk array, according to a statement.

For testing purposes, the companies used SAP ERP data from sales delivery database tables. HANA was able to deliver 10,000 queries each hour on 1.3TB of data and brought back results in seconds, the companies claimed.

The results were achieved without assists like pre-aggregating the data or database tuning, in order to reflect real-world qualities of enterprise information, SAP said.

Such figures reflect both the promise and current limitations of HANA, according to analyst Curt Monash of Monash Research.

"If you only care about conventional business intelligence and only care about SAP data, SAP HANA could be a great choice for performance," he said. "There are multiple ways to get great business intelligence performance. If your use case is sufficiently limited, HANA is one of them. But the core idea of throwing things in RAM to make them run fast is legit."

It's not surprising that SAP moved to link HANA with IBM first, Monash added. "SAP and IBM are in tight co-opetition, with Oracle as their joint mortal enemy," he said. Still, "if HANA ever becomes sufficiently compelling, Oracle will have to partner with it."

SAP and IBM's announcement speaks to HANA's immaturity in another way, albeit only when one reads between the lines, according to Forrester Research analyst James Kobielus.

"What this press release is referring to is not utilization of DB2 for storage/persistence within the HANA platform," he said in an e-mail. This will eventually happen when HANA is integrated with SAP's Business Warehouse product, "which in fact does leverage the storage, indexing, partitioning, query and other features of third-party DBMSs," he added.

Friday's announcement "merely refers to HANA's ability to replicate data (via the Sybase tool) from external instances (non-BW-integrated) of DB2 (into HANA's in-memory columnar database) and/or from SAP ERP apps that themselves integrate with DB2," Kobielus said.

HANA currently consumes information from most databases via the ETL (extract, transform, and load) process, a spokesman said. The near-real-time data synchronization provided by Sybase Replication Server will be enabled for databases besides DB2 in the future, he added.

Overall, SAP has placed high hopes on not only HANA but an array of specialized in-memory applications, a number of which were announced earlier this week at an event in Boston. The company is expected to reveal yet more details of its plans in May at the annual Sapphire conference in Orlando.

Chris Kanaracus covers enterprise software and general technology breaking news for The IDG News Service. Chris's email address is

Copyright © 2011 IDG Communications, Inc.