More details emerged Wednesday about SAP's plans to build new in-memory database appliances for high-speed analytics, which could set the vendor up against platforms like Oracle's Exadata.
SAP is currently seeking a group of initial test customers to help finish up development on the appliance before a general release by the end of the year, SAP CTO Vishal Sikka said during a keynote address at the Sapphire conference ongoing in Orlando and Frankfurt, Germany, which was webcast. It is working with Hewlett-Packard on the system's hardware elements, according to a statement.
The appliances will be compatible with SAP's modern ERP (enterprise resource planning) platforms, as well as older versions like R/3, Sikka said. He repeatedly stressed that customers will be able to adopt the appliances without radical effort or changes to their systems.
"Enterprises today are looking to become unencumbered, are looking to become unbound, to unleash the power of their existing landscapes. And we will do that without bringing disruption to the systems you already own," he said.
Many of SAP's customers have been interested in adopting in-memory technology but due to the recession, have lacked money, people and time, company co-founder Hasso Plattner said later in the keynote. "And even if they had these three assets, they don't want to risk what they currently have."
To that end, SAP is working to make sure the new capabilities work with current and legacy ERP systems, instead of waiting for a future release, Plattner said. "Instead of moving forward like we traditionally do in high-tech, we have to move backward."
Sikka took a barely veiled shot at rival Oracle, which is positioning itself as a one-stop shop for customers now that it has hardware by acquiring Sun Microsystems.
"We believe that customers always look for choice. There are companies that believe you can have a landscape that comes from one vendor...from screen to metal."
SAP believes customers should have "the landscape of the best," he said. "It's not one [vendor]. Never ever."
In-memory databases move information to be processed into main memory, instead of reading and writing it off of disks, giving a performance boost. In the past, the capacity constraints and cost of RAM held back the potential of in-memory processing. That obstacle has been surmounted by the advent of servers capable of containing multiple terabytes of RAM, as well as additional power provided by multicore processors, Sikka said.
SAP will likely initially target its own installed base with the appliances, but future versions "will be much more general and could be sold as 'query accelerators' or 'BI [business intelligence] accelerators' to any BI customer," Forrester Research analyst Boris Evelson said via e-mail. "It is also a hit against any traditional data warehousing vendor without in-memory technology."
The upcoming appliances have some technical advantages over other in-memory products, such as Tibco Spotfire and Microsoft PowerPivot, Evelson added. "Most of the other in-memory technologies are positioned as data marts or analytical sandboxes ... mostly because the size of the data you can load is limited to a single memory space. I haven't seen an in-memory model/data mart from these vendors larger than a few hundred GB."
Meanwhile, SAP's appliance "can have a number of blades and a single data model can span multiple memory spaces. They've tested it with models of 2TB and plan to go to as much as 4TB," he said.
Meanwhile, along with SAP's claimed advanced data compression capabilities, customers could conceivably push enterprise data warehouse-sized amounts of information into memory, Evelson added.
SAP is also planning to create a front end that allows many BI (business intelligence) tools to use it, compared to other platforms which support only their own proprietary front-ends, he said.