SAPs move to expand its in-memory technology beyond its business intelligence confines to broader enterprise data sets has sparked debate over the potential threat to database vendors.
eWEEK reported in its Aug. 7 issue that, according to sources close to the company, SAP is broadening the scope of its in-memory technology to areas where users need fast query and transaction capabilities and, in the process, is potentially disrupting Oracles core relational database business (about 55 percent of SAPs customers applications sit on an Oracle database) by negating the need to store data externally.
The consensus, among database vendors at least, is that SAPs plan wont work. While there is definite value in utilizing in-memory data capabilities, the underlying supporting data needs to reside in a standard relational database.
“An in-memory database is limited by the available RAM,” said Steven Graves, president and co-founder of McObject, which develops the eXtremeDB in-memory database system, in Issaquah, Wash. “With 64-bit memory its possible to have a terabyte size, but the time it takes to provision it is rather large. And also theres the question of the survivability of the database. If someone trips over a cord, that in-memory goes away. So in-memory is not going to replace conventional databases; it can work side-by-side and replace real time,” database needs.
Indeed, both McObject and MySQL are working on hybrid models that will combine in-memory database technology with traditional data storage capabilities.
In beta now, the 5.1 release of MySQL Cluster—the companys in-memory offering—will allow users to store in memory data that overflows the databases capabilities. “The new release says, These objects need to reside in memory all the time, but for others that will exceed the amount of memory, they can be stored,” said Robin Schumacher, director of product management at MySQL, in Cupertino, Calif. “Depending on how well the beta goes, it should be available, conservatively, [in] the first half of 2007.”
Separately, a beta version of McObjects eXtremeDB in-memory database will enable users to indicate which data should be in-memory and which needs to be persistent, according to Graves.
“We created it because there are many different applications that have both transient and persistent data,” said Graves. “When eXtremeDB was purely in-memory, it had to use two different products. You could periodically check point data or use our transaction logging edition, but neither was as elegant as a single product. And neither addressed the scalability issue, for the database to scale beyond the available store.”
While SAP has no designs to be a database vendor—it tried and failed with MaxDB, which is now part of the MySQL community—sources close to the company have suggested that SAP could be working on a hybrid approach or something similar. “Basically there is no reason SAP couldnt work with IBM and others on making [in-memory] work well on hardware,” said a source close to the company who requested anonymity. “Probably thats the company that would be the most benefited. The vendor that [it] would be most disruptive for would be Oracle.”
“Our applications generate a lot of data. We are not a database company, but when it comes to generating data, we know what we are doing there,” said Vishal Sikka, SAPs chief software architect, in Walldorf, Germany.
Oracle, however, isnt staking its future on a hybrid approach. Rather, its taking what it has up and down the technology stack—traditional and in-memory database technology (the latter from the 2005 acquisition of TimesTen), middleware and applications—and optimizing each layer to work together.
“If you take a look at where TimesTen comes from—technology that from the very beginning was designed to be embedded inside a system versus running on a dedicated server—its very natural to run it on the same system as the application server and use it for caching,” said Jim Groff, senior vice president of business strategy in Oracles Database Server Technology Group, in Redwood Shores, Calif., and the former CEO of TimesTen.
Rather, Groff seemed to suggest that Oracle is looking at embedding the TimesTen in-memory capability at the middleware layer (Oracles Fusion Middleware, that is) and, in return, being able to tap in-memory capabilities at both the metadata and application-specific level.
“If you look at Fusion, theres a huge role for metadata and caching in the middle tier,” said Groff. “Theres also real advantages using in-memory to cash data and allow it to be accessed in real time, for example with products, using it to pull a product catalog forward, or in manufacturing to pull a BOM [bill of material] forward.”
Groff said he sees in-memory playing in a number of different fields, from defense and intelligence with the real-time tracking of everything from supplies using RFID to threats, to manufacturing and logistics. “Now theres information about individual items—toothpaste—that generates a step-function increase in the amount of data,” said Groff. “All these things are about real time.”
When asked if SAP could feasibly have a similar plan to utilize its in-memory capability within its NetWeaver middleware platform—at both the metadata and application-specific level, Groff said, unequivocally, no. “Its possible to take something and claim it is in-memory and have it be very application-specific, but the real power is to have it be standards-based and in-memory,” he said.
Oracle, in the process of rebuilding its Java-based Fusion Application stack to sit on top of its Fusion Middleware stack—and likely its database stack as well—is relying heavily on standards as a big differentiator in the technology market.
At least one SAP partner believes SAPs deeper inclusion of in-memory technology—whatever it winds up looking like—can only be a good thing.
“If SAP started making a nice suite of BI [business intelligence] capabilities that worked with streaming data [akin to in-memory], thats going to be great for us,” said Gregg LaBlanc, technical strategist at OSIsoft, which develops a real-time performance management platform for various manufacturing industries. “NetWeaver has a nice large queue of data—a big playground for all these applications to work together—so if they have a big cube for all this data to work together, with data streams coming in from all over …, we would like to plug into that from the manufacturing end.”
LeBlanc said a SAP in-memory offering would probably end up as some sort of server that has database-like capabilities. “It would probably be a set of NetWeaver XI interfaces—ways to get data in and out of the server—with a big calculation engine hung on it,” said LaBlanc, in San Leandro, Calif. “It would have to be pretty high performance. It would have to be a separate box.”