IBM joins in-memory database party

IBM has joined the in-memory database (IMDB) party after launching the IBM solidDB, which it claims will deliver data 10 times faster than a conventional relational database.


IBM has joined the in-memory database (IMDB) party after launching the IBM solidDB, which it claims will deliver data 10 times faster than a conventional relational database.

It is worth noting that in memory databases are not new technology, and IBM makes no secret of the fact that it has used the technology from Finnish company Solid Information Technology, which it acquired back in December 2007.

Solid made a name for itself selling an embedded database with an in-memory database engine (IMDB), which meant it could store and retrieve data from main memory, giving faster performance than traditional disk-based systems. This has made IMDBs popular for applications that require very fast processing times, such as routing calls in a phone network or trading stocks.

IBM solidDB version 6.1 can be deployed as a cache to IBM's main relational database products, namely DB2 and Informix Dynamic Server (IDS). Alternatively, it can also be deployed as a standalone in-memory database. Because it keeps data in main memory, IBM claims that solidDB support tens of thousands of transaction per second with microsecond response times.

To use solidDB as a cache for DB2 or IDS, the user identifies a set of database tables that would benefit from faster access. For example, the user may select only ten performance critical tables such as shopping cart management etc. Once these tables are specified in a schema in DB2 or IDS, the solidDB connector can load the data from DB2 or IDS and make it available to applications at "extreme speed."

Scale is achieved in that solidDB cache can be partitioned across multiple servers. For example a large customer database containing 1 million customers can be partitioned into four solidDB Cache instances (each storing 250,000 customers).

Reliability is promised thanks to a two node, hot standby configuration, which allows solidDB to maintain two copies of data that is synchronised at all times between the solidDB nodes.

"We offer high availability," said Paola Lubet, director of product marketing for IBM solidDB. "It can operate on two node environment, and transactions can shift over to second node in case of hardware failure, in less than one second."

But there is no doubt that IBM's rivals have already been involved in the IMDS market for some time now. Oracle, for example, acquired Solid Information Technology competitor . Open source IMDB competitors also exist, such as MonetDB, H2, and HSQLDB. However, some feel that IBM's late entry to this market is not much of a setback.

"With acquisitions you cannot such flip a switch," said Carl Olofson, Research VP, Information Management and Data Integration Software Research, at analyst house IDC. "In truth, the IMDB market is actually still very immature."

He estimates that IMDB occupies only somewhere between 0.5 to 1 percent of the total database market.

"IMDBs are a means to an end initially," Olofson continued, pointing out that IMDBs should not be viewed as a replacement for enterprise databases. "It is only in certain circumstances when IMDBs are a requirement. It is a means to an end, and fills out a vendor's data management portfolio."

IBM's IMDB product is scheduled to ship on 24 June, but IBM refuses to disclose pricing at this stage. Traditionally, IMDB have been on the expensive side, but Lubet insists that this is no longer the case.

"We are not disclosing pricing per se," Lubet told Techworld. "But we are cheaper than disk-based databases from some enterprise market leaders." Indeed she claims that in some cases the IBM offering is 30 to 40 percent cheaper.

"Recommended For You"

IBM releases DB2 version 10, the first big upgrade in four years IBM zEnterprise EC12 mainframe eyes Big Data analytics and hybrid clouds