Databases rev up

Microsoft, Sybase blending OLTP, analytics

There's nothing like the human brain. No nonbiological entity can process massive amounts of different types of data, integrate the data, analyze it, and respond to it in a split second. Yet to hear the major database vendors tell it, they're coming close.

To that end, some database providers are working toward the convergence of OLTP (online transaction processing) and analytics in a single database that handles both transaction and analysis simultaneously.

The marriage of analytics and OLTP, indeed, goes hand in glove with the move to consolidate systems in the datacenter, according to Barry Shuler, CTO of Marriott International.

"We want to store and retrieve data faster and take it out of a warehouse. There is no need to have separate reporting from the transaction system. We want to run the transaction and reporting system on a single physical entity," Shuler said.

In an exclusive to InfoWorld, Sybase revealed that this fall it will release ASE (Adaptive Server Enterprise) 15, an OLTP relational database that will include a sophisticated query-processing engine with algorithmic similarities to the analytics done by Sybase IQ, its data-warehousing tool. 

"It's not replacing IQ, but it is expanding what ASE can do in a mixed environment," said Tom Traubitz, senior marketing manager at Sybase.

By mixed environment, Traubitz is referring to changing requirements in telecommunications, health care, and retail. In these industries and others, customers want a database to process, read, and write data, as is typical of an OLTP database, while simultaneously retrieving and analyzing the data, in a fashion similar to data warehouses.

Microsoft also is moving slowly and steadily toward convergence of OLTP and analytics, according to Tom Rizzo, Microsoft's director of product management for SQL Server. Rizzo said that the forthcoming version of SQL Server, due in the second half of 2005, will have analysis services on top of relational rows and columns, and nonrelational, unstructured data. Within the ETL (extraction, transformation, and loading) tool, Microsoft will integrate the analytics right in the pipeline as the data is flowing in real time, Rizzo added. 

In addition, Microsoft will add what it calls "proactive caching."

In this design, as data comes into the OLTP system it is cached and aggregated according to standard statistical algorithms that perform sum, deviation, min and max, average, and ceiling, Rizzo explained.

"We can slice the data according to whatever you tell us," Rizzo said.

Representatives from IBM and Oracle claimed to do both OLTP and analytics now, but Donald Feinberg, vice president and distinguished analyst at Gartner, said it just isn't so.

"They're dreaming," Feinberg said about any database vendor making that claim today. Feinberg points out it can be done on a small to midsize database, but in an enterprise data warehouse with thousands of OLTPs, the database can't do both -- just yet.

"It won't be possible [in a large database] until 2010," Feinberg added.