It's easy to see why Oracle is hungry to get a piece of the enterprise search pie: Sue Feldman, an analyst at IDC, estimates the 2005 market was about $900 million and is expected to grow by more than 20 percent a year.

With that kind of money at stake, niche players such as Autonomy and Fast Search & Transfer are getting nervous, and with good reason, said Feldman in Framingham, Mass. "[Oracle] set up an incredible collection of search experts to develop an enterprise search product that has all the usual search technology and some interesting stuff as well," she said. That includes phrase identification (the ability to discern a word's accuracy in relation to its proximity to other phrase words) and the ability to identify the names of people, places and things.

Those capabilities may sound like no-brainers, but they're lacking in search products offered by Yahoo, Google, MSN or America Online, although such consumer search players will likely offer it soon, Feldman said.

The problem with Oracle search came down to scalability. "We'd gotten up to 60 or 70 queries per second out of an Oracle database," said Larry Korb, a lead architect for AutoTrader.com, in Atlanta. "We found to add [features] was very, very difficult with Oracle. It just couldn't scale."

Two years ago, AutoTrader. com wanted to tweak Microsoft's SQL to get union joins-the combination of multiple queries into one query-and subqueries. Such search finessing would have given site visitors aggregated views of, for example, one Ford F150 truck that's representative of the 30 trucks in stock at a given dealer, Korb said.

Korb said that in testing Oracle ConText, AutoTrader.com found that adding more search criteria made the search product scale "less and less."

Source: 1

Prashanth Rai

Tag(s):Oracle, Search