Four-month-old start-up Inktomi Corp, of Berkeley, California, staffed by University of California, Berkeley scientists, is planning to cash in on the boom for all things Internet and intranet with its parallel processing search engine and Web crawler products. It has already co-developed the HotBot search engine with Wired Ventures Inc subsidiary Hot Wired Ventures LLC and reckons that it has a bunch of partners in the bag to push its stuff into the mainstream market. Inktomi’s core technology is based on parallel processing work that its founders Eric Brewer and Paul Gauthier developed at Berkeley. HotBot uses a custom-built high-performance primary database integrated with Informix Inc’s Online Dynamic Server for tracking user preference profiles, advertisement placement and accounting. It relies on a network of workstations and Arcadia, Los Angeles-based Myricom Inc’s Myrinet local network technology for communication and packet switching within concurrent and para llel supercomputers. Myricom claims Gigabits-per-second packet communication starting at $1,500 per host. Bandwidth is 160Mb per second per processor with an injection time of 12 microseconds, claims Inktomi. The search engine comes with a software package, Audience 1, including dynamic tags for database access enabling arbitrary persistent information and customised HyperText Mark-up Language tracking and support for SQL. The crawler recognizes multimedia file types including ShockWave, Java and RealAudio. Access to the Inktomi search engine can be encapsulated in a single dynamic tag, hiding the complexities of parallel programs and enabling multi-threaded, interleaving long latency queries and customer content selection. Inktomi reckons HotBot provides a more comprehensive index than the likes of Digital Equipment Corp’s AltaVista search engine, which it estimates covers 30m pages out of the 50m Mark-up Language pages on the Web. It describes the full text index of AltaVista as haphazard, missing lots of entries based on individual words. In comparison, HotBot delivers an index covering the whole of the Web, with a level of fault-tolerance that the single big computers from the likes of InfoSeek Corp or set of independent machines such as AltaVista and Lycos cannot manage because of replication requirements. Point your browser at http://www.hotbot.com.
