By Timothy Prickett Morgan
As part of the Net.Commerce announcements last week, Karl Salnoske, general manager of electronic commerce for IBM’s Software Solutions division told analysts that IBM is firmly behind the forthcoming TPC-W e-business benchmark test and that it intends to be the first vendor out the door with test results on the new e-business benchmark.
The TPC says that the TPC-W test is designed to represent the transactions performed by any e-business – retail sales, airline reservations, stock trading and so forth – as well as web-based intranet systems where end users browse catalogs and buy products. The TPC-W test was launched in January 1998 by the members of the Transaction Processing Performance Council and was supposed to be completed by late 1998. A revised schedule put forth last year had pushed delivery of the TPC-W test to the second quarter of 1999, which has come and gone.
The TPC-W test will have a mix of transaction types, just like the TPC-C test does, including browsing product web pages, registering as a user on a site, using a search engine to find information, placing orders using a shopping cart and requesting the status of an existing order. The test was initially conceived as having both online and batch transaction modes and the same rigorous transaction integrity – the so-called ACID tests – built into the TPC-C test, as well as internet security and authentication procedures. TPC-W will measure web interactions per second (WIPS), and have price/performance ratings just like the TPC-C and its predecessors did.
The test will also have two secondary performance metrics, splitting web browsing hits from web online transactions. IBM will very likely announce TPC-W tests using its Net.Commerce software running on RS/6000 Pulsar servers, if the TPC can get its act together and get the benchmark out the door. IBM will probably also provide TPC-W results for Netfinity and AS/400 servers as well, but probably will never offer results for its S/390 mainframes because their poor bang for the buck would shock the mainframe base, regardless of the reliability of S/390s and their OS/390 software.
