The Transaction Processing Performance Council will abandon its TPC-A and -B benchmarks on December 6 next year when it withdraws all -A and -B results. No new -A and -B numbers will be accepted after June 5 1995. The reason? Cost and desire to get vendors up on to the more complex, ‘real-world’ TPC-C suite. Benchmark-watcher Benchpress observes that the council might be a tad hasty given that TPC-A and -B numbers are being withdrawn faster than TPC-C marks are being added. It also argues for a unified benchmark that could gradually incorporate all of the Council’s planned series of benchmarks, so that fewer of the high configuration costs have to be met. The standards body says that this simply will not work because the market is too complex and vendors want more specific comparison tests. It claims there are a slew of TPC-C results coming through – plus many more that vendors have not published because they are not commercially advantageous enough – their systems are slow. It recognises, however, that the costs of configuring vast amounts of multiple client equipment for the benchmark is hampering its use, and it is currently reworking TPC-C as a server-oriented suite. Also forthcoming are TPC-D benchmarks in the first quarter of 1995 for measuring decision support performance. TPC-E for enterprise server environments, plus a server-only versions, is due in the second quarter. An unnamed database benchmark is being prepared for the second or third quarter and a client-server benchmark for the fourth quarter. By the end of 1995 it will have five benchmarks but estimates it will be mid-1996 before results on all come to be published: it is still waiting for first Windows NT-based results.
