Previous | Table of Contents | Next |
A pseudo-code definition of the debit-credit workload is given in Figure 4.5. There are four record types in the database: account, teller, branch, and history. History records are 50 bytes; others are 100 bytes. Each transaction consists of reading a 10byte message from a block mode terminal connected via X.25. The system performs presentation services to map the input for a
Begin-Transaction Read message from the terminal (100 bytes) Rewrite account (100 bytes, random) Write history (50 bytes, sequential) Rewrite teller (100 bytes, random) Rewrite branch (100 bytes, random) Write message to the terminal (200 bytes) Commit-Transaction |
FIGURE 4.5 Debit-credit transaction pseudo-code.
COBOL program, which in turn uses a database system to debit a bank account, do the standard double-entry bookkeeping, and then reply to the terminal. Fifteen percent of the transactions require access to remote branch databases; others are local.
Due to increased popularity of the transaction processing systems, an effort is underway to define transaction processing benchmarks more precisely. A council of several transaction processing system vendors, users, and consultants, called the Transactions Processing Performance Council (TPC), was formed in August 1988. Its first benchmark, called TPC Benchmark A, is a variant of the debit-credit benchmark (see TPC 1989). In TPC benchmark, the throughput is measured in terms of TPS such that 90% of all transactions provide 2 seconds or less response time.
The Systems Performance Evaluation Cooperative (SPEC) is a nonprofit corporation formed by leading computer vendors to develop a standardized set of benchmarks. Release 1.0 of the SPEC benchmark suite (see SPEC 1990) consists of the following 10 benchmarks drawn from various engineering and scientific applications:
These benchmarks, which stress primarily the CPU, Floating Point Unit (FPU), and to some extent the memory subsystem, are meant for comparing CPU speeds. Benchmarks to compare I/O and other subsystems may be included in future releases.
The elapsed time to run two copies of a benchmark on each of the N processors of a system (a total of 2N copies) is measured and compared with the time to run two copies of the benchmark on a reference system (which is VAX-11/780 for Release 1.0). For each benchmark, the ratio of the time on the reference system and the system under test is reported as SPECthruput using a notation of #CPU@Ratio. For example, a system with three CPUs taking only 1/15th as long as the reference system on GCC benchmark has a SPECthruput of 3@15. The ratio is a measure of the per processor throughput relative to the reference system. The aggregate throughput for all processors of a multiprocessor system can be obtained by multiplying the ratio by the number of processors. For example, the aggregate throughput for the above system is 45.
The geometric mean (discussed in Section 12.5) of the time ratios for single copies of the 10 benchmarks is used to indicate the overall performance for the suite and is called SPECmark. Again, the notation #CPU@Ratio is used to report the SPECmark.
The TPC and SPEC are the beginning of a new trend in the industry to develop standard benchmarks for comparing all types of computer systems including networks, image processing systems, and databases.
Previous | Table of Contents | Next |