問66
頻出度★★★
システム評価の方法であるベンチマークテストに関する記述として, 最も適切なものはどれか。
平成23年度 特別試験 ITパスポート試験 問66
テクノロジ系システム構成要素 › システムの評価指標
This question's explanation and your answers so far.
解説
問 66
Incorrectア
This describes availability. Availability is expressed as the proportion of the total time during which the system is operating normally, and it is calculated as MTBF ÷ (MTBF + MTTR). The statement itself is correct as a definition of availability, but what it evaluates is availability, namely whether the system can be used without going down, not processing capability, namely how fast and how much it can process. It also lacks the procedural feature of a benchmark test, which is to actually run standard programs and data and measure the result, so it is wrong.
- イ
正解 The correct answer is イ. A benchmark test is a method of evaluating the processing capability, that is, the performance, of a system: you prepare standard programs (benchmark programs) and standard data, actually run them on the system being evaluated, and measure things such as the time required for processing and the throughput. Choice イ says exactly that you “evaluate the processing capability of a system using standard programs and data”, which matches this definition directly. There are two key points about a benchmark test. First, it is a measurement-based evaluation that actually runs programs, not a theoretical value on paper or a number from a catalog spec sheet. Second, the programs and data used as the load are standardized so that everyone measures on the same footing. Because they are standardized, machines with different CPU and memory configurations can be compared relatively on the same scale, which makes benchmarks useful for selecting equipment (comparing performance at procurement time) and for confirming performance improvements before and after a system change. Well-known benchmarks include the TPC family, which measures online transaction processing performance, and the SPEC family, which measures CPU computing performance. Keep in mind as well that a benchmark cannot fully reproduce the load characteristics specific to your own organization, so a benchmark value does not by itself guarantee the performance of the production environment.
Incorrectウ
This describes MTBF (Mean Time Between Failures). It is the mean time from when the system recovers and starts running again until the next failure occurs and stops it; the larger the value, the less often failures occur, meaning higher reliability. The statement is correct as a definition, but what it evaluates is reliability rather than processing capability, and its methodology itself differs from a benchmark test, which measures performance by actually running a standard load, so it is wrong.
Incorrectエ
This describes MTTR (Mean Time To Repair). It is the mean time required from when the system stops because of a failure until it is restored; the smaller the value, the faster it can be repaired, meaning higher maintainability (serviceability). The statement is correct as a definition of MTTR, but what it measures is the speed of recovery from a failure, not the amount or speed of processing the system can handle per unit of time. It has nothing to do with measuring performance with a benchmark program, so it is wrong.
ここがポイント
- Definition of a benchmark test — a method that actually runs standard programs (benchmark programs) and standard data, measures the time required and the throughput, and thereby evaluates and compares the processing capability (performance) of a system. Because the load is standardized, different machines can be compared relatively, which is used for equipment selection and for confirming performance improvements. Well-known examples are TPC and SPEC.
- Telling performance evaluation apart from RASIS indicators — availability is expressed by the availability ratio (MTBF ÷ (MTBF + MTTR)), MTBF expresses reliability (the mean time from recovery until the next failure), and MTTR expresses maintainability (the mean time from a stop until restoration). Their axis of evaluation differs from that of a benchmark test, which measures processing capability. Sort out this type of question by asking what is actually being evaluated: processing capability, availability, reliability, or maintainability.
用語