Objectives: Introduction Over View of System Analysis and Design
Download 0.94 Mb. View original pdf
|
- Navigate this page:
- HARDWARE SOFTWARE
- Table 10.1 Representative benchmarks for hardware software. 10.4.3 Comparison of Benchmarks
| 10.4.2 Design of Synthetic Programs A synthetic job is a program written to exercise a computer’s resources in away that allows the analyst to imitate the expected job steam and determine the results. Then the artificial job stream can be can be adjusted and rerun to determine the impact. The process can be repeated, as many times as necessary to see which tasks a comparison set of computer handles well and which they do not handle as well. The synthetic jobs can be adjusted to produce the same type of activity as actual programs, including perhaps random access of files, sequential searching of files with varying size records, input and output activities, and file accessing in varying random patterns. The types of hardware and software features that are often simulated are listed in Table 10.1. HARDWARE SOFTWARE CPU processing speed. Memory access speed. Interrupt handling abilities. Peripheral channel speed. Printer speeds. Seek time for magnetic disk. Rotational delay for magnetic disk. Communication speeds. Scheduling algorithm. Compilation algorithm. Code efficiency. Virtual storage management algorithm File handling efficiency. Interrupt handling. Indexing methods. Multiple buffer handling. Communication processing procedure Table 10.1 Representative benchmarks for hardware & software. 10.4.3 Comparison of Benchmarks Although some comparison on the basis of equipment performance is better than no comparison at all, there are drawbacks to the use of benchmarks, first of all, the comparisons are made on purely quantitative grounds. They do not relate the learning time needed to become accustomed to the system or the quality of the systems software such as the quality of the diagnostics produced during compilation or the efficiency of the object code produced. In addition, benchmarks do not provide reasonable assurances that programs currently being used on an existing system can be converted to the new system or that the new machine will run them efficiently even if they are converted. Vendors may also make sales claims that a specific system can handle additional tasks that another system cannot. Since benchmarks cannot directly verify these claims, the purchaser may insist that statements of certain sales claims be attached in writing to the sale contract. Download 0.94 Mb. Share with your friends:
|