home *** CD-ROM | disk | FTP | other *** search
-
- MDisk is not like other disk benchmarks -- its primary
- purpose is not to measure basic disk subsystem performance.
- MDisk is an attempt at quantifying two parameters not measured
- by any other disk benchmark that I have seen: the amount of cpu
- time consumed by disk I/O; and the I/O performance of two drives
- at the same time, particularly when using two physical hard
- drives.
-
- Cpu usage is unimportant under Dos, as Dos will not allow
- anything else to use the cpu during disk access. Under a
- multitasking os such as OS/2, this time can be used by other
- processes. The net effect is overall cpu performance equivalent
- to what one would get under Dos with a somewhat faster cpu.
- OS/2 does not provide a direct way to measure how much cpu
- time is used by a particular process/thread. MDisk makes a
- rough measurement by running a thread at idle priority which
- basically counts how many times it loops during each test. This
- thread is run once by itself to obtain a baseline figure, and
- again during each disk test. The results of each run during
- disk I/O is then used to caculate a percentage of available cpu
- time relative to no disk I/O conditions.
-
- The performance of simultaneous I/O to two drives is
- intended to show the benefit of advanced disk subsystems.
- Setups with multiple SCSI drives should provide some interesting
- results; those with multiple ST506 type drives spread across
- multiple controllers may also show benefits.
- The disk I/O thread is based on IOSTONE. This benchmark
- uses a set of test files that total about 1.5 MB, resulting in
- test results that are heavily influenced by the amount of disk
- caching. Numbers obtained with cache sizes larger than the test
- files will not bear any resemblence to the performance one would
- get when using files significantly larger than the cache.
-
-
- Have fun!!!
-