DISK BENCHMARK UTILITY Release Level 1.01.029 The Disk Benchmark utility is used to measure the performance of CTOS disk drives. This application consists of one runfile: DiskBenchmark.run Product Name : Supported On : B24, B26, B27, B28, B38, B39, and all SuperGen hardware Supported OS : BTOS II, CTOS I, CTOS II, CTOS III Description : The tested volume is accessed read-only, no writes are done during the tests. The measurement includes all CTOS overheads (and some program overheads) in addition to the actual low-level performance of the drive. The results are close to what a standard CTOS environment program can expect for reads of single 512-byte sectors. The purpose is to provide a basis for the comparison of different drives rather than to validate specific manufacturer's claims. Standard CTOS calls are used, thereby allowing the program to be run from any station in a CTOS network, measuring the achievable speed of any accessible drive in the network. When testing a non-local drive, the accesss times given include any network overheads as well. Note: The use of CTOS file system caching will give impressive results which are generally invalid. Directions/Use : The Disk Benchmark utility may be run by executing the /Disk Benchmark command. The following list of field names and explanations describe the use of the program: Volume Id This required field may be a logical volume name such as [Sys] or [Scr], a volume name such as [MyVol] or [Win], or a device name such as [D0] or [s0]. The Volume Id must include the square brackets [ ], and may be prefixed with a node name (eg. {NodeName}[Sys]). If the current path password does not allow the opening of the volume for read access, the volume password must be appended to the Volume Id entry (eg. [Win]^SysPassword). [# of Seconds (20)] This parameter is optional, the default is 20 seconds. This specifies the amount of time that each of the 5 tests is run for. Specifying a time, allows you to control the length (and therefore accuracy) of the test (very handy if testing WAN access speeds). [Q Depth (1)] This paramter is optional, the default is 1 entry in the Queue. This specifies the number of simultaneous requests the program will issue before waiting for I/Os to complete. This allows simulation of heavy I/O situations which arise in a multiple requestor environment. This generally shows the optimization that CTOS performs. It can also be used to eliminate some network overhead and determine actual performance of non-local disks. Note that Q depths of other than 1 can give misleading results for the first four (rather artificial) tests. [Short seek #Trks (1)] This parameter is optional, the default is 1 track. This is to allow testing of other than a single track seek. [Log File] This parameter is optional, the default is [vid]. This allows the results of the test to be written to a file or printer. Description of results: This test run was done against a volume on a remote LAN node. Ping's Disk BenchMark - V1.01.029 Testing Volume {nztce01}[sys] To ensure accuracy of results, there must be no other users of this volume Disk Information: Volume Name = NZTCE01 Sectors Per Track = 32 Tracks Per Cylinder = 15 Cylinders Per Disk = 998 Disk Capacity = 245268480 bytes Q Depth = 1, sampled for 2 seconds Average Latency + Transfer = 62.9 milliseconds ( 35 I/O's) Short Cylinder Seek = 54.8 milliseconds ( 42 I/O's) Max Cylinder Seek = 63.9 milliseconds ( 36 I/O's) 1/3 disk Seek = 64.7 milliseconds ( 34 I/O's) Random Seeks = 73.3 milliseconds ( 30 I/O's) The 'Average Latency + Transfer' is the average time measured for random sector reads within a single cylinder. This is actually misnamed as it includes the measurement program, CTOS (and network) overheads. The 'Short Cylinder Seek' is the average time measured for random sector reads between two tracks. For the default, these will be tracks 1 and 2 (track 0 is not used, as this causes problems with ECC drives). The 'Max Cylinder Seek' is the average time measured for random sector reads between two tracks. These will be tracks 1 and the last track on the device (track 0 is not used, as this causes problems with ECC drives). The '1/3 disk Seek' is the average time measured for random sector reads between two tracks. For the default, these will be tracks 1 and track (Max/3) (track 0 is not used, as this causes problems with ECC drives). The 'Random Seeks' is the average time measured for random sector reads over the entire disk capacity (excluding track 0 for reasons that by now should be obvious). This is the truest indication of the performance your applications could achieve. Known Bugs: The program selects sectors at random, it does not exclude known bad spots when selecting a sector to read. If a selected sector is a bad spot, invalid results may be obtained. COMMAND Disk Benchmark Volume Id [# of Seconds (20)] [Q Depth (1)] [Short seek #Trks (1)] [Log File]