dmurdoch@watstat.waterloo.edu (Duncan Murdoch) (03/06/90)
I'd like to measure the performance of a couple of disk cache programs in real use. The obvious measurement as to how well they're performing is the number of cache hits versus misses; however, neither of them keeps track. (I've got MS SMARTDRV.SYS and Mace MCACH-EM.SYS. According to the near-useless test supplied by Mace, they're about equal. But who cares how long it takes to read a long file backwards?) It would seem straightforward enough to write: I'd install a TSR or device driver before the cache program that kept count of BIOS disk read requests, and another copy after the cache. The difference in the two counts would tell me how many requests had been satisfied from the cache. Does anyone know of an already existing program to do this? If not, can you see any obvious flaws in the "design" given above? Please reply either by email or by posting here; I'll summarize any information I get by email that doesn't appear in a posting. Duncan Murdoch dmurdoch@watstat.waterloo.edu
ching@pepsi.amd.com (Mike Ching) (03/06/90)
In article <1641@maytag.waterloo.edu> dmurdoch@watstat.waterloo.edu (Duncan Murdoch) writes: > >I'd like to measure the performance of a couple of disk cache programs in real >use. The obvious measurement as to how well they're performing is the number >of cache hits versus misses; however, neither of them keeps track. A data point that you might find interesting is that I just booted my Compaq which has a cache program that keeps statistics and my autoexec.bat does 132 reads of which 26 are from the disk and 106 are from the cache. Seems like ANY cache program would be a big win. Mike Ching
dmurdoch@watstat.waterloo.edu (Duncan Murdoch) (03/06/90)
In article <29381@amdcad.AMD.COM> ching@pepsi.AMD.COM (Mike Ching) writes: > >A data point that you might find interesting is that I just booted my >Compaq which has a cache program that keeps statistics and my autoexec.bat >does 132 reads of which 26 are from the disk and 106 are from the cache. >Seems like ANY cache program would be a big win. How many lines in your .bat? I think DOS goes back to disk to read each line; or perhaps just each line following an external command. But yes, batch files should really improve from a cache. Duncan Murdoch
ching@pepsi.amd.com (Mike Ching) (03/08/90)
In article <1651@maytag.waterloo.edu> dmurdoch@watstat.waterloo.edu (Duncan Murdoch) writes: >In article <29381@amdcad.AMD.COM> ching@pepsi.AMD.COM (Mike Ching) writes: >> >>A data point that you might find interesting is that I just booted my >>Compaq which has a cache program that keeps statistics and my autoexec.bat >>does 132 reads of which 26 are from the disk and 106 are from the cache. >>Seems like ANY cache program would be a big win. > >How many lines in your .bat? I think DOS goes back to disk to read each line; >or perhaps just each line following an external command. But yes, batch files >should really improve from a cache. > There are 8 lines in my autoexec.bat (2 are comments) but there's more going on. Device=cache.exe is the last line in my config.sys file so I assumed the statistics were for the autoexec.bat but if I clear the statistics and run autoexec.bat again, I get 43 reads, 10 from disk and 33 from the cache. Compiling hello, world with Turbo C results in 85 reads, 39 from disk, 46 from cache. Just a few more data points. mike ching