--- Dennis Cote <[EMAIL PROTECTED]> wrote: > All, > > More mysteries. To investigate this low insert > performance under WinXP I > wrote a simple test program that loops writing one > character to a file > and then flushing the file. The strange thing is > that it seems to > alternate between two different run times as shown > below. This is for 6 > consecutive runs.
Don't know for sure, but I can offer some educated guesses for the scattering of results: - NTFS is a fairly complex file system (more complex than traditional FS-es), with several tables into which file data must be written and cross linked, as well as a data journal. The scattering of locations for these journals means that seek times are different, depending where the data is laid out on disk. In particular because there's a journal, consecutive program runs never actually write to the same place on the disk even if it seems so to the application. The results you have seen (alternating between two values) can also be provoked on complex database systems (in particular PostgreSQL) with simple benchmarks (e.g. pgbench) on relativly simple non-journaled file systems (such as UFS) when one benchmark run nearly fills a write-ahead log and the log gets processed/commited in the next run. - Windows has many background disk users / writers - Explorer, registry and other components are known to "wake up" periodically and write their data (whatever it is). __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com
