On Mon, Dec 8, 2014 at 5:56 PM, Roger Binns <rog...@rogerbinns.com> wrote: > On 12/07/2014 04:43 PM, David Barrett wrote: >> so I'm curious if you can think of a way using the API (or any >> other way) to essentially "nice" the process by inserting a short >> "sleep" into whatever loop runs inside the VACUUM command. > > Using OS provided functionality will be the most reliable. Other than > that, a simple way is to sleep in the progress callback, although that > will make I/O lumpy.
I wonder whether I/O "sleeping" possible in the first place. Correct me, but what we usually call "sleeping" is about CPU that already sleeps during most I/O operations waiting for rotating media finishing its slow tasks. As a consequence, the more fragmented the data on disk, the less relative cpu time will be spent trying to read and write data. In this case the strategy might be to measure cpu consumption value for consecutive blocks of data and if it's 100% (or other heuristically adequate value) then never sleep (this probably means either cached data or non-fragmented data on a fast disk). But when the cpu consumption drops significantly (much time spent waiting for I/O), the sleeping indeed might be needed. Max _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
