sysmaster question

poking about. I have a table, in a single dbspace, in a single chunk. (Cooked). According to sysptprof, there have been 4.2 million page reads against the table. According to syschkio, there have been 3.6 million pages read from the chunk. Obviously this is a discrepancy, how is it getting resolved? A request is made to disk to read a page, and the page has already been read? Then why doesn't it count in bufreads? This particular table accounts for 88% of my i/o against this disk, so add another 400K or so page reads to the 4.2 million number. Page writes seem to be more consistent with what I'd expect. cheeers j.

Jack, This is to be expected. There are other I/O activities that take place at the chunk/dbspace/disk level that do not pertain to any specific table. Look for John Miller III presentation on Understanding Onstat. I think it was a chat with the lab last year so it will be online. You will also see a difference when comparing onstat -g iof and onstat -D. onstat -D shows logical pages where onstat -g iof show total disk I/O. http://www.ibm.com/software/data/informix/webcasts/#1 Thanks Kernoal On Jan 25, 2:41 pm, "Jack Parker" <jack.park...@verizon.net> wrote: > poking about. > > I have a table, in a single dbspace, in a single chunk. (Cooked). > > According to sysptprof, there have been 4.2 million page reads against the > table. > > According to syschkio, there have been 3.6 million pages read from the > chunk. > > Obviously this is a discrepancy, how is it getting resolved? A request is > made to disk to read a page, and the page has already been read? Then why > doesn't it count in bufreads? > > This particular table accounts for 88% of my i/o against this disk, so add > another 400K or so page reads to the 4.2 million number. > > Page writes seem to be more consistent with what I'd expect. > > cheeers > j.