Re: onstat -g sch , Busy Waits ?

Jay Aymond wrote: > rdavis@rmi.net wrote: > >In article <78kcue$kng$1@nnrp1.dejanews.com>, > > darren_m_rogers@my-dejanews.com wrote: > >> Hi, > >> Could anyone explain what may cause a large number of 'busy waits' ? > >> Informix Dynamic Server Version 7.30.UC5 (HPUX-11.00) > >> VP Scheduler Statistics: > >> vp pid class semops busy waits spins/wait > >> 1 1495 cpu 29 31 967 > >> 2 1508 adm 0 0 0 > >> 3 1509 cpu 280334 280893 999 > >> 4 1510 lio 2 0 0 > >> 5 1515 pio 2 0 0 > >> 6 1517 aio 19 0 0 > >> 7 1522 msc 1287 0 0 > >> 8 1524 aio 14 0 0 [SNIP] > The CPU that is running the NET thread for shared memory will usually have very > few busy waits since it is responsible for the polling process. The other CPU's > always have some idle time and therefore have busy waits. OOO I missed that. If you are running your TCP listeners in CPU VPs, and that seems likely, CPU VPs MUST poll to listen since they cannot block like NET VPs can. Shift to NET VPs in the NETTYPE for tcp/ip. BTW similarly NET VPs cannot block on shared memory listeners and have nothing else to do so they will spin furiously if you move our ipcshm NETTYPE entry from CPU to NET leave ipcshm as CPU. Art S. Kagel