Hi,
ASE 16.0 PL1 running on Linux (SLES 11) as guest within a VMWare environment:
The non-voluntary context switches of the syb_default_pool are high. We tried to reduce 'idle timeout' and the number of threads (8->4) of the syb_default_pool. Both changes have no effect in phases of high server load including a lot of BCP IN action.
Typical sp_sysmon output during high server load:
Context Switches at OS per sec per xact count % of total
------------------------- ------------ ------------ ---------- ----------
ThreadPool : syb_default_pool
Voluntary 214.2 17.0 192794 2.8 %
Non-Voluntary 5626.6 446.4 5063951 72.2 %
Low server load:
Context Switches at OS per sec per xact count % of total
------------------------- ------------ ------------ ---------- ----------
ThreadPool : syb_default_pool
Voluntary 303.4 26.6 273082 59.0 %
Non-Voluntary 84.5 7.4 76009 16.4 %
What can be done to reduce the non-voluntary context switches?
Many thanks
Robert