> > We have a client with 320GB database (running FB CS v2.5) > > Is it really so? FB does not support LockHashSlots more than 64K, it would > truncate your 90001 down to 65521.
<SL> Is the engine that smart to trim to an exact prime number? > > Now, today, I checked the lock print values again, and got even worse > > numbers!!! > > The change didn't help, the reason is elsewhere. 21% vs 23% could be > explained by e.g. printing the stats at 3PM and 5PM. <SL> The stats were printed at 3am. > And we don't know what was the mutex wait ratio before they reported the > performance issues. <SL> True, not something that came to mind for us to be monitoring. > 688 peak connections. Is this expected? Could the load be growing recently? <SL> Yes, that seems a reasonable value. <SL> Growth? Can't really speak to that, the client system is accessible by a large number of users, their work activity has day of the week and seasonal variation. <SL> Curious, how would the number of owners (even free owners) impact the lock manager? <SL> Today's numbers are much worse, again. But the activity load is substantially lower (based on the "Enqs" and "Acquires" values). LOCK_HEADER BLOCK Version: 145, Active owner: 0, Length: 67108864, Used: 5269256 Flags: 0x0001 Enqs: 202113295, Converts: 934229, Rejects: 96674, Blocks: 615455 Deadlock scans: 103, Deadlocks: 0, Scan interval: 10 Acquires: 326947224, Acquire blocks: 186433223, Spin count: 0 Mutex wait: 57.0% Hash slots: 90001, Hash lengths (min/avg/max): 0/ 0/ 4 Remove node: 0, Insert queue: 0, Insert prior: 0 Owners (47): forward: 732776, backward: 2105384 Free owners (54): forward: 3893360, backward: 4041872 Free locks (6339): forward: 734616, backward: 2979632 Free requests (31952): forward: 2283760, backward: 1299072 Lock Ordering: Enabled > > Any suggestions on how I can improve the numbers? <SL> Would a larger/smaller page cache size have any impact? Currently = 400 <SL> Could reducing the cached pages increase lock operations performance? Sean