Does it mean that backup (either full or tran) may potentially consume fairly large amount of proc cache memory (with the bunch of servers I inspect it may range anywhere between 10 to 90 % of it)?
The one server that has had this huge percentage of PC consumed by backups also reports 20+% proc_cache spinlock contention - from time to time. Is this related to the pressure on the PC memory caused by concurrent backup (actually, the only concurrent dump at that time was the tran log dump which takes place every 5 minutes)?
Which brings me back to inability to use this MDA conclusively - at least in 15.0.3. The counters in it increase all the time - so if at some point of time some activity pushed one of the modules into using 90% out of all available PC memory it will stay there until the server is bounced (for both Active and HWM counters - I would expect the HWM value being pushed up while the Active value being brought back to real non-cumulative value). All the subsequent attempts to see which module is most active and potentially has greater influence on the overall behavior of ASE at the time of inspection will be unsuccessful. Measuring delta values will also make no sense. Is there no way to reset the counters without bouncing the server? How can I make sense of the values the table lists?
