History log, what to think about
We have a 5.1 system with separate AS (3) and CS (2) and 10 AC450.
Today it is TTD about 800 trendsignals and Log Configuration for 1600 other signals.
This takes about 20gb of 100gb storage on the CS-servers.
They want to start trending all AI-signals and all PID-signals, total of 4500 signals every 3 seconds for 12 weeks. (they will get rid of the TTD in a few month)
So the system will consist of about 6000 trends.
What to think about:
Storage, is it a formula to calculate the nedd for storage?
Do the system have any problem logging 6000 signals?
Voted best answer
- An average AC450 controller can easily handle 1000 subscriptions on 3 seconds
- OPC logging at 1, 3 or 9 seconds uses less CPU in controller than TTD logging
- A dedicated (not combined with other server role) connectivty server can host 10.000 logs, possibly more if high end storage is used (fast 15K SAS disks in RAID5 or 6 + adequate amounts of RAID cache).
- Typically, a Basic History log will allocate 21 bytes per sample it stores + some overhead.
- The "hard" cap of 90 days storage time for Basic History has been removed. Logging several years typically works just fine with Basic History.
Keep an eye at the controller and RTA CPU load and channel saturation.
- Perform ANPER Task + Channel Utilization analyses BEFORE and AFTER implementing the new logging, both in controllers and RTAs
- Average RTA load shall not exceed 50%
- No Full Events shall occur (The "Min Free SQB" and "Avg Free SQB" figures will tell "average saturation" and "high water mark" during analyse period).
Keep controller (ANALYZE PC element) and RTA (RTA Load object) logged. The latter need configuration of RTA net & node after creation as child to any RTA.
From previous experience I am going to take the unusual step of disagreeing with Stefan on this answer
I think it is unwise to remove TTD logging from the AC450 controllers. While 1,3 and 9 second subscriptions are "more efficient" for mb300 comms, I have seen occasions where significant amounts of data was lost from OPC logs by overloading the subscriptions. TTD logging can also spread CPU loads by delaying logging of data slightly to timeswhen the CPU is idle
If your AC450 controllers are already heavily loaded, then it is very likely that you will loose data using OPC logging.
of course Im quite happy to be proved wrong on this one ;)