3 Replies Latest reply on Jun 4, 2018 2:30 PM by vwolf

    Af Automatic Backfilling Mechanics


      Good morning,


      i have a question about the automatic backfilling function:

      in a scenario where i have an af server and a separate server with a ufl interface with buffering on. On af server there is a series of analysis  in cascade e.g.:  input A from ufl --> analysis --> result in tag B  which is the new input for ---> analysis2 ---> result in Tag C--->  and so on...  some of those analysis are event triggered, and some periodic. Now i activate the automatic backfilling on the various analysis for example.

      At this point I suppose to have a network communication problem between AF server and interface node for 1 day.... 


      when the communication returns up, the "old" data buffered by ufl respect the current time are simply old data and not out of order, so is the mechanism of auto-backfillling triggered anyway subsequently for all the analisys and in correct order ?

      And in that case is there a difference in the behaviour between the analysis which are event triggered based and the periodic ones?


      Thanks in advance

        • Re: Af Automatic Backfilling Mechanics

          Hello Paolo,


          If your interface is holding data on buffer queues, the event-triggered analysis would not write anything to the tags until data starts to flow again. Once the communication issue is resolved, the data starts to flow in the correct order and trigger the event-triggered analyses as expected, without the need of auto-backfilling.


          The periodic analysis, on the other hand, will continue to evaluate as if nothing was wrong and would probably repeat the same value for the period of the communication problem. The auto-backfilling mechanism is able to perform the recalculation of the "gap" on the periodic analysis. But then, in your case, this recalculation would trigger another recalculation and so on. Using event-triggered analysis means better performance.


          In short, if you do have many analysis that rely on tags and on each other, it is highly recommended that all of them are event-triggered and also output to a PI Point.





          1 of 1 people found this helpful
            • Re: Af Automatic Backfilling Mechanics

              Hi Victor,

              sorry for the late answer.

              Every analysis output to a pipoint, but unfortunately in our environment we can't use only event triggered analisys, and in addition to that in the past the use of subsequent event trigger analysis made some errors in calculations, but that was with previous versions of analysis service. Now we use the last one and we haven't tested this case yet. 


              Anyway if i understand well, enabling the auto-backfilling in advanced options on the periodic analysis will make them recalculate when new data appears in the input tags in the past, generating a new input for another subsequent analysis,  correct?. On the other hand obviously, the event triggered analysis already do this even without the autobackfill option because they work differently, and in this way i should have consistent data in the various output of the analysis chain. Right?