PatrickHiggins607207

impact of dynamically setting ERT

Discussion created by PatrickHiggins607207 on Nov 6, 2017
Latest reply on Nov 6, 2017 by Pete Wirfs
I have been looking through the various discussions on how ERT and MRT can be set.
We are running on 11.2.

The doc states that the default values for
     ERT_METHOD in UC_CLIENT_SETTINGS = 'linear regression' (LR)
     ERT_CALCULATION in UC_CLIENT_SETTINGS = 'IMMEDIATE'
  
We do NOT have any of the ERT entries set in any UC_CLIENT_SETTINGS, so we are going to use the default, unless some other setting is used in the individual objects. 

Another entry in either the doc or in a community discussion indicated that if the ERT option effective for any object is NOT a fixed value, the ERT is recalculated as part of the task completion process for every task.

In looking at just our JOBS object, about 95% are using defaulting ERTMethod (implies LR)  and 5% using fixed or average.

Three questions
1) Is the above accurate in how the estimated run time is determined?
2) How does the CA/Automic/UC4 LR actually work - the complete formula/calculation process?
3) What is the impact on the engine if 95% of the tasks are recalculating the ERT based on LR with every run?
4) Are we causing performance 'opportunities'  by doing this?

When I was young, I worked with LR and optimization and this could pin a Cyber mainframe for hours(95% CPU for 2-3 hours) for a single solution.  So depending on exactly what is happening inside the AE, LR calculations can be very costly.

Outcomes