Issue with OPC UA .Net implementation for running Aggregate|OPC UA Implementation: Stacks, Tools, and Samples|Forum|OPC Foundation

Saurabh,

  Yes all of the algorithms listed in the specification can be calculated without storing all of the raw data.  They do require storing of some intermediate data or depending on the calculation storing the previous value.  It does require a little effort to setup each calculation to run without storing all of the values, since what the intermediate storage items are will vary based on what is being computed.  For example:  a Min or Max only requires that the current value is compare to the latest Max or Min and if it is larger/smaller the value is updated.  The TimeAverage requires the previous value (and time stamp) be stored and sum for the average can be calculated as a new value is received.  When the interval end is detected then the average can be calculated from the sum  and the time interval.

It is also import to note that a separate calculation is required to compute the Quality/Status associated with the aggregate, but at least there are fewer variation of this calculation.

If there are any specific Aggregates that you don't know how to calculated in a one point at a time manner, just ask about them.

Paul 

Saurabh,

  Are you having a problem with a specific aggregate and how it would be implemented in a one point at a time manner? If so let post which aggregate? And I can help you with the algorithm you would need to implement.

 If you are looking for someone to provide sample code for how to calculate aggregates in general then I would post your request to GitHub.

Paul

Saurabh,

So if you want to calculate an interpolated value at some interval - that is very easy - As you get values in, save the last value only, check each new value and see if the interval boundary has been crossed(or landed on), if so take the previous saved value and compute the value at the interval boundary.  The reason I was asking about an aggregate - is that the algorithm for how to calculate it on a point by point basis varies.  In a previous post I described how to calculate a Min and Max which is a simple compare as the values are reported. There is no short cut, you have to look at each aggregate you are going to support and look at the calculation for it.  Some will require a summation and a count, some are compare operations and a saved value or an increment of a counter or timer. If you can't figure out what the math is for a point by point for an aggregate I can help you.  Once you have each of them figured out - you will find that some of them overlap and you can reuse parts.

If you are asking about bounding then again it depends on the type of bounding and the aggregate.  The key point is that if you know you are going to calculate aggregates you need to keep the last value around (but depending on the aggregate it might be two values if the last value is bad status  - since you need to keep the last good status value also for some aggregates).

To calculate status information - you also need to track number of good and bad values and the total time duration of good and bad, but again this varies from aggregate to aggregate.  Make a list of the aggregate that need to be supported and based on the list you can generate the items you need to track.

The key point to remember is that aggregate are available for subscription and for history.  In history you have a big block of data that you process, which is what the examples provide.  In subscription you get the values in one at a time.  The history block code might not work for one point at a time (depending on implementations and algorithms), it would be different code (ask GitHub if you need someone to write this).   The Subscription code is typically just filter code - i.e. I get values in and need to spit something out into the subscription, but I only do then once I cross a boundary depending on the algorithm.

Paul