WinCC OA is suitable for very large applications. I've seen distributed applications where each server (pair) was handling many 10 thousands of dp elements.
But how much is 'ok' for a single server (pair). What is your experience ? What about:
* 50.000 alarms
* 100.000 alarms
* 200.000 alarms
Does anyone have good or bad experiences with such big numbers.
Thanks for the info
Any help is appreciated
Share the fun
Frenk Mulder
How big is big ?
Search
Re: How big is big ?
Hello Frenk!
The size of the data model has an effect on the memory used to store the identification in every manager after the initialization.
It has also has an impact on the disk space used to save the data model (DP types, datapoints, configuration).
The number of pending alerts has an impact on the memory usage of the Event Manager as the information must be stored during runtime in the process image managed by the Event Manager.
When talking about the amount of data which can be handled by a server the data traffic is of interest.
Therefore, you must differ exactly between the size of the data model and the amount of value changes / second and alert changes / second.
Best Regards
Leopold Knipp
Senior Support Specialist
The size of the data model has an effect on the memory used to store the identification in every manager after the initialization.
It has also has an impact on the disk space used to save the data model (DP types, datapoints, configuration).
The number of pending alerts has an impact on the memory usage of the Event Manager as the information must be stored during runtime in the process image managed by the Event Manager.
When talking about the amount of data which can be handled by a server the data traffic is of interest.
Therefore, you must differ exactly between the size of the data model and the amount of value changes / second and alert changes / second.
Best Regards
Leopold Knipp
Senior Support Specialist
Re: How big is big ?
I understand your statement. Our projects are all about traffic where we get integer 'State' data from barriers, doors, traffic lights, smoke detectors. Also some analog values from air quality, temperate and such. Normally the number of changes per second is relatively low. Doors don't open much, barriers don't move much.
Lets rephrase the question.
Lets assume that we get 1000 changes per second (mostly analogue values). Alarms should not occur much so can be neglected for now.
Does anyone have reference figures for large applications with:
* 50.000 alarms ?
* 100.000 alarms ?
* 200.000 alarms ?
Assuming the changes per second are low, is there any limitation to the number of datapoint types, datapoints, alarms, adress configs other than the memory consumption and the fact that startup takes longer ?
Thanks
Frenk Mulder
Share the fun
Lets rephrase the question.
Lets assume that we get 1000 changes per second (mostly analogue values). Alarms should not occur much so can be neglected for now.
Does anyone have reference figures for large applications with:
* 50.000 alarms ?
* 100.000 alarms ?
* 200.000 alarms ?
Assuming the changes per second are low, is there any limitation to the number of datapoint types, datapoints, alarms, adress configs other than the memory consumption and the fact that startup takes longer ?
Thanks
Frenk Mulder
Share the fun
Re: How big is big ?
The limitations concerning the size of the data model you can find in the WinCC OA Documentation:
https://www.winccoa.com/documentation/W ... tions.html
Defining a project with 200.000 _alert_hdl configs is possible.
As already stated the number of pending alerts and alert changes is more interesting than just the number of active _alert_hdl configs.
Best Regards
Leopold Knipp
Senior Support Specialist
https://www.winccoa.com/documentation/W ... tions.html
Defining a project with 200.000 _alert_hdl configs is possible.
As already stated the number of pending alerts and alert changes is more interesting than just the number of active _alert_hdl configs.
Best Regards
Leopold Knipp
Senior Support Specialist
Re: How big is big ?
The only way to find out if this will work is to carry out an NFR test using the exact software / hardware configuration.
For example, we have projects that only work on Linux because Windows isn't fast enough.
For example, we have projects that only work on Linux because Windows isn't fast enough.