Skip to main content
N
nahuel14-Alexandrite
14-Alexandrite
June 18, 2021
Solved

Anomaly Detection Server Sizing

  • June 18, 2021
  • 1 reply
  • 2159 views

Hi community,

 

Given the following context:

There are 600 assets and each of them have more than 100 properties and each property need to have at least one Anomaly Detection alert configured.

 

1) How should the ThingWorx Analytics server (where the Anomaly Detection functionality will be running) be sized?

2) Is any guideline/rule/formula to size the ThingWorx Analytics server appropriately once the context changes?

 

PD: Also any rules or guidelines on how to size the ThingWorx Analytics server taking into account not only the Anomaly Detection feature but all ThingWorx Analytics server features will be much appreciated.

Best answer by cmorfin

@nahuel 

 

The hardware recommendation should mostly be similar.

Unfortunately there is no sizing guide for 9.0 so we do not have specific information, but  I would think that should be similar.

 

Thanks

Christophe

1 reply

C
cmorfin19-Tanzanite
19-Tanzanite
June 21, 2021

Hi @nahuel 

 

There is an Anomaly Detection sizing guide at https://www.ptc.com/support/-/media/65149E71E87E49D4AC96D00946FD681E.pdf?sc_lang=en

Note that once the anomaly detection model has been created by Analytics Server , the monitoring (scoring against this anomaly model ) is done by ThingWatcher engine that runs inside ThingWorx, not in ThingWorx Analytics Server. That is the reason why the above sizing guide discuss ThingWorx sizing.

 

I do not understand your second question about "when the context change", could you clarify this ?

 

Regarding the sizing of Analytics Server, all I could find is the minimum requirements in Help Center

 

Thanks

Christophe

 

N
nahuel14-AlexandriteAuthor
14-Alexandrite
June 28, 2021

Hi @cmorfin 

 

By the "when the context change" I meant how the server should be scaled up or down according to the increase or decrease of the number of anomaly alerts. But don't worry, I think that point has been covered by the guide you shared.

 

I read the guide and I have some follow up questions:
* Since the tests were done on ThingWorx 8.x, do the Hardware Recommendations apply to ThingWorx 9.x?
* On the Testing Assumptions section it says that the Training and Results microservices were run on a separate server. What was the server size? Was sized to the minimum requirements?
* On the Other Observations section it mentions there are some options that allow multiple Anomaly Alerts to be trained asynchronously. What are those options and how could I configure them?

 

Regarding the Analytics Server sizing, thanks for making that search. At this moment I don't need further details, but I'll create another post related to ThingWorx Anlytics if necessary.

 

Regards,
Nahuel

C
cmorfin19-Tanzanite
19-Tanzanite
June 30, 2021

Hi @nahuel 

 

Thank you for clarifying.

 

I do not have the information about the sizing of the Training/result microserver machines, however since only one training could be done at a time, i am assuming that it was a minimal requirement.

The statement in Other Consideration is likely no longer true. The Anomaly Detection now uses ThingWorx Analytics Server Training service, so it can do one training job per worker concurrently. By default there are 3 workers.

 

There is however no more details about the sizing here. I have informed R&D that this guide does need an update but I can't tell when/if that will happen.

 

Thanks

Christophe

 

Terms & ConditionsAccessibility statement