CPU Utilization issue if we start the Decipher services

[SriharshaVanama]

Hi Community Team,

We are facing a CPU Utilization issue if we start the Decipher services. CPU utilization goes to 100 % ↑ and the system is hanging.
From day 1 we installed, we are facing issues in slowness and we never checked the CPU utilization at that time. from last 2 or 3 days no application at all working on the server, so we checked the CPU utilization and it is fluctuating from 80 to 100.

 BP APP server and Decipher are installed on the same server (VM).

• Blue prism version: 6.10.1, Decipher Version: 1.2,
• Environment: Multi System,
• VM Configuration: Windows Server 2016,
• Processor: Intel Xeon silver 4110, RAM: 48 GB, HDD: 200 GB

 DataCaptureWorker is utlizing all the CPU, inspecting this process is located in "C:\Program Files (x86)\Blue Prism\Decipher Automated Clients\DataCaptureML\structured\DataCaptureWorker.exe"

► My finding is when the Decipher services are started CPU utilization is hiking and when I stop the services  (RabbitMQ, Decipher Licensing Service, Decipher Server, Decipher Web SDK Service, Decipher Automated Client Manager) it's coming to normal which is below 10.

☼ Kindly suggest a solution to solve this Issue 

​​

------------------------------
Sriharsha Vanamala
RPA Developer
------------------------------

[Ben.Lyons1]

Hi Sriharsha,

That process is the one creating the ML Capture model(s). What are your ML training settings?

Thanks

------------------------------
Ben Lyons
Product Consultant - Decipher Specialist
Blue Prism
UK
------------------------------

[SriharshaVanama]

Hi Ben,

Thank you for a response back @Ben.Lyons1

After training the Documents. ML is getting created 15 – 20 hrs later, It is getting reflecting in Document Type, Util it is showing N/A  

At this time the Cpu is hiking

The below image shows the ML Settings

Thank you 

​

------------------------------
Sriharsha Vanamala Rpa Developer
------------------------------

[Ben.Lyons1]

Hi Sriharsha,

So what's happening is every 50 documents Decipher's trying to create a new model/update the existing model. If you push multiples of 50 documents through in a day, you will experience the scenario you're getting. This is one of the reasons we generally recommend setting the training count to much higher volumes.

The amount of time spent training will be a direct link of your CPU's capabilities and the complexity of the required model. The complexity will be impacted by document count, field count, field definitions and the presence of an existing model.

Again to improve, you can either consider increasing your CPU size or having multiple capture servers. You can install the capture client on multiple servers, and then only 1 would get tied up working on the ML model creation. If you need more info on these specific topics, I would recommend raising an Expert Connect request, where we will be able to assist you more closely.

Please refer to our best practice guidance on how best to utilise the ML capture model capabilities.

Thanks

------------------------------
Ben Lyons
Product Consultant - Decipher Specialist
Blue Prism
UK
------------------------------