Reducing the Mbuild or model build process execution runtime in DSP Machine Learning

Products

• Mainframe Operational Intelligence

Issue/Introduction

If you find that the Mbuild or model build machine learning process is running longer than expected,  the steps included in the Resolution section below to change the thread count from the default value of 1 to a new value of 4 need to be performed.    This will decrease the elapsed runtime of the Mbuild process and allow other utility functions such as the Postgres vacuum cleanup process to run and complete successfully. 

Environment

Mainframe Operational Intelligence

Release  2.0.*

Resolution

To reduce the elapsed runtime of the Mbuild Machine Learning process, please apply the below changes to your MOI appliance: 

1. Change to directory
 cd /opt/moi/

2. Edit the docker-compose.yml

 vi dsp/normal/docker-compose.yml 

3. Change the below property to use 4 threads 

 KDE_THREAD_COUNT=1
 to 
 KDE_THREAD_COUNT=4

4. Restart the DSP microservice
 

bin/plex down -- dsp
bin/plex up -d -- dsp

5.  The next scheduled run of the Mbuild process should have a shorter elapsed runtime. 

6.   To check the Mbuild process history, please refer to  Knowledge Document at the following link:  https://ca-broadcomcsm.wolkenservicedesk.com/wolken/esd/knowledgebase_search?articleId=230372