AIOps on IBM Z - Group home

IBM Z Anomaly Analytics v5.1.10 provides common componentry


As customer adoption of IBM Z Anomaly Analytics continues to grow, we are seeing more and more customers with unique environments driving more requirements for what IBM Z Anomaly Analytics can achieve. With these new requirements we have continued to invest in improving the installation, configuration and management of the product for our customers. We also have helped reduce IBM Z Anomaly Analytics (IZAA) overhead when installed with it’s “sister product” IBM Z Log and Data Analytics (IZLDA) through the use of common componentry when used together to detect and diagnose a problems.

When installing and configuring IZAA and IZLDA together, customer’s can significantly reduce the amount of work by leveraging common core pieces of infrastructure. Today, we support sharing of Kafka, Zookeeper, Keycloak and OpenSearch. This helps to achieve the following gains across ZAA and ZLDA:
•    Duplication of 4 services removed
•    Installation processes reduced from 3 to 1
•    Configuration processes reduced from 3 to 1
•    Configuration files reduced from 5 to 1
•    Number of administrative CLIs reduced from 3 to 1
•    Number of places in which to maintain user IDs reduced from 3 to 1
•    Number of keystores / truststores / certification reduced from 4 to 1
•    Other improvements:
          - Configuration settings are now automatically retained when a new release is installed
          - Problem Insights server data are now automatically retained when a new release is installed. 

In addition, we have invested in making the log-based machine learning model training more manageable for IT Admins. When it comes to model training management there are two key concepts:
     1. Training Periods: This is the number of consecutive calendar days to include in training models. By default, the configuration is set to 90 days, with a configuration range of 2 to 365 days.
     2. Training Intervals: This is the number of consecutive calendar days between automatic builds of the model. By default, the configuration is set to 30 days, with a configurable range of 30 to 365.

For illustration purposes, the following example uses non-default values for the training period and the training interval:
•    Training Period is set to 10 days
•    Training Interval is set to 7 days

The UI to manage the training dates help ensure that you don’t automatically train a new model with dates with known issues. 

Here you have the ability to remove or exclude dates that you want to ensure are not included in the new model. 

Other enhancements and fixes include:

Log-based Machine Learning:
•    Variable analysis enhancements and changes
•    Data cleanup
•    Train management improvements
•    Feedback dialog for NPS

Metric Based Machine Learning:
•    Scorecard UI responsiveness improvements
•    Validation changes
•    Bug fixes
•    Upgrading

ZAA Common Updates:
•    Dashboard extensibility and re-architecture
•    Uplift to Angular 15.1
•    Support of material icons using a browser running in an air-gapped environment
•    OCI image changes: deployment management and shared services

Learn more from our Product Doc here.