AI/ML Workspace

Introduction

Welcome to the AI/ML Workspace, where you can navigate your journey graphs with ease. The Workspace offers a structured approach, providing a clear separation between different components. This segmentation ensures a customized RCA experience tailored to your specific needs. Within the Workspace, you can define the role of each signal, encapsulate multiple signals, and explore their relationships and dependencies. These features empower you to gain deeper insights and effectively analyze the factors influencing your system's performance.

Overview

Our system architecture revolves around key components working harmoniously to provide a robust root cause analysis (RCA) solution. It consists of 4 main layers which are loosely connected to complete the setup

The Data Model forms the core, facilitating the identification of essential metrics and signals crucial for RCA. Built on our Data Store, these models enable users to extract critical insights into system performance. Configured within the AI/ML Workspace, users define specific journeys and analyze associated data, leveraging insights from Data Models to explore system dependencies and relationships.

As the RCA algorithm processes data and identifies potential issues, it generates Incidents accessible through the RCA Incident module. This seamless integration ensures users have timely access to actionable insights, fostering informed decision-making and proactive issue resolution.

Prerequisites

The user must have configured Data Models and Data Store on vuSmartMaps.

Why This Feature Is Useful

• AI/ML Modelling Workspace provides a centralized environment for configuring machine learning-driven analysis, anomaly detection, root cause analysis (RCA), and alert correlation workflows. It helps organizations identify abnormal behavior, detect incidents proactively, correlate events across systems, and uncover probable root causes using automated machine learning techniques.
• By organizing signals, components, journeys, graphs, and correlation models into dedicated workspaces, teams can gain operational insights, reduce alert noise, accelerate troubleshooting, and improve service reliability across complex business and application environments.

Example Scenario

A financial services organization wants to proactively identify application issues before they impact customer transactions. The operations team creates an RCA workspace using application, infrastructure, and database metrics from the configured Data Models. They define the business journey, map service dependencies, configure signalizers, and activate the workspace.

As live data is analyzed, the AI/ML models detect abnormal behavior, correlate related events, and identify the probable root cause of a database latency issue. The generated RCA Storyboard and incidents enable the operations team to quickly investigate and resolve the issue, reducing service downtime.

When to Use This Feature

Use AI/ML Modelling Workspace when you need to:

• Detect anomalies across infrastructure, applications, and business services.
• Perform automated Root Cause Analysis (RCA).
• Correlate related alerts and events across multiple systems.
• Analyze business journeys and service dependencies.
• Identify lead indicators that directly impact business operations.
• Create machine learning models for incident prediction and analysis.
• Reduce alert fatigue using alert correlation techniques.
• Build topology-aware service dependency models.
• Monitor operational and external indicators alongside business metrics.
• Visualize anomaly trends, forecasts, and signal health.
• Improve incident response and operational efficiency through AI-assisted analysis.
• Generate actionable insights using RCA Storyboards and Alert Storyboards.

Comprehensive Understanding

The AI/ML Modelling Workspace page provides a centralized interface to create, configure, and manage AI/ML workspaces used for Root Cause Analysis (RCA), Time Series Analysis, 3T Alert Correlation, and ML Alert Correlation.

The page allows users to:

• View all configured workspaces and their current status.
• Create, edit, activate, deactivate, or delete workspaces.
• Configure journeys, components, graphs, signalizers, bot settings, and storyboards.
• Monitor AI/ML processing status and access generated incidents for further analysis.

Step-by-step instructions

Follow these steps to create and configure an AI/ML Modelling Workspace:

1. Navigate to Observability Studios → AI/ML Modelling Workspace.
2. Click Add Workspace.
3. Enter workspace details including Name, Description, Category, and Run Type.
4. Create the workspace.
5. Configure the business Journey and Signals.
6. Add Components and assign metrics.
7. Configure Graphs to define service topology and dependencies.
8. Submit the Schema configuration.
9. Review the generated Signalizers.
10. Configure signalizer hyperparameters if required.
11. Activate signalizers globally or individually.
12. Configure Bot Settings based on the selected workspace type.
13. Activate the AI/ML processing pipelines.
14. Review Storyboards and generated insights.
15. Validate workspace health and signal status.
16. Complete the configuration and activate the workspace.
17. Monitor incidents, anomalies, forecasts, or alert correlations generated by the workspace.

Workspaces

Workspaces read the information provided by the Data Model and map them as part of an application or business journey.

We have 4 main Workspaces.

1. RCA
2. Time Series Analysis
3. 3T Alert Correlation
4. ML Alert Correlation

Ready to dive in? Just make sure your Data Models and Data Store are set up in vuSmartMaps—it's the foundation for a smooth AI/ML Workspace experience!

What happens after the step

After a workspace is successfully configured and activated:

• AI/ML processing pipelines are automatically created.
• Signalizers begin monitoring configured metrics.
• Anomaly detection models start analyzing incoming data.
• RCA models evaluate abnormal conditions and identify probable root causes.
• Alert correlation engines group related alerts and events.
• Storyboards begin displaying health, anomaly, and correlation insights.
• Forecasting models generate baseline and trend predictions where applicable.
• Incidents are automatically generated when configured thresholds or anomaly conditions are met.
• RCA Incidents become available for investigation and analysis.
• Alert reduction and noise suppression mechanisms become active based on workspace configuration.

Tips and Best Practices

• Ensure the required Data Models and Data Store are configured before creating a workspace.
• Categorize Lead, Operational, and External Indicators accurately to improve analysis results.
• Build clear component relationships and topology graphs to enhance root cause identification.
• Activate only the required signalizers to optimize resource utilization.
• Review generated Storyboards regularly to validate model behavior and operational health.
• Periodically retrain AI/ML models as application behavior and workloads evolve.

Troubleshooting

1. Issue: Unable to create a workspace.
   • Possible Cause: Required workspace details are missing.
   • Solution: Verify that the Workspace Name, Category, and Run Type are specified before creating the workspace.
2. Issue: Data Models are not available while configuring the Schema.
   • Possible Cause: Required Data Models or Data Store are not configured.
   • Solution: Verify that the required Data Models and Data Store are configured before creating the workspace.
3. Issue: Unable to add Signals.
   • Possible Cause: Invalid Data Model or unsupported metric selection.
   • Solution: Ensure that the selected Data Model contains valid metric columns and select supported metrics.
4. Issue: Components are not available while configuring Graphs.
   • Possible Cause: Components have not been created or saved.
   • Solution: Create and save the required components before configuring graph connections.
5. Issue: Signalizers remain inactive after activation.
   • Possible Cause: Signalizer activation or AI/ML pipeline creation was unsuccessful.
   • Solution: Activate the signalizers again and verify that the required AI/ML pipelines are created successfully.
6. Issue: No anomalies or incidents are generated.
   • Possible Cause: Insufficient historical data or Lead Indicators are not configured correctly.
   • Solution: Verify data availability, review the Lead Indicator configuration, and allow sufficient time for the AI/ML models to train.
7. Issue: Storyboards do not display any data.
   • Possible Cause: The workspace is inactive or AI/ML pipelines are not processing data.
   • Solution: Verify that the workspace is active, signalizers are running, and data is being ingested successfully.
8. Issue: Alert correlation or RCA results are not generated.
   • Possible Cause: Journey, Graph, or correlation configuration is incomplete.
   • Solution: Review the configured Journey, Components, Graph topology, and correlation settings before activating the workspace.
9. Issue: Hyperparameter changes are not applied.
   • Possible Cause: Updated hyperparameters were not saved or signalizers were not reactivated.
   • Solution: Save the updated hyperparameters, reactivate the affected signalizers, and verify that the changes have taken effect.
10. Issue: Workspace activation fails.
    • Possible Cause: Schema configuration is incomplete or required components are missing.
    • Solution: Verify that the Journey, Components, Graphs, Signalizers, and Bot Settings are configured correctly before activating the workspace.
11. Issue: RCA incidents are not visible.
    • Possible Cause: No incidents have been generated or the workspace is inactive.
    • Solution: Verify that the workspace is active, AI/ML processing is running, and abnormal conditions exist to generate incidents.
12. Issue: Forecasts or anomaly trends are not displayed.
    • Possible Cause: Insufficient historical data or forecasting models have not completed training.
    • Solution: Allow sufficient historical data to be collected and verify that the forecasting models are active and processing data.

FAQs

What is the AI/ML Workspace, and why should I use it?

The AI/ML Workspace is a dedicated environment within vuSmartMaps that helps you visualize and analyze the relationships between signals, events, and system behaviors. This workspace works in tandem with pre-configured Data Models and the RCA algorithm to provide incident-driven insights through the RCA Incident Console.

What do I need before using the AI/ML Workspace?

Before using the AI/ML Workspace, ensure you have set up the following:

• A Data Store (e.g., PostgreSQL, MySQL, Elasticsearch) where your source data resides
• One or more Data Models that define the signals and metrics relevant to your system or application
• Proper user permissions to access the RCA module and configure workspaces