Alert grouping

Alert grouping is the process of organizing and consolidating related alerts into sets based on common characteristics or criteria. This helps in simplifying alert management by reducing noise, making it easier to prioritize, track, and address issues efficiently. Grouped alerts provide a clearer overview of related incidents, facilitating quicker root cause analysis and remediation.

Approaches to alert grouping

There are several approaches available for alert grouping. Some methods rely on user-defined logic, such as Manual, Rule-based, or Tag-cluster, while others use advanced algorithms that can be fine-tuned, including Automatic, CMDB, Text-based, and Log Analytics.

Table 1. Alert grouping types and use cases

Type	Description	Use case
Log Analytics Grouping	Alerts are grouped based on the analysis of log data. This involves correlating log entries to identify related incidents and issues. By leveraging log patterns and sequences, this method can detect complex, multi-step problems across the IT environment.	An online gaming company enhances server stability by implementing proactive log analytics. They monitor logs from game servers in real-time and use analysis tools to detect patterns of errors that occur before crashes. For instance, the analysis reveals that certain error patterns appear about 30 minutes prior to server crashes. By setting up automated alerts for these patterns, the company can initiate remediation actions, such as restarting services or reallocating resources, before a crash occurs. This proactive approach prevents disruptions, minimizes downtime, and improves the gaming experience by addressing issues before they impact players.
Rule-based Grouping	Alerts are grouped according to predefined rules and criteria set by users. These rules might include specific conditions, such as thresholds or event types. This method is effective for consistent and repeatable patterns but requires maintenance of the rules.	In a data center managing an e-commerce website, rule-based alert grouping helps handle high traffic during events like flash sales. Alerts about server issues, such as high CPU usage, are designated as parent alerts. These parent alerts are linked to child alerts that report related problems, like slow database queries. The rules ensure that server-related alerts are grouped with their symptoms, allowing the IT team to quickly identify and address server overload issues. This approach improves issue resolution efficiency and minimizes downtime.
Automated Grouping	Advanced algorithms automatically identify and group related alerts based on patterns and similarities in the alert data. This method leverages machine learning and AI to adapt to new and unknown issues, providing proactive alert management. Event Management groups alerts that are similar, but not necessarily identical, based on the proximity in time of the last event generation. Alerts with the same CI and the same pattern identifier are grouped together. Automatic alert grouping consists of the following components. Alert Aggregation Learner (Service Analytics Alert Aggregation Learner - Daily): This offline job runs daily to process past alerts and perform statistical analysis to build alert patterns. For details, see Configure pattern based alert aggregation. Real-time Alert Aggregation job (Service Analytics group alerts using RCA/Alert Aggregation): This job runs every minute to generate alert aggregation groups based on alert patterns, CMDB relationships, text similarity, user-defined alert clustering tags, and network traffic connection between processes.	A large financial institution uses machine learning to manage alerts from numerous servers and applications. The system analyzes historical alert data to recognize patterns, such as database server failures frequently being accompanied by client connection errors. It then automatically groups related alerts together. For instance, when a new database server failure alert is detected, it is grouped with previous connection error alerts. This automated grouping helps the IT and security teams quickly identify and address issues, improving response times and reducing downtime.
CMDB-based Grouping	Alerts are grouped based on Configuration Item (CI) relationships and dependencies from the Configuration Management Database (CMDB). This approach ensures that alerts related to specific infrastructure components or services are grouped together, providing context-aware alert management.	A telecommunications company uses CMDB data to manage alerts related to their network infrastructure. Alerts related to a specific network router and its connected devices are grouped together based on their CMDB relationships, enabling the network team to see all related issues and address the root cause efficiently.
Text-based Grouping	Alerts are grouped by analyzing the text content of alerts to identify similarities and related issues. Natural language processing (NLP) techniques are used to find commonalities in alert description, metric name, and ci class, making this method effective for unstructured data.	In an organization that uses Zoom rooms for virtual meetings, the IT team receives numerous alerts when the Zoom room server experiences an outage. Each alert might indicate a different Zoom room being down, such as Zoom room no 10 is down, Zoom room no 11 is down, and so on, with the only difference being the room number. For organizations with a CMDB, these alerts can be grouped using CMDB relations, as the system can correlate the alerts based on the server's impact on all associated Zoom rooms. However, for organizations without a CMDB, text-based grouping can be used. The system applies natural language processing to group alerts with similar descriptions, helping the IT team quickly identify that multiple Zoom rooms are affected by the same underlying server issue. This approach allows the IT team to efficiently address the root cause of the problem, reducing downtime and improving response times.
Tag Cluster Grouping	Alerts are categorized and grouped using tags or labels that represent common attributes, such as application, server type, or geographic location. This method allows for flexible and dynamic grouping based on evolving tagging strategies.	An organization without a CMDB manages a Linux server running various services. The IT team uses a Node field in each alert to identify the server, and they group all events related to services on the same server based on this node value. For example, they cluster alerts like Service A down and Service B high CPU usage together if they share the same node value. This approach helps the IT team address server-related issues more efficiently. By clustering alerts for the same node, application, or IP address, the team streamlines their response efforts and resolves issues more effectively, even without a CMDB.
Manual Grouping	Users manually select and group related alerts based on their expertise and understanding of the system. This approach allows for precise control but can be time-consuming and may miss automated correlations.	A system administrator receives multiple alerts about different services failing on a single server. The admin manually groups these alerts, recognizing that they are all related to a single hardware failure on that server, and prioritizes fixing the hardware issue to restore all services.

Manual and rule-based alert grouping differs from algorithm-based grouping mainly in how the parent alert is chosen. In manual, rule-based, or log analytics grouping, one of the real alerts is designated as the parent alert. In Automatic, CMDB, Text-based, and Tag Cluster modes, a virtual alert, representing the oldest and most severe alert in the group, is created as the parent alert.

Note:

In domain-separated environments, alert groups are created only for alerts within the same domain.

For information on scheduled jobs and parameters, refer to Scheduled jobs and parameters for alert grouping. For detailed information on different grouping types, see Alert grouping types.

Benefits of alert grouping

• Creating automated alert groups by aggregating alerts based on predefined patterns.
• Correlating alerts using timestamps and CI identification to form automated alert groups.
• Forming CMDB based alert grouping by correlating alerts based on CI relationships in the CMDB.
• Correlating alerts based on text similarity of alerts using NLP (Natural Language Processing).

Related tasks

• Synchronizing alert response with automated alert grouping