IBM Z Observability Connect extends enterprise observability to the mainframe without requiring organizations to replace their existing observability tools. Although many observability platforms monitor distributed environments, they often lack visibility into application processing on the mainframe. IBM Z Observability Connect captures traces, metrics, and logs as applications run on z/OS and sends this operational data to designated endpoints. It exports spans using the OpenTelemetry Protocol (OTLP), enabling integration with any OTLP-compliant application management solution.
IBM Z Observability Connect runs in an OpenShift or Kubernetes cluster and supports Linux on IBM Z and x86 platforms. Its scalable architecture serves as an OpenTelemetry control plane, consolidating metrics, logs, and traces from multiple LPARs into a single distribution service. This service is ideally deployed on IFLs to take advantage of high-speed data movement within IBM Z. On the mainframe, the product collects native OpenTelemetry spans generated by IBM subsystems. Where native support is unavailable, CPU-efficient collectors create the required execution spans. In both cases, the spans are sent to the Telemetry Controller for correlation and distribution to designated analytics endpoints.
Total cost of ownership (TCO) is a critical consideration for management tools. TCO includes more than hard dollar costs, as the cost of data collection for mainframe tooling can be significant. IBM Z Observability Connect is designed to minimize the overhead of gathering application execution spans. Higher-cost metrics are disabled by default and can be enabled selectively, giving clients control over collection costs. Organizations can also selectively limit operational data sent to analytics platforms to reduce data volume. The product streams metrics and logs that add system context to application traces, helping accelerate problem determination. Because Z Observability Connect integrates with observability tools already deployed at customer sites, users need minimal additional training. The familiar user experience remains intact while gaining visibility into mainframe application processing.