Real-time monitoring of application integration processes
AI & DATA
Real-time monitoring of application integration processes
event13/07/2026
THE CHALLENGE
Dozens of specialized applications—admissions, laboratory, radiology, report repositories, and departmental systems—continuously communicate through an integration bus. These are invisible flows: their existence becomes apparent only when they stop working.
When disruptions occurred, entire departments experienced slowdowns or came to a halt. However, the greatest cost was the lack of diagnostic visibility: without dedicated monitoring, the IT team spent hours simply identifying which integration had failed and where, followed by additional time determining who was responsible—whether an internal team, the ESB vendor, or the supplier of the specific application—in an endless cycle of responsibility handoffs. The result was slower issue resolution, risks to service continuity, and strained relationships with suppliers.
THE SOLUTION
Extra Red implemented the Integrations Realtime Monitoring service: a cross-domain, flexible, and modular system that does not interfere with the communication channel in use and is limited to observing, correlating, and providing meaning to the data already in transit. By using the community version of the open-source stack, it was possible to bring the cost down to zero, allowing investment to focus primarily on developing the solution itself.
- On-premises ELK Stack (Elasticsearch, Logstash, Kibana): installed, configured, and supported by Extra Red on hardware located within the organization.
- Three monitoring levels: Basic (process health), Advanced (placer-to-filler transport times and errors), and Semantic (message content analysis).
- Placer/Filler-based logging model with a unique transaction identifier: the flow of each integration can be reconstructed end-to-end—a model aligned with healthcare integration standards.
- Custom extraction connector designed to connect with the organization's integration bus without invasive modifications.
- Kibana dashboards: an overview of all integrations, sender-to-recipient Sankey diagrams, KPIs on messages, latency, and errors, with drill-down capabilities down to the individual message level.
- Proactive alerting based on historical data: thresholds are determined by the historical behavior of each integration—a trend falling outside expected ranges triggers an alert before the anomaly becomes critical. Multiple notification channels are supported, including email, SMS, and enterprise messaging platforms.
KPIs
Metrics are currently being collected during the post go-live period and will be consolidated over time.
| Metric | Result | Driver |
|---|---|---|
| Time required to identify anomalies | -70% | Alerts are triggered before issues occur thanks to trend analysis |
| MTTR | -55% | The ticketing system automatically escalates incidents to the relevant operator or supplier |
| Errors resulting in service interruptions or slowdowns | -80% | Historical-data-based thresholds enable proactive intervention |
| Uptime level | 99.95% | Service continuity is ensured without impacting operational flows |
| Multi-level visibility of integration flows | 100% | Today, every organizational level—from governance to operations—can monitor the areas relevant to their responsibilities |
THE EVOLUTION: FROM STATISTICAL TO INTELLIGENT
- Predictive AI: a model that continuously learns from data and correlates multiple factors simultaneously to anticipate anomalies with increasing accuracy, reducing false positives and widening the prevention window.
- Agentic ticketing automation: ticket creation messages already include a proposed resolution, generated from an ever-evolving knowledge base fed by historical data and previously resolved and documented tickets. Every resolved incident makes the system more effective in handling the next one.
other experiences on the subject
