Data with Added Value: How Raw Monitoring Signals Become Reliable SLA Reports

Monitoring data often has a short half-life: An alert pops up, the issue is resolved, and the alert disappears. However, for a managed hosting provider or a critical infrastructure operator, these data hold much more potential. They are the objective proof of the performance delivered.

The challenge lies in processing the enormous amounts of metrics produced by global endpoint monitoring every second in a way that is understandable to both technicians and customers. The solution is seamless integration into the existing observability stack using Prometheus and Grafana.

The Problem: Data Silos and Manual Reports

Without central integration, two separate worlds often emerge within a company:

1. The Technicians’ World: They use specialized tools, view live graphs, but lack a historical overview spanning months.
2. The Business World: Account managers must painstakingly gather data from various sources for monthly service reviews, transfer it to Excel sheets, and manually calculate availability. This is error-prone and appears unprofessional.

The Solution: Metrics Export and Visual Preparation

Instead of operating endpoint monitoring as an isolated island, all results—from response time in milliseconds to TLS status—flow directly into the central time-series database (e.g., Prometheus or VictoriaMetrics).

1. Prometheus as Central Storage (Single Source of Truth)

Every check of the global PoPs is exported as a Prometheus metric. This has significant advantages:

• Long-term Archiving: We can analyze availability not just for today, but for the entire past year.
• Correlation: We can directly compare external response time with internal metrics (e.g., CPU load of the web server) in one chart.
• Standard Queries: With PromQL (Prometheus Query Language), complex questions can be answered, such as: “What was the average availability of all API endpoints for customer X in the last quarter?”

2. Grafana for Dashboarding

Grafana is the window to the data. Here, we create different views for various target audiences:

• The Operations Dashboard: Focus on real-time data, latency spikes, and TLS warnings for the on-call team.
• The Management Dashboard: High-level view of all customer SLAs with a “traffic light system” (Green/Yellow/Red).
• The Customer Dashboard: A filtered view that transparently shows the customer that their leased infrastructure meets the agreed targets.

3. Automated SLA Reports

The greatest operational leverage is the automation of reporting. Since the data is structured, reports can be generated at the push of a button or on a scheduled basis:

• Availability Percentage: Calculated based on actual uptime (e.g., 99.95%).
• Performance Trends: Graphical representation of whether the application has slowed down over the month.
• Incident History: Listing of all verified outages including duration and affected regions.

Conclusion: Transparency Builds Trust

By freeing monitoring data from their silos and transforming them into professional dashboards and reports, technology becomes tangible for all involved. For the customer, it is the reassuring feeling that the promised quality is measurably maintained. For the provider, it is the efficient way to demonstrate professionalism without additional manual effort. Monitoring is ultimately not just a technical warning system but a central tool for customer retention.

FAQ

Can we give the customer access to our Grafana? Yes, Grafana supports multi-tenancy. Customer accounts can be configured to see only the data of their own endpoints. This is a massive vote of confidence in one’s own service.

How do we handle maintenance windows in SLA reports? In Prometheus, maintenance times can be marked or excluded from calculations via specific metrics. This way, availability in the report is not distorted by planned work.

Is Prometheus suitable for long-term storage of SLA data? Prometheus itself is optimized for short- to medium-term data. For true SLA histories over years, connecting to a long-term storage like VictoriaMetrics or Thanos is recommended.

Can we also track error rates (Error Budgets)? Absolutely. In line with Google’s SRE principles, “Error Budgets” can be defined. The dashboard then shows not only if there is currently an issue but also how much “downtime” is left in the month before the SLA is violated.