Logs, Metrics, Traces…Assemble! Building a Context-Driven Mesh
Observability isn’t just about seeing, it’s about understanding. While dashboards and traces help, most systems today still operate in silos, with disconnected signals and fragmented context. And that’s a problem not just for debugging, but for security, performance, and cost-efficiency too. What if your infrastructure could explain itself? In this talk, we’ll explore how we built a telemetry mesh powered by Model Context Protocol (MCP) servers, each acting as a localized signal processor for logs, metrics, traces, profiling data, policies, and more. Using lightweight, serverless components and message buses, we federated these into a distributed context graph an always-on intelligence layer that powers smarter alerting, faster incident triage, and even security enforcement. No sidecars. No vendor lock-in. Just modular MCP nodes that collaborate like neurons turning raw signals into insight. We’ll wrap up with a live demo that shows how this mesh enables real-time correlation across layers, alerting based on context (not just thresholds), and a foundation that supports everything from SLOs to security posture audits.
Speaker
-
Prerit MunjalGrouponPrerit is working as the CTO of KubeCloud, an umbrella company with multiple products in the Cloud-Native Space, directing his expertise towards harnessing Cloud Native Technologies to design resilient architectures that can seamlessly scale, all while prioritizing technical cost, security, availability and end-user experience.