China Mobile's Panji Platform: Observability Practices and Implementations for LLM Applications Base
Talkat KubeCon + CloudNativeCon China 2025 · on Wednesday, 2025-06-11· at 11:45· Level 16 | Grand Ballroom I·
summary
Describes how China Mobile’s Panji platform and DeepFlow use eBPF and Wasm plugins to provide zero-code, full-stack observability for large language model deployments. Explains collecting kernel-level real-time data with eBPF, parsing streaming requests with Wasm, and integrating those signals into service maps, distributed tracing, and multidimensional metrics to monitor and optimize LLM application performance.
AI-written summary · about
related talks
17meBPF in Microservices Observability
40mOpen Source LLMs in the Cloud: Scalable Solutions
28mNavigating LLM deployment: Tips, tricks, and techniques
36mPanel Disucssion: Is There Actually a Byte Behind All the Buzz? eBPF in Production!
11mCan LLMs Understand the Linux Kernel?
31meBPF for Observability: The Good, the Bad, and the Ugly
22mHarnessing Observability for 5G Performance: eBPF and OpenTelemetry I...
22mDWARFs and ELFs on the Cloud - eBPF Profiling Adventures in Observability at...