eBPF-based APM and Observability for Cloud-native Infrastructure
Primary Investigator:
Deepak Nadig
Bhavye Sharma
Abstract
Extended Berkeley Packet Filter (eBPF) is a powerful and versatile technology that allows developers and system administrators to trace, monitor, and debug the behavior of applications running on Linux-based systems. eBPF is particularly useful for cloud-native observability, as it enables the collection of fine-grained metrics and insights into the performance and behavior of microservices, containers, and other cloud-native infrastructure components.
We built an open-source eBPF-based observability pipeline that enables sophisticated troubleshooting and debugging workflows, allowing developers to identify and resolve issues in complex, dynamic environments quickly. Using this pipeline of labeled data, we propose an AI model that can classify anomalous behavior of our applications.