YoVDO

xprobes: Hybrid User/Kernel eBPF Probes for Cross-Layer Observability

Offered By: Linux Plumbers Conference via YouTube

Tags

eBPF Courses Observability Courses CPU Profiling Courses Distributed Tracing Courses Flamegraphs Courses

Course Description

Overview

Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only!
Explore the concept of xprobes, a hybrid approach combining uprobes and kprobes for enhanced performance diagnostics in production environments. Delve into the limitations of existing tools like flamegraphs, distributed tracing, and BCC probes for identifying root causes of performance issues. Learn how xprobes provide cross-layer observability by maintaining aggregate distributions for functions of interest, updated by both userspace and kernelspace events. Examine the cpu-latency xprobe as an example, which measures on-cpu time of userspace functions while accounting for context switching. Discover how xprobes can offer powerful performance insights by comparing wall clock latency distributions with actual on-cpu and I/O latency distributions. Consider potential strategies for managing xprobe overheads and their implications for future performance analysis techniques.

Syllabus

xprobes: Hybrid User/Kernel eBPF Probes for Cross-Layer Observability - Lucas Castanheira


Taught by

Linux Plumbers Conference

Related Courses

Hands-on Introduction to Pixie - Kubernetes Observability
Rawkode Academy via YouTube
Production Profiling - What, Why and How
Devoxx via YouTube
Set Your Code on Fire with FlameGraphs
Devoxx via YouTube
Using FlameGraphs to Illuminate the JVM
Devoxx via YouTube
A Tale of Two Flamegraphs - Continuous Profiling in Ruby
Ruby Central via YouTube