An Online Probabilistic Distributed Tracing System

TL;DR

This paper introduces Astraea, an online probabilistic distributed tracing system that uses Bayesian learning and multi-armed bandits to reduce tracing overhead while maintaining diagnostic accuracy in cloud environments.

Contribution

Astraea is the first system to adaptively steer distributed tracing instrumentation using probabilistic models, significantly lowering overhead without sacrificing diagnostic utility.

Findings

Reduces tracing overhead to 10-28% of instrumentation

Decreases storage and compute costs substantially

Maintains high accuracy in performance diagnosis

Abstract

Distributed tracing has become a fundamental tool for diagnosing performance issues in the cloud by recording causally ordered, end-to-end workflows of request executions. However, tracing in production workloads can introduce significant overheads due to the extensive instrumentation needed for identifying performance variations. This paper addresses the trade-off between the cost of tracing and the utility of the "spans" within that trace through Astraea, an online probabilistic distributed tracing system. Astraea is based on our technique that combines online Bayesian learning and multi-armed bandit frameworks. This formulation enables Astraea to effectively steer tracing towards the useful instrumentation needed for accurate performance diagnosis. Astraea localizes performance variations using only 10-28% of available instrumentation, markedly reducing tracing overhead, storage, compute costs, and trace analysis time.