Skip to content

Latest commit

 

History

History
31 lines (18 loc) · 2.43 KB

README.md

File metadata and controls

31 lines (18 loc) · 2.43 KB

Falcon: Practical Log-based Analysis for Distributed Systems

Falcon is a tool aimed at easing the understanding of distributed system executions. Falcon ingests events logged by popular tracing tools (e.g. eBPF, strace, log4j) and merges them into a single causally-coherent execution trace. To further improve the reasoning about the runtime behavior of the system, Falcon also generates a space-time diagram depicting the events and their happens-before dependencies.

The architecture of Falcon is composed of three main components that operate together as a pipeline:

  • falcon-tracer uses eBPF to trace events of interest (e.g. start, end, fork, join, send, receive, etc) at runtime.
  • falcon-solver combines the events into a global execution trace that preserves causality. This is achieved by i) building a symbolic constraint model that encodes the happens-before relationships between events, ii) using an SMT solver to solve the constraints and assign a logical clock to each event such that all causal dependencies are satisfied.
  • falcon-visualizer draws a space-time diagram that enables a visual analysis of the whole execution.

For additional details about Falcon, please check our DSN'18 paper.

Installation

Falcon's components are independent and, therefore, can be leveraged by different projects. To set up the full pipeline, install each particular component according to the following instructions:

Usage

Falcon operates in three different phases, namely event tracing, causality inference, and space-time diagram visualization. Each phase is performed by a particular component of Falcon:

For a concrete example of how to use falcon in practice, please check our Tutorial - Falcon with Zookeeper.