LeitMotif: An Abstraction for Debugging Distributed Applications
summary
Claims an actionable abstraction called workflow motifs that are mined from end-to-end traces to capture recurring processing patterns, their latency distributions, and critical paths. Implementation runs alongside apps, ingests X-Ray and Jaeger traces, slices graphs, and uses frequent subgraph mining built from Arabesque, Gaston, Graeme and sampling ideas to find motifs. Example: a scan expected at 10s actually takes 50s because reads are issued sequentially vs parallel and motif-based analysis points to parallelization as the fix.
AI-written summary · about
chapters (9)
- 0:00The debugging challenge
- 3:06Introducing workflow motif
- 5:35Methodology and enablers
- 8:31Practical application
- 10:46Formalizing the model
- 13:36Core design requirements
- 16:21Mining tools and strategies
- 20:08Results and conclusion
- 22:34Q&A session
related talks
57mDistributed Tracing: How the Pros Debug Concurrent and Distributed Systems
44mNow You See Me Too: Visual Tooling for Advanced System Analysis
24mDistributed Tracing : New DevOps Foundation.
51mNavigating Distributed Traces
52mPerfecting reliable code delivery for the cloud
38mDistributed Tracing: Practices, Patterns & Technologies for Debugging Microservices
35mTracing is More Than Traces: The Insights in Trace Aggregates
45mTracing: A journey to tactical insights