Efficient and Correct Test Scheduling for Ensembles of Network Policies
summary
Mikado is a scheduling layer that takes test traces from black‑box tools (bust, ATPG, simnet per uploader) and guarantees correctness by formalizing stateful network behavior and detecting read-write conflicts between traces. It builds a conflict graph and uses graph coloring to produce a safe, parallel schedule. Evaluation shows Mikado is open source, schedules thousands of policies across networks with 1000+ nodes in minutes, and yields up to orders of magnitude speed‑ups over sequential runs while preserving per‑trace behavior.
AI-written summary · about
chapters (19)
- 0:00Intro
- 0:27Background: Network Testing
- 1:40Network Testing in Practice
- 3:09Problem: How to Run Test Traces in Stateful Networks?
- 4:23Example: Incorrect Testing Results
- 7:18Mikado: Efficient and Correct Test Scheduling for Ensembles of Network Policies
- 8:27Ideal Scheduling for Running Test Traces
- 9:53Key Challenges
- 10:31Formal Framework of Stateful Networks
- 13:08Correctness of Running Multiple Tests
- 14:35Naive Scheduling Algorithm
- 15:53Read-write Conflict
- 17:58Scheduling Generation via Graph Coloring
- 18:25Implementation
- 18:50Reduction in Testing Time
- 20:46Scheduling Generation Scalable to 10K Test Traces
- 21:11Mikado Extensions
- 21:44Conclusion
- 22:46THE ADVANCED COMPUTING SYSTEMS ASSOCIATION
related talks
14mFeedback-guided Adaptive Testing of Distributed Systems Designs
12mMetha: Network Verifiers Need To Be Correct Too!
16mAutomatic Parallelization of Software Network Functions
14mMeshTest: End-to-End Testing for Service Mesh Traffic Management
16mToward Orchestration of Complex Networking Experiments
16mHeteroSketch: Coordinating Network-wide Monitoring in Heterogeneous and Dynamic Networks
19mNetcastle: Network Infrastructure Testing At Scale
21mModel-Agnostic and Efficient Exploration of Numerical State Space of Real-World...
