Testing large scale networks with Testground
summary
Testground runs reproducible, large-network experiments on Kubernetes and let teams spin up 100 validators and 1,000 light nodes to automate workload, network faults and resource assignments; Celestia uses it to submit ~9,911 transactions, validate 8 MB blocks with thousands of blobs, prove block reconstruction with ~400 light nodes, and show a 6x median bandwidth reduction in MLE. Tests use manifest.toml, composition.toml and behavior code, but are expensive on EKS and aren’t suited as a CI drop-in.
AI-written summary · about
chapters (12)
- 0:00Large network testing intro
- 2:14Testground and Celestia’s stack
- 4:00Testground’s key features
- 6:25Designing complex tests
- 8:12Actors and network structure
- 10:17Scalable actor behaviors
- 12:28Real-world API testing
- 14:59Monitoring test results
- 16:03Performance benchmarks
- 17:30Large network testing tips
- 19:15Testground future use cases
- 20:40Closing & Q&A
related talks
18mtesting in production gathering actually useful data
31mTests Are What You Want, What You Really Really Want
56mTesting at scale
56mTestOps at scale
26mCorralling the CNFs
27mTesting Before You Scale & Making Friends While You Do It
45mDistributed load testing with k6
46mDistributed load testing with k6