Methodology
Overview
We measure Polkadot's reliability using enterprise-standard metrics: end-to-end transaction latency, confidence (success rates), and cost stability. We track the complete user experience across three lifecycle stages—Ready, InBlock, Finalized—using long-tail percentiles to capture consistent performance, not just averages.
Why Long-Tail Percentiles?
Reliability means consistent performance across the tail. Like airline punctuality: 90% on-time isn't good if 10% are hours late. We use long-tail percentiles (P95) to show what the vast majority of users actually experience, not just the average.
Transaction Lifecycle
Transactions progress through stages in a distributed system:
- Ready - Transaction validated, dependencies met, eligible for next block inclusion.
- InBlock - Transaction included in a block and imported by node.
- Finalized - Transaction and block irreversibly committed to relay chain.
Metrics
Latency
End-to-end time from submission to each stage. Captures all distributed system delays—network, pool, forks—not just block production.
Block Production
Time between consecutive blocks. Measures network consensus health.
Data Sources
Latency: Prometheus metrics via a Thanos persistence layer.
Block Production: Block data from DotLake.