Published June 14, 2026
| Version v1
Dataset
Open
Online Boutique Telemetry Dataset (Prometheus Metrics + Jaeger Traces), Minute-level Tables, ClarkNet/NASA Replay
Authors/Creators
Description
Minute-level telemetry tables for the Online Boutique microservices benchmark (11 services), collected under fixed replica counts (no autoscaling) while replaying two real-world web traffic shapes:
- ClarkNet-HTTP workload replay
- NASA-HTTP workload replay
Under varying injected load, we record how service-level metrics change over time and how inter-service calls evolve, derived from Prometheus (Istio/service metrics) and Jaeger traces, and exported as ready-to-use CSV tables.
Files (per run)
- service_minute_features.csv — per-service minute features (RPS, latency, error/timeout rates, CPU/memory, replica state, dependency aggregates)
- service_call_edge_minute.csv — per-minute caller→callee edges (call_count, latency, errors)
- service_metadata.csv, run_metadata.csv — service list and run/scenario metadata
Granularity: 1-minute UTC buckets.
Intended uses
- Workload/demand forecasting under changing traffic
- Service performance & reliability analysis (latency/error vs load)
- Dependency / call-graph dynamics analysis
- Cross-workload comparison (ClarkNet vs NASA) with the same schema