OpenMessaging Benchmark
Explore results across
krabka releases
Compare published releases, then follow each trial over time. Keep throughput, latency and test conditions in view.
Each published release has one round per workload. The interval samples within a trial are not independent repeated trials. Resource timelines are available only where captured.
Go to the trial timelineWorkload by workload
Read the pattern across releases
Each cell compares krabka with a broker in the same run. Select a cell to inspect the captured intervals.
100% is the comparison broker. Favorable by more than 5% · within 5% · unfavorable by more than 5%. These bands are visual guides, not statistical significance.
Rates on fixed-load workloads can be capped by offered load. Max-rate workloads give a clearer view of throughput limits.
| Workload | 1.0.11 round · trim uncontrolled | 1.0.31 round · trim uncontrolled | 1.0.51 round · SSD trimmed |
|---|---|---|---|
| simple-workload | 100% | 100% | 100% |
| 1-topic-1-partition-1kb | 100% | 100% | 100% |
| 1-topic-1-partition-100b | 100% | 100% | 100% |
| 1-topic-16-partitions-1kb | 100% | 100% | 100% |
| backlog-1-topic-1-partition-1kb | 100% | 100% | 100% |
| backlog-1-topic-16-partitions-1kb | 100% | 100% | 100% |
| max-rate-1-topic-1-partition-1p-1c-1kb | 74% | 116% | 112% |
| max-rate-1-topic-1-partition-1p-1c-100b | 100% | 105% | 103% |
| 1-topic-3-partition-100b-3producers | 99% | 100% | 93% |
| max-rate-1-topic-16-partitions-1kb | 97% | 123% | 106% |
| max-rate-1-topic-16-partitions-100b | 113% | 92% | 53% |
| max-rate-1-topic-100-partitions-1kb | 95% | 110% | 112% |
| max-rate-1-topic-100-partitions-100b | 118% | 71% | 165% |
| simple-workload | 100% | 100% | 100% |
| 1-topic-1-partition-1kb | 100% | 100% | 100% |
| 1-topic-1-partition-100b | 100% | 100% | 100% |
| 1-topic-16-partitions-1kb | 100% | 100% | 100% |
| backlog-1-topic-1-partition-1kb | 100% | 100% | 100% |
| backlog-1-topic-16-partitions-1kb | 100% | 100% | 100% |
| max-rate-1-topic-1-partition-1p-1c-1kb | 75% | 72% | 101% |
| max-rate-1-topic-1-partition-1p-1c-100b | 92% | 92% | 96% |
| 1-topic-3-partition-100b-3producers | 81% | 69% | 94% |
| max-rate-1-topic-16-partitions-1kb | 72% | 98% | 104% |
| max-rate-1-topic-16-partitions-100b | 105% | 104% | 115% |
| max-rate-1-topic-100-partitions-1kb | 100% | 97% | 103% |
| max-rate-1-topic-100-partitions-100b | 108% | 101% | 87% |
Bars share a 0–200% scale; longer values stop at the edge and keep their numeric label. Every workload has equal weight in the median. Missing comparisons are excluded and counted.
Disk preparation changed in the latest run. Cross-release differences include test conditions. The full suite uses krabka’s shared page cache, not its direct I/O mode.
Loading captured data…
Temporal comparison
Exact samples and CSV export
Latency distribution
Aggregate quantiles are recorded by OMB across the measurement. They are not averages of interval percentiles.
Full upstream cumulative distribution
Latency uses log(1 + value). The vertical axis is the percentile at or below that latency. Every upstream quantile is preserved in result.json.
Trial drilldown
Choose a captured file and select Inspect file to browse its fields or lines.
Complete capture
Trial catalog
Archive integrity and provenance
OMB results retain every upstream interval. Older captures include resource timelines and client logs from the verified Git LFS archive. Newer publications contain OMB files and compact trial summaries; missing telemetry is not reconstructed.
Read the archive verification instructions