Qwen Bailian Coder
Code generation workload. Source: qwen_coder_blksz_16.jsonl. Timestamps are trace-relative seconds; no absolute wall-clock times or request durations are included.
Requests
43,011
Duration
2.000 h
Mean RPS
5.975
Peak 1s RPS
19.00
Input p50/p99
4,540.00 / 14,406.90
Input min/max
9 / 25,777
Output p50/p99
469.00 / 5,842.20
Sessions approx.
36,318
Hash reuse
66.4%
Concurrency note: this trace gives request arrival timestamps and token lengths, but not service durations. The RPS plots describe arrival load; actual in-flight concurrency depends on the serving system latency during replay.
Arrival Rate
Token Lengths
Structure
Statistics
| Metric | Mean | p50 | p90 | p95 | p99 | p99.9 |
|---|---|---|---|---|---|---|
| Input tokens | 5,747.66 | 4,540.00 | 12,776.00 | 13,406.50 | 14,406.90 | 17,283.99 |
| Output tokens | 808.26 | 469.00 | 1,789.00 | 2,699.00 | 5,842.20 | 12,710.77 |
| Total tokens | 6,555.91 | 5,505.00 | 13,213.00 | 13,889.00 | 16,772.00 | 25,382.94 |
| Turn | 2.174 | 1.000 | 5.000 | 7.000 | 13.00 | 25.00 |
| Session requests | 1.184 | 1.000 | 2.000 | 2.000 | 2.000 | 2.000 |
| Hash blocks/request | 359.70 | 284.00 | 799.00 | 838.00 | 901.00 | 1,081.00 |
Arrival RPS by Window
| Window | Mean | p50 | p90 | p95 | p99 | Max |
|---|---|---|---|---|---|---|
| 1s | 5.975 | 6.000 | 9.000 | 10.000 | 12.000 | 19.000 |
| 10s | 5.974 | 6.000 | 7.300 | 7.700 | 8.400 | 9.200 |
| 60s | 5.974 | 6.067 | 6.790 | 7.018 | 7.503 | 7.683 |
| 5min | 5.974 | 6.160 | 6.577 | 6.622 | 6.638 | 6.640 |
| 10min | 5.974 | 6.156 | 6.543 | 6.574 | 6.603 | 6.610 |
Request Types
This file uses a single native workload type, coder, so its per-type RPS matches the overall RPS.
| Type | Requests | Fraction | Mean RPS |
|---|---|---|---|
| coder | 43,011 | 100.00% | 5.975 |
Hash Blocks
| Metric | Value |
|---|---|
| Total hash blocks | 15,470,907 |
| Unique hash blocks | 5,202,517 |
| Global hash reuse ratio | 66.37% |
| Requests with any previously seen hash | 99.64% |