Results
Pilot content: the Tier 1 default-transport validation (measured 2026-10-18). Every cell below ships with all of its rounds; nothing is cherry-picked.
Verdict — gate: mixed. gate: mixed
Correctness-backed Tier 1 default selection; not an overall performance win. Ordinary-route throughput vs old Tier 1 ranges -5.1% to +3.7% in median with overlapping spreads; POST c512 and known-length stream c512 p99 medians regress. No leaderboard update.
This is a correctness-backed default selection, not an overall performance win: POST c512 median p99 is worse than old Tier 1, and known-length streaming c512 median p99 is worse than both stock controls. Those signals sit beside the throughput figures, not behind them (report, Interpretation and tradeoff).
Throughput (requests/second, latency collection disabled)
| Cell | Old stock | Old Tier 1 | Current default | Raw NIO | ||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GET / · c64 | 203,6644 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 202,2484 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 199,1484 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 196,8474 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | ||||||||||||||||||||||||||||||||||||||||
| GET / · c256 | 206,6804 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 200,9724 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 195,2484 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 205,4034 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | ||||||||||||||||||||||||||||||||||||||||
| GET / · c512 | 187,5964 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 198,5184 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 188,3944 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 203,9124 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | ||||||||||||||||||||||||||||||||||||||||
| GET /user/42 · c64 | 172,5264 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 201,1504 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 203,7084 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 195,7824 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | ||||||||||||||||||||||||||||||||||||||||
| GET /user/42 · c256 | 177,6964 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 201,8764 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 195,3134 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 185,0494 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | ||||||||||||||||||||||||||||||||||||||||
| GET /user/42 · c512 | 175,2124 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 189,3784 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 196,3874 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 186,6854 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | ||||||||||||||||||||||||||||||||||||||||
| POST /user · c64 | 192,2664 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 191,2264 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 188,9564 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 194,4284 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | ||||||||||||||||||||||||||||||||||||||||
| POST /user · c256 | 190,6804 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 187,6804 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 178,0704 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 192,3084 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | ||||||||||||||||||||||||||||||||||||||||
| POST /user · c512 | 184,6854 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 183,6164 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 183,2384 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 183,9304 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment throughput table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json |
p99 latency (microseconds, separate latency-enabled runs)
| Cell | Old stock | Old Tier 1 | Current default | Raw NIO | ||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GET / · c64 | 6734 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 8634 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 7644 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 7294 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | ||||||||||||||||||||||||||||||||||||||||
| GET / · c256 | 2,5484 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 2,9294 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 2,5384 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 2,6894 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | ||||||||||||||||||||||||||||||||||||||||
| GET / · c512 | 4,9414 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 5,2024 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 5,0374 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 6,1604 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | ||||||||||||||||||||||||||||||||||||||||
| GET /user/42 · c64 | 8014 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 6954 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 7274 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 7554 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | ||||||||||||||||||||||||||||||||||||||||
| GET /user/42 · c256 | 2,8514 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 2,6984 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 2,6314 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 2,6614 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | ||||||||||||||||||||||||||||||||||||||||
| GET /user/42 · c512 | 6,8984 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 5,0534 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 5,0284 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 5,9404 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | ||||||||||||||||||||||||||||||||||||||||
| POST /user · c64 | 7154 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 7094 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 7884 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 7004 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | ||||||||||||||||||||||||||||||||||||||||
| POST /user · c256 | 2,6384 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 2,3924 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 2,4484 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 2,9644 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | ||||||||||||||||||||||||||||||||||||||||
| POST /user · c512 | 5,7344 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 3,6284 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 5,1754 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 5,3284 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment p99 table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json |
Response-stream controls (throughput RPS / p99 µs)
Baseline stock, current forced stock, and current default on known-length and chunkedHello, World! responses. The baseline Tier 1 cannot serve these streams, so it is not a stream control (report, Network experiment).
| Cell | Old stock | Current forced stock | Current default | |||||||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Known length · c64 | 127,650 / 1,2844 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment stream table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 120,894 / 1,0494 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment stream table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 141,240 / 9384 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment stream table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | |||||||||||||||||||||||||||||||||||||||||||||
| Known length · c256 | 130,528 / 3,0204 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment stream table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 128,141 / 3,5374 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment stream table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 141,240 / 3,1934 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment stream table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | |||||||||||||||||||||||||||||||||||||||||||||
| Known length · c512 | 127,944 / 5,0014 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment stream table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 128,762 / 5,5964 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment stream table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 138,387 / 8,7664 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment stream table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | |||||||||||||||||||||||||||||||||||||||||||||
| Chunked · c64 | 109,574 / 1,1024 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment stream table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 112,207 / 1,1314 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment stream table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 122,856 / 1,0984 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment stream table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | |||||||||||||||||||||||||||||||||||||||||||||
| Chunked · c256 | 110,691 / 3,9314 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment stream table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 111,648 / 4,6854 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment stream table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 123,906 / 3,6734 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment stream table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | |||||||||||||||||||||||||||||||||||||||||||||
| Chunked · c512 | 109,386 / 7,6924 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment stream table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 110,284 / 9,8634 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment stream table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json | 128,152 / 5,6024 rounds + source
Median of 4 runs. Source: Pilot report, Network experiment stream table (medians of 4 runs); per-round values from tier1-default-2026-10-18.json |
Known-length streams at c512: the four current-default p99 values are 10,927.625 / 15,658.458 / 6,604 / 5,545.167 µs (rounds 1–4), versus old stock 5,344.584 / 16,762.083 / 4,363.209 / 4,657.125 µs. The medians indicate a substantial tail cost despite higher throughput; the overlapping excursions prevent a clean mechanism attribution (report, Network experiment).
Exporter chart (GET /, all concurrencies)
CPU and memory
Median process CPU-seconds per 250,000 throughput requests (100 ticks/second), and maximum observed process RSS per server. RSS is a process-lifetime high-water mark including earlier warm-up and load, not a fresh-process peak per cell, and this finite workload is not a long-duration leak test (report, Network experiment).
| Route / connections | Old stock | Old Tier 1 | Current default | Raw NIO |
|---|---|---|---|---|
| GET /, 256 | 2.395 | 2.185 | 2.170 | 2.380 |
| GET /user/42, 512 | 2.840 | 2.515 | 2.420 | 2.685 |
Known-length c512 streams use 3.910, 3.880, 3.615CPU-seconds on old stock, current forced stock, current default respectively. Maximum observed process RSS: 24,968 KiB (old stock), 24,084 KiB (old tier 1), 24,192 KiB (current default), 24,996 KiB (current forced stock), 24,332 KiB (raw nio).
How the matrix was run
- Each cell: 20,000-request warm-up, then separate throughput and latency samples of 250,000 requests.
- Four rounds in alternating server order; concurrency 64, 256, and 512.
- Servers: baseline stock, baseline Tier 1, current default Tier 1, and unchanged raw NIO (plus current forced stock for streams).
- Two event-loop threads; server CPUs 0–3, client CPUs 4–7; HTTP/1.1 persistent connections, no TLS or compression.
- The client requests four threads but uses two event loops (double
group.next()); held constant across servers. - All 432 samples completed: 108 million measured requests with no client failures.
Two generator caveats apply to this matrix: the pre-fix loop distribution above, and the socket-option erratum (historical TCP_NODELAY=0 labels did not prove the effective value). See the report archive for errata and the full report for commands.
WhiteThroated