# zoo-bench performance report

raw data: `results/zoo-framework-0.10.0/20261011T093201Z.json`

## Run environment

| item | value |
|---|---|
| CPU model | AMD EPYC 7763 64-Core Processor |
| logical cores | 4 |
| platform | Linux-x86_64 |
| operating system | Linux 6.17.0-1022-azure (#22-Ubuntu SMP Mon Jul 27 17:24:03 UTC 2026) |
| Python | CPython 3.13.16 |
| interpreter | /opt/hostedtoolcache/Python/3.13.16/x64/bin/python |
| GIL | default |
| framework under test (distribution metadata) | 0.10.0 |
| framework under test (module __version__) | 0.10.0 (footnote only, not the version criterion) |
| drive-surface generation of the framework under test | current-generation drive surface (scheduling core + scheduling model; the completion signal comes from the result reactor) |
| harness | zoo-bench 0.0.1 @ 9a42be24ac43f75e2e5d7b16434f91ba3d0717b6 |
| command to reproduce | `zoo-bench run --framework zoo-framework==0.10.0` |

> dist_version (the distribution metadata) is the source of truth; module_version is only a footnote

## Workload

- composition: one unit of work = JSON serialize + deserialize + string concatenation, repeated until the target duration is reached; the tiers are declared by body_tiers_us in matrix.yaml
- **the workload is a stand-in, not a real trace**: its shape resembles business actions such as 'take a frame of data -> serialize -> parse -> transform strings', but this repository holds no real trace, so every conclusion is conditional on that

## Measurement methodology

- **the body duration is instrumented within the same run**: the body measures itself internally and reports back; framework overhead = end-to-end - it. **No cross-run subtraction**  -  two runs differ in state (caches, frequencies, scheduling noise), so subtracting introduces systematic bias.
- **warmups are excluded from statistics**: each unit first runs 3 warmup rounds that are discarded, then only the following 30 measured rounds are counted; the sample count equals 'measured rounds x concurrency'.
- **absolute durations are not comparable across runs**: they only show the order of magnitude; across versions/machines read the **relative quantities within the same run** (overhead share, multiples against the baselines), which are insensitive to overall speed.

## Conclusion summary

- Once the framework under test zoo runs a body of about 2700 microseconds or more, its framework overhead drops below 15% of end-to-end; below that tier, the main price of choosing it is the framework overhead itself.
- At concurrency 1 and a 40 microsecond body, the framework under test is 2.11x faster than process_pool (the fastest measured tier at that concurrency).
- At concurrency 1, once the body exceeds about 40 microseconds, apscheduler is no longer faster than the framework under test.
- At concurrency 4, once the body exceeds about 40 microseconds, apscheduler is no longer faster than the framework under test.
- At concurrency 16, once the body exceeds about 40 microseconds, apscheduler is no longer faster than the framework under test.
- At concurrency 64, once the body exceeds about 40 microseconds, apscheduler is no longer faster than the framework under test.
- At concurrency 1, within the measured tiers asyncio_pool is **always faster** than the framework under test  -  tiers like these are exactly the openly published unfavourable data of this report.
- At concurrency 4, within the measured tiers asyncio_pool is **always faster** than the framework under test  -  tiers like these are exactly the openly published unfavourable data of this report.
- At concurrency 16, within the measured tiers asyncio_pool is **always faster** than the framework under test  -  tiers like these are exactly the openly published unfavourable data of this report.
- At concurrency 64, within the measured tiers asyncio_pool is **always faster** than the framework under test  -  tiers like these are exactly the openly published unfavourable data of this report.
- At concurrency 1, once the body exceeds about 10000 microseconds, bare_thread is no longer faster than the framework under test.
- At concurrency 4, within the measured tiers bare_thread is **always faster** than the framework under test  -  tiers like these are exactly the openly published unfavourable data of this report.
- At concurrency 16, once the body exceeds about 40 microseconds, bare_thread is no longer faster than the framework under test.
- At concurrency 64, within the measured tiers bare_thread is **always faster** than the framework under test  -  tiers like these are exactly the openly published unfavourable data of this report.
- At concurrency 1, once the body exceeds about 40 microseconds, process_pool is no longer faster than the framework under test.
- At concurrency 4, once the body exceeds about 40 microseconds, process_pool is no longer faster than the framework under test.
- At concurrency 16, within the measured tiers process_pool is **always faster** than the framework under test  -  tiers like these are exactly the openly published unfavourable data of this report.
- At concurrency 64, once the body exceeds about 40 microseconds, process_pool is no longer faster than the framework under test.
- At concurrency 1, within the measured tiers thread_pool is **always faster** than the framework under test  -  tiers like these are exactly the openly published unfavourable data of this report.
- At concurrency 4, within the measured tiers thread_pool is **always faster** than the framework under test  -  tiers like these are exactly the openly published unfavourable data of this report.
- At concurrency 16, within the measured tiers thread_pool is **always faster** than the framework under test  -  tiers like these are exactly the openly published unfavourable data of this report.
- At concurrency 64, within the measured tiers thread_pool is **always faster** than the framework under test  -  tiers like these are exactly the openly published unfavourable data of this report.

> a single-point speedup has no selection meaning; the conclusion is phrased as 'which scheme to choose at what body duration' (the same framework overhead is 70% of a 40 microsecond task and a few percent of a 10 millisecond one)

### Overhead threshold crossings

| scheme | concurrency | threshold | first tier at or below the threshold (microseconds) | measured tiers (microseconds) | note |
|---|---|---|---|---|---|
| apscheduler | 1 | 15.00% | 2700 | 40, 300, 2700, 10000 | this is the smallest measured tier that satisfies the condition; the true crossing point lies between it and the previous tier |
| apscheduler | 4 | 15.00% | - | 40, 300, 2700, 10000 | this concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values). End-to-end, throughput and the body's self-reported values remain valid, so they are given as usual. |
| apscheduler | 16 | 15.00% | - | 40, 300, 2700, 10000 | this concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values). End-to-end, throughput and the body's self-reported values remain valid, so they are given as usual. |
| apscheduler | 64 | 15.00% | - | 40, 300, 2700, 10000 | this concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values). End-to-end, throughput and the body's self-reported values remain valid, so they are given as usual. |
| asyncio_pool | 1 | 15.00% | 2700 | 40, 300, 2700, 10000 | this is the smallest measured tier that satisfies the condition; the true crossing point lies between it and the previous tier |
| asyncio_pool | 4 | 15.00% | - | 40, 300, 2700, 10000 | this concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values). End-to-end, throughput and the body's self-reported values remain valid, so they are given as usual. |
| asyncio_pool | 16 | 15.00% | - | 40, 300, 2700, 10000 | this concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values). End-to-end, throughput and the body's self-reported values remain valid, so they are given as usual. |
| asyncio_pool | 64 | 15.00% | - | 40, 300, 2700, 10000 | this concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values). End-to-end, throughput and the body's self-reported values remain valid, so they are given as usual. |
| bare_thread | 1 | 15.00% | 2700 | 40, 300, 2700, 10000 | this is the smallest measured tier that satisfies the condition; the true crossing point lies between it and the previous tier |
| bare_thread | 4 | 15.00% | - | 40, 300, 2700, 10000 | this concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values). End-to-end, throughput and the body's self-reported values remain valid, so they are given as usual. |
| bare_thread | 16 | 15.00% | - | 40, 300, 2700, 10000 | this concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values). End-to-end, throughput and the body's self-reported values remain valid, so they are given as usual. |
| bare_thread | 64 | 15.00% | - | 40, 300, 2700, 10000 | this concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values). End-to-end, throughput and the body's self-reported values remain valid, so they are given as usual. |
| process_pool | 1 | 15.00% | 2700 | 40, 300, 2700, 10000 | this is the smallest measured tier that satisfies the condition; the true crossing point lies between it and the previous tier |
| process_pool | 4 | 15.00% | - | 40, 300, 2700, 10000 | this concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values). End-to-end, throughput and the body's self-reported values remain valid, so they are given as usual. |
| process_pool | 16 | 15.00% | - | 40, 300, 2700, 10000 | this concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values). End-to-end, throughput and the body's self-reported values remain valid, so they are given as usual. |
| process_pool | 64 | 15.00% | - | 40, 300, 2700, 10000 | this concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values). End-to-end, throughput and the body's self-reported values remain valid, so they are given as usual. |
| thread_pool | 1 | 15.00% | 300 | 40, 300, 2700, 10000 | this is the smallest measured tier that satisfies the condition; the true crossing point lies between it and the previous tier |
| thread_pool | 4 | 15.00% | - | 40, 300, 2700, 10000 | this concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values). End-to-end, throughput and the body's self-reported values remain valid, so they are given as usual. |
| thread_pool | 16 | 15.00% | - | 40, 300, 2700, 10000 | this concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values). End-to-end, throughput and the body's self-reported values remain valid, so they are given as usual. |
| thread_pool | 64 | 15.00% | - | 40, 300, 2700, 10000 | this concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values). End-to-end, throughput and the body's self-reported values remain valid, so they are given as usual. |
| zoo | 1 | 15.00% | 2700 | 40, 300, 2700, 10000 | this is the smallest measured tier that satisfies the condition; the true crossing point lies between it and the previous tier |
| zoo | 4 | 15.00% | - | 40, 300, 2700, 10000 | this concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values). End-to-end, throughput and the body's self-reported values remain valid, so they are given as usual. |
| zoo | 16 | 15.00% | - | 40, 300, 2700, 10000 | this concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values). End-to-end, throughput and the body's self-reported values remain valid, so they are given as usual. |
| zoo | 64 | 15.00% | - | 40, 300, 2700, 10000 | this concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values). End-to-end, throughput and the body's self-reported values remain valid, so they are given as usual. |

### Relative turning points

| baseline | concurrency | first tier where it is no longer faster (microseconds) | note |
|---|---|---|---|
| apscheduler | 1 | 40 | this is the smallest measured tier that satisfies the condition |
| apscheduler | 4 | 40 | this is the smallest measured tier that satisfies the condition |
| apscheduler | 16 | 40 | this is the smallest measured tier that satisfies the condition |
| apscheduler | 64 | 40 | this is the smallest measured tier that satisfies the condition |
| asyncio_pool | 1 | - | **not observed**: within the measured tiers this baseline is always faster than the framework under test |
| asyncio_pool | 4 | - | **not observed**: within the measured tiers this baseline is always faster than the framework under test |
| asyncio_pool | 16 | - | **not observed**: within the measured tiers this baseline is always faster than the framework under test |
| asyncio_pool | 64 | - | **not observed**: within the measured tiers this baseline is always faster than the framework under test |
| bare_thread | 1 | 10000 | this is the smallest measured tier that satisfies the condition |
| bare_thread | 4 | - | **not observed**: within the measured tiers this baseline is always faster than the framework under test |
| bare_thread | 16 | 40 | this is the smallest measured tier that satisfies the condition |
| bare_thread | 64 | - | **not observed**: within the measured tiers this baseline is always faster than the framework under test |
| process_pool | 1 | 40 | this is the smallest measured tier that satisfies the condition |
| process_pool | 4 | 40 | this is the smallest measured tier that satisfies the condition |
| process_pool | 16 | - | **not observed**: within the measured tiers this baseline is always faster than the framework under test |
| process_pool | 64 | 40 | this is the smallest measured tier that satisfies the condition |
| thread_pool | 1 | - | **not observed**: within the measured tiers this baseline is always faster than the framework under test |
| thread_pool | 4 | - | **not observed**: within the measured tiers this baseline is always faster than the framework under test |
| thread_pool | 16 | - | **not observed**: within the measured tiers this baseline is always faster than the framework under test |
| thread_pool | 64 | - | **not observed**: within the measured tiers this baseline is always faster than the framework under test |

## Charts

![overhead_ratio](figures/en-overhead_ratio.svg)

> Body tiers at concurrency 4 only; ratios hold within a single run. The overhead numbers of concurrency 16, 64 were withheld (beyond this machine's parallelism; see the caliber section) and are not charted

![throughput](figures/en-throughput.svg)

> Body tier 40 microseconds only; throughput holds within a single run

![relative_multiple](figures/en-relative_multiple.svg)

> One panel per measured concurrency (4 in total); multiples hold within a single run

## Dimension: Latency percentiles and jitter

> end-to-end / concurrency; percentiles and relative spread are given in the same table  -  a metric that only gives the median is not acceptable

| concurrency | body tier (microseconds) | zoo | apscheduler | asyncio_pool | bare_thread | process_pool | thread_pool |
|---|---|---|---|---|---|---|---|
| 1 | 40 | 157.73 microseconds | 288.55 microseconds | 140.57 microseconds | 141.79 microseconds | 332.18 microseconds | 83.56 microseconds |
| 1 | 300 | 424.91 microseconds | 547.57 microseconds | 404.67 microseconds | 405.40 microseconds | 606.50 microseconds | 323.16 microseconds |
| 1 | 2700 | 2.797 milliseconds | 3.120 milliseconds | 2.605 milliseconds | 2.717 milliseconds | 3.017 milliseconds | 2.667 milliseconds |
| 1 | 10000 | 9.931 milliseconds | 9.890 milliseconds | 9.476 milliseconds | 9.972 milliseconds | 10.015 milliseconds | 9.618 milliseconds |
| 4 | 40 | 183.23 microseconds | 253.08 microseconds | 159.80 microseconds | 162.08 microseconds | 188.60 microseconds | 94.96 microseconds |
| 4 | 300 | 439.01 microseconds | 519.95 microseconds | 410.40 microseconds | 409.00 microseconds | 295.22 microseconds | 362.52 microseconds |
| 4 | 2700 | 2.848 milliseconds | 3.035 milliseconds | 2.749 milliseconds | 2.659 milliseconds | 1.398 milliseconds | 2.706 milliseconds |
| 4 | 10000 | 10.050 milliseconds | 10.021 milliseconds | 9.977 milliseconds | 9.793 milliseconds | 4.598 milliseconds | 9.798 milliseconds |
| 16 | 40 | 170.72 microseconds | 251.40 microseconds | 138.08 microseconds | 174.45 microseconds | 160.88 microseconds | 90.52 microseconds |
| 16 | 300 | 439.04 microseconds | 501.69 microseconds | 394.14 microseconds | 430.66 microseconds | 280.12 microseconds | 351.18 microseconds |
| 16 | 2700 | 2.926 milliseconds | 2.952 milliseconds | 2.822 milliseconds | 2.724 milliseconds | 1.395 milliseconds | 2.718 milliseconds |
| 16 | 10000 | 10.507 milliseconds | 10.271 milliseconds | 9.992 milliseconds | 9.834 milliseconds | 4.711 milliseconds | 10.001 milliseconds |
| 64 | 40 | 184.98 microseconds | 224.83 microseconds | 123.94 microseconds | 169.50 microseconds | 195.74 microseconds | 89.55 microseconds |
| 64 | 300 | 472.05 microseconds | 505.68 microseconds | 373.95 microseconds | 424.17 microseconds | 297.44 microseconds | 349.61 microseconds |
| 64 | 2700 | 2.937 milliseconds | 2.974 milliseconds | 2.777 milliseconds | 2.722 milliseconds | 1.417 milliseconds | 2.768 milliseconds |
| 64 | 10000 | 10.515 milliseconds | 10.311 milliseconds | 10.449 milliseconds | 9.812 milliseconds | 4.709 milliseconds | 10.022 milliseconds |

per-unit values (percentiles, spread, raw per-round samples) are in 'Appendix: all values' at the end.

## Dimension: Framework overhead share

> overhead = end-to-end / concurrency - the body duration measured within the same run (no cross-run subtraction). **cells with -**: that concurrency exceeds this machine's parallel capacity; the body's self-reported duration contains scheduling waits from oversubscription and is not comparable with the row's end-to-end per task, so no overhead figure is given (end-to-end and throughput live in their own dimensions)

| concurrency | body tier (microseconds) | zoo | apscheduler | asyncio_pool | bare_thread | process_pool | thread_pool |
|---|---|---|---|---|---|---|---|
| 1 | 40 | 67.93% | 82.14% | 66.96% | 63.81% | 82.88% | 38.45% |
| 1 | 300 | 28.57% | 44.63% | 26.17% | 28.08% | 50.03% | 10.75% |
| 1 | 2700 | 5.17% | 9.74% | 4.50% | 4.20% | 11.29% | 1.80% |
| 1 | 10000 | 1.81% | 4.13% | 1.41% | 1.26% | 3.64% | 0.60% |
| 4 | 40 | 72.88% | 77.73% | 66.65% | 68.11% | 49.63% | 47.40% |
| 4 | 300 | 31.79% | 40.54% | 25.54% | 27.64% | - | 15.92% |
| 4 | 2700 | 5.27% | 9.67% | 4.68% | 4.32% | - | 3.15% |
| 4 | 10000 | - | - | - | - | - | - |
| 16 | 40 | - | - | - | - | - | - |
| 16 | 300 | - | - | - | - | - | - |
| 16 | 2700 | - | - | - | - | - | - |
| 16 | 10000 | - | - | - | - | - | - |
| 64 | 40 | - | - | - | - | - | - |
| 64 | 300 | - | - | - | - | - | - |
| 64 | 2700 | - | - | - | - | - | - |
| 64 | 10000 | - | - | - | - | - | - |

per-unit values (percentiles, spread, raw per-round samples) are in 'Appendix: all values' at the end.

## Dimension: Throughput and concurrency scaling

> throughput = concurrency / median end-to-end; whether it keeps rising with concurrency is the scaling behaviour

| concurrency | body tier (microseconds) | zoo | apscheduler | asyncio_pool | bare_thread | process_pool | thread_pool |
|---|---|---|---|---|---|---|---|
| 1 | 40 | 6340.1 | 3465.7 | 7114.0 | 7052.9 | 3010.4 | 11967.2 |
| 1 | 300 | 2353.5 | 1826.2 | 2471.2 | 2466.7 | 1648.8 | 3094.4 |
| 1 | 2700 | 357.6 | 320.5 | 383.9 | 368.0 | 331.4 | 374.9 |
| 1 | 10000 | 100.7 | 101.1 | 105.5 | 100.3 | 99.9 | 104.0 |
| 4 | 40 | 21830.7 | 15805.1 | 25031.0 | 24678.8 | 21208.8 | 42122.3 |
| 4 | 300 | 9111.4 | 7693.1 | 9746.5 | 9780.0 | 13549.0 | 11034.0 |
| 4 | 2700 | 1404.6 | 1317.9 | 1455.1 | 1504.0 | 2861.8 | 1478.5 |
| 4 | 10000 | 398.0 | 399.2 | 400.9 | 408.4 | 869.9 | 408.3 |
| 16 | 40 | 93723.0 | 63644.3 | 115870.7 | 91714.4 | 99452.7 | 176759.3 |
| 16 | 300 | 36443.4 | 31892.4 | 40594.7 | 37152.4 | 57119.0 | 45560.6 |
| 16 | 2700 | 5468.7 | 5419.3 | 5670.2 | 5873.4 | 11467.3 | 5885.6 |
| 16 | 10000 | 1522.8 | 1557.7 | 1601.3 | 1626.9 | 3396.2 | 1599.9 |
| 64 | 40 | 345987.8 | 284656.4 | 516381.4 | 377578.2 | 326969.3 | 714659.3 |
| 64 | 300 | 135579.6 | 126561.3 | 171144.5 | 150881.9 | 215170.2 | 183061.8 |
| 64 | 2700 | 21794.5 | 21519.8 | 23045.6 | 23513.0 | 45160.7 | 23119.9 |
| 64 | 10000 | 6086.6 | 6207.0 | 6125.1 | 6522.6 | 13591.7 | 6385.7 |

per-unit values (percentiles, spread, raw per-round samples) are in 'Appendix: all values' at the end.

## Dimension: Overhead attribution

> scope: splits the per-task end-to-end into submit side / handoff / body / return and, at the same tier and concurrency, **compares each segment against every baseline**, answering 'which segment the excess falls on'. It is **diagnostic, not the main evidence**: it samples a small set of tiers, and does not compare total overhead against the baselines. **A negative handoff is not an error**: it means the submit call had not returned yet while the body was already running  -  schemes that count thread creation as submit side behave this way (measured: bare_thread handoff -343 microseconds). **The return segment includes waiting for the completion signal** (the last stretch before draining returns); it shares the main measurement's scope

**status: ok**

> this dimension is a **diagnosis**, not the main evidence: the sampled tiers and concurrencies are a fixed small batch, not the whole matrix. The concurrency ceiling is the machine's parallel capacity - beyond it, what is measured is contention;

sampled tiers (microseconds): 40, 10000; concurrency: [1, 4]; this machine's parallel capacity: 4

| scheme | concurrency | tier (microseconds) | submit side | handoff | body | return | status |
|---|---|---|---|---|---|---|---|
| apscheduler | 1 | 40 | 110.01 microseconds | 83.43 microseconds | 58.01 microseconds | 43.81 microseconds | ok |
| apscheduler | 4 | 40 | 101.53 microseconds | 508.15 microseconds | 56.80 microseconds | 221.34 microseconds | ok |
| apscheduler | 1 | 10000 | 168.54 microseconds | 113.17 microseconds | 9.639 milliseconds | 63.68 microseconds | ok |
| apscheduler | 4 | 10000 | 132.04 microseconds | 10.592 milliseconds | 14.575 milliseconds | 9.825 milliseconds | ok |
| asyncio_pool | 1 | 40 | 17.04 microseconds | 26.12 microseconds | 49.00 microseconds | 58.29 microseconds | ok |
| asyncio_pool | 4 | 40 | 19.30 microseconds | 249.31 microseconds | 55.20 microseconds | 304.58 microseconds | ok |
| asyncio_pool | 1 | 10000 | 26.82 microseconds | 41.45 microseconds | 9.214 milliseconds | 93.47 microseconds | ok |
| asyncio_pool | 4 | 10000 | 23.10 microseconds | 12.397 milliseconds | 9.982 milliseconds | 10.142 milliseconds | ok |
| bare_thread | 1 | 40 | 146.55 microseconds | -77.58 microseconds | 52.32 microseconds | 32.58 microseconds | ok |
| bare_thread | 4 | 40 | 155.98 microseconds | -75.15 microseconds | 51.04 microseconds | 317.86 microseconds | ok |
| bare_thread | 1 | 10000 | 5.184 milliseconds | -5093.18 microseconds | 9.592 milliseconds | 41.48 microseconds | ok |
| bare_thread | 4 | 10000 | 5.212 milliseconds | -5085.74 microseconds | 14.658 milliseconds | 9.873 milliseconds | ok |
| process_pool | 1 | 40 | - | - | - | - | not_measurable: the body's start/end timestamps are unavailable - it did not run in this process (cross-process units cannot be split this way) |
| process_pool | 4 | 40 | - | - | - | - | not_measurable: the body's start/end timestamps are unavailable - it did not run in this process (cross-process units cannot be split this way) |
| process_pool | 1 | 10000 | - | - | - | - | not_measurable: the body's start/end timestamps are unavailable - it did not run in this process (cross-process units cannot be split this way) |
| process_pool | 4 | 10000 | - | - | - | - | not_measurable: the body's start/end timestamps are unavailable - it did not run in this process (cross-process units cannot be split this way) |
| thread_pool | 1 | 40 | 8.16 microseconds | 10.40 microseconds | 43.71 microseconds | 14.95 microseconds | ok |
| thread_pool | 4 | 40 | 11.37 microseconds | 174.81 microseconds | 47.72 microseconds | 141.19 microseconds | ok |
| thread_pool | 1 | 10000 | 13.01 microseconds | 17.56 microseconds | 9.336 milliseconds | 28.64 microseconds | ok |
| thread_pool | 4 | 10000 | 13.47 microseconds | 10.148 milliseconds | 15.055 milliseconds | 5.076 milliseconds | ok |
| zoo | 1 | 40 | 52.47 microseconds | 24.98 microseconds | 52.22 microseconds | 59.61 microseconds | ok |
| zoo | 4 | 40 | 62.36 microseconds | 303.76 microseconds | 46.66 microseconds | 275.39 microseconds | ok |
| zoo | 1 | 10000 | 92.37 microseconds | 35.80 microseconds | 9.700 milliseconds | 117.86 microseconds | ok |
| zoo | 4 | 10000 | 97.22 microseconds | 9.968 milliseconds | 14.989 milliseconds | 9.553 milliseconds | ok |

### where the excess over each baseline falls

| baseline | concurrency | tier (microseconds) | excess per task | mostly on | gap in that segment |
|---|---|---|---|---|---|
| apscheduler | 1 | 40 | -114.49 microseconds | return | 15.80 microseconds |
| asyncio_pool | 1 | 40 | 39.74 microseconds | submit side | 35.44 microseconds |
| bare_thread | 1 | 40 | 34.64 microseconds | handoff | 102.56 microseconds |
| thread_pool | 1 | 40 | 113.52 microseconds | return | 44.66 microseconds |
| apscheduler | 4 | 40 | -235.45 microseconds | return | 54.05 microseconds |
| asyncio_pool | 4 | 40 | 23.70 microseconds | handoff | 54.46 microseconds |
| bare_thread | 4 | 40 | 223.28 microseconds | handoff | 378.91 microseconds |
| thread_pool | 4 | 40 | 304.11 microseconds | return | 134.20 microseconds |
| apscheduler | 1 | 10000 | -54.37 microseconds | body | 61.78 microseconds |
| asyncio_pool | 1 | 10000 | 565.33 microseconds | body | 486.14 microseconds |
| bare_thread | 1 | 10000 | 229.49 microseconds | handoff | 5.129 milliseconds |
| thread_pool | 1 | 10000 | 541.75 microseconds | body | 364.01 microseconds |
| apscheduler | 4 | 10000 | -424.87 microseconds | body | 414.07 microseconds |
| asyncio_pool | 4 | 10000 | -419.34 microseconds | body | 5.007 milliseconds |
| bare_thread | 4 | 10000 | 8.528 milliseconds | handoff | 15.053 milliseconds |
| thread_pool | 4 | 10000 | 1.09 microseconds | return | 4.477 milliseconds |

> instrumentation seams used (temporarily wrapped on the harness side, no framework code changed): BaseWaiter._dispatch_worker (handing a worker to the scheduling model); WorkerDispatchCore.resolve_period / resolve_phase / resolve_run_timeout (per-round policy lookups); BaseWaiter.execute_service (one scheduling round)

> **the cost of instrumentation itself** (on the framework-under-test side, paired from two rounds in the same subprocess, so it can be discounted honestly): zoo 40 microseconds / concurrency 1: without instrumentation 186.64 microseconds vs with 191.03 microseconds, a difference of 4.39 microseconds; zoo 40 microseconds / concurrency 4: without instrumentation 671.61 microseconds vs with 673.41 microseconds, a difference of 1.79 microseconds; zoo 10000 microseconds / concurrency 1: without instrumentation 9.915 milliseconds vs with 9.953 milliseconds, a difference of 37.32 microseconds; zoo 10000 microseconds / concurrency 4: without instrumentation 39.693 milliseconds vs with 39.568 milliseconds, a difference of 125.02 microseconds (faster with instrumentation: this round's machine drift exceeded the instrumentation cost)

> **additivity of the sub-measurements**: the sub-measurements are **independent measurements**: each one's duration contains its own preemption waits, so their sum can exceed the submit side they all live in (about 3% measured). Do not read them as an additive partition.measured overlap this round: zoo 10000 microseconds / concurrency 1 overlaps 11.8%; zoo 10000 microseconds / concurrency 4 overlaps 93.5%

### what the submit side of the framework under test consists of

- concurrency 1, tier 40 microseconds: rest of the scheduling round (due checks, locks, list upkeep) 0.05 microseconds; dispatch (handed to the scheduling model, incl. pool enqueue) 22.43 microseconds; per-round policy lookups (period / phase / timeout each read config) 20.74 microseconds; submit side, other (the adapter's own bookkeeping and completion-signal state) 8.43 microseconds
- concurrency 4, tier 40 microseconds: rest of the scheduling round (due checks, locks, list upkeep) 0.00 microseconds; dispatch (handed to the scheduling model, incl. pool enqueue) 31.12 microseconds; per-round policy lookups (period / phase / timeout each read config) 28.63 microseconds; submit side, other (the adapter's own bookkeeping and completion-signal state) 4.32 microseconds
- concurrency 1, tier 10000 microseconds: rest of the scheduling round (due checks, locks, list upkeep) 1.26 microseconds; dispatch (handed to the scheduling model, incl. pool enqueue) 37.35 microseconds; per-round policy lookups (period / phase / timeout each read config) 37.74 microseconds; submit side, other (the adapter's own bookkeeping and completion-signal state) 15.28 microseconds
- concurrency 4, tier 10000 microseconds: rest of the scheduling round (due checks, locks, list upkeep) 0.00 microseconds; dispatch (handed to the scheduling model, incl. pool enqueue) 43.77 microseconds; per-round policy lookups (period / phase / timeout each read config) 54.88 microseconds; submit side, other (the adapter's own bookkeeping and completion-signal state) 5.48 microseconds

### Attribution conclusions

- At concurrency 1 and a 40 microsecond body, the framework under test is not slower overall than apscheduler (-114 microseconds per task): the two sides have different timing structures, and the per-segment differences have both signs, so no 'which segment' is given.
- At concurrency 1 and a 40 microsecond body, the framework under test spends 40 microseconds more per task than asyncio_pool, most of it on 'submit side' (+35 microseconds).
-   note: the per-segment differences in this group do not all share a sign  - 'handoff' is -1 microseconds, so read them together.
- At concurrency 1 and a 40 microsecond body, the framework under test spends 35 microseconds more per task than bare_thread, most of it on 'handoff' (+103 microseconds).
-   note: the per-segment differences in this group do not all share a sign  - 'submit side' is -94 microseconds, so read them together.
- At concurrency 1 and a 40 microsecond body, the framework under test spends 114 microseconds more per task than thread_pool, most of it on 'return' (+45 microseconds).
- In this group the framework's submit side consists of (microseconds): rest of the scheduling round (due checks, locks, list upkeep) 0, dispatch (handed to the scheduling model, incl. pool enqueue) 22, per-round policy lookups (period / phase / timeout each read config) 21, submit side, other (the adapter's own bookkeeping and completion-signal state) 8.
- At concurrency 4 and a 40 microsecond body, the framework under test is not slower overall than apscheduler (-235 microseconds per task): the two sides have different timing structures, and the per-segment differences have both signs, so no 'which segment' is given.
- At concurrency 4 and a 40 microsecond body, the framework under test spends 24 microseconds more per task than asyncio_pool, most of it on 'handoff' (+54 microseconds).
-   note: the per-segment differences in this group do not all share a sign  - 'return' is -29 microseconds, so read them together.
- At concurrency 4 and a 40 microsecond body, the framework under test spends 223 microseconds more per task than bare_thread, most of it on 'handoff' (+379 microseconds).
-   note: the per-segment differences in this group do not all share a sign  - 'submit side' is -94 microseconds, so read them together.
- At concurrency 4 and a 40 microsecond body, the framework under test spends 304 microseconds more per task than thread_pool, most of it on 'return' (+134 microseconds).
-   note: the per-segment differences in this group do not all share a sign  - 'body' is -1 microseconds, so read them together.
- In this group the framework's submit side consists of (microseconds): rest of the scheduling round (due checks, locks, list upkeep) 0, dispatch (handed to the scheduling model, incl. pool enqueue) 31, per-round policy lookups (period / phase / timeout each read config) 29, submit side, other (the adapter's own bookkeeping and completion-signal state) 4.
- At concurrency 1 and a 10000 microsecond body, the framework under test is not slower overall than apscheduler (-54 microseconds per task): the two sides have different timing structures, and the per-segment differences have both signs, so no 'which segment' is given.
- At concurrency 1 and a 10000 microsecond body, the framework under test spends 565 microseconds more per task than asyncio_pool, most of it on 'body' (+486 microseconds).
-   note: the per-segment differences in this group do not all share a sign  - 'handoff' is -6 microseconds, so read them together.
- At concurrency 1 and a 10000 microsecond body, the framework under test spends 229 microseconds more per task than bare_thread, most of it on 'handoff' (+5129 microseconds).
-   note: the per-segment differences in this group do not all share a sign  - 'submit side' is -5092 microseconds, so read them together.
- At concurrency 1 and a 10000 microsecond body, the framework under test spends 542 microseconds more per task than thread_pool, most of it on 'body' (+364 microseconds).
- In this group the framework's submit side consists of (microseconds): rest of the scheduling round (due checks, locks, list upkeep) 1, dispatch (handed to the scheduling model, incl. pool enqueue) 37, per-round policy lookups (period / phase / timeout each read config) 38, submit side, other (the adapter's own bookkeeping and completion-signal state) 15.
- At concurrency 4 and a 10000 microsecond body, the framework under test is not slower overall than apscheduler (-425 microseconds per task): the two sides have different timing structures, and the per-segment differences have both signs, so no 'which segment' is given.
- At concurrency 4 and a 10000 microsecond body, the framework under test is not slower overall than asyncio_pool (-419 microseconds per task): the two sides have different timing structures, and the per-segment differences have both signs, so no 'which segment' is given.
- At concurrency 4 and a 10000 microsecond body, the framework under test spends 8528 microseconds more per task than bare_thread, most of it on 'handoff' (+15053 microseconds).
-   note: the per-segment differences in this group do not all share a sign  - 'submit side' is -5114 microseconds, so read them together.
- At concurrency 4 and a 10000 microsecond body, the framework under test spends 1 microseconds more per task than thread_pool, most of it on 'return' (+4477 microseconds).
-   note: the per-segment differences in this group do not all share a sign  - 'handoff' is -180 microseconds, so read them together.
- In this group the framework's submit side consists of (microseconds): rest of the scheduling round (due checks, locks, list upkeep) 0, dispatch (handed to the scheduling model, incl. pool enqueue) 44, per-round policy lookups (period / phase / timeout each read config) 55, submit side, other (the adapter's own bookkeeping and completion-signal state) 5.

## Dimension: The cost of scheduling semantics

> scope: this dimension uses an **internal switch comparison** (same workload, semantics all off vs. enabled one by one) and **does not compare across schemes**  -  a bare writeup has no counterpart for priority / timeout / retry, so it can only answer 'what is this framework's semantics worth', not 'how much more expensive than a bare writeup'

**status: ok** (the conclusion was reached on the 'current' drive-surface generation installed during this measurement)

| semantics item | layer | how probed | measurable | evidence / reason |
|---|---|---|---|---|
| timeout | dispatch layer | behavioral probe (driven repeatedly through scheduling rounds; measures the moment the worker leaves the in-flight table) | yes | timeout reap observed: the worker left the in-flight table before the body ended (0.0123s) (body 0.15s, run_timeout 0.01s); judgment action at dispatch_core.py:427; what is in effect is observation and circuit-breaking (recording, marking unhealthy, removing the in-flight registration, stopping dispatch), **not termination** of the still-running worker - CPython cannot safely interrupt a running thread |
| priority | event layer | structural probe (checks whether the symbols required for execution exist) | no | not re-checked: this item's conclusion was reached on the 'previous' generation; taking it as this generation's fact asserts wiring in the report that the current generation does not have. Re-deriving it needs a probe design of its own (each item needs clean on/off states before it can be priced) |
| retry | event layer | structural probe (checks whether the symbols required for execution exist) | no | not re-checked: this item's conclusion was reached on the 'previous' generation; taking it as this generation's fact asserts wiring in the report that the current generation does not have. Re-deriving it needs a probe design of its own (each item needs clean on/off states before it can be priced) |

> this conclusion was reached on the version under test and its coordinates hold for that version only; after appending a version to matrix.yaml, re-run this probe

## Dimension: Native task line: three execution forms of the same subject

> scope: this line is measured on its own: it enters neither the main matrix's pivot tables nor its crossings, and its numbers must not be read against the baseline adapters listed there

**verdict: not_applicable**

the framework under test has no such capability (symbols not found: adaptive_params, dual_arm_worker, native_params, native_task_worker) - this line does not apply to this version, and that is not a failure

equivalence gate: not run (the probe did not reach it)

> loaded task modbus_rtu.parse_response - Modbus RTU response-frame parsing: the input is a bare RTU ADU response frame (no transport framing), and the tier's frame bytes are its length

> tiers R1, R8, R125 (frame bytes 7, 21, 255); concurrencies 1, 4, 16, 64; each cell runs 100 warm-up plus 100 measured rounds, one child process per concurrency (4 in all)

## tiers where the framework under test is faster

> Tiers where the framework under test is ahead (the two sides differ by more than that tier's band). This comes from the **same within-run comparison** as the openly published unfavourable data  -  one side takes the faster tiers, the other the slower ones, so both cover the same ground and neither can pick its own tiers. **This section being empty is not a failure**: it should be empty when nothing is distinguishable or when the framework loses everywhere.

| baseline | concurrency | body tier (microseconds) | gap | median of the framework under test |
|---|---|---|---|---|
| apscheduler | 1 | 40 | the framework under test is 1.83x faster than apscheduler | 157.73 microseconds |
| process_pool | 1 | 40 | the framework under test is 2.11x faster than process_pool | 157.73 microseconds |
| apscheduler | 1 | 300 | the framework under test is 1.29x faster than apscheduler | 424.91 microseconds |
| process_pool | 1 | 300 | the framework under test is 1.43x faster than process_pool | 424.91 microseconds |
| apscheduler | 1 | 2700 | the framework under test is 1.12x faster than apscheduler | 2.797 milliseconds |
| process_pool | 1 | 2700 | the framework under test is 1.08x faster than process_pool | 2.797 milliseconds |
| apscheduler | 4 | 40 | the framework under test is 1.38x faster than apscheduler | 183.23 microseconds |
| apscheduler | 4 | 300 | the framework under test is 1.18x faster than apscheduler | 439.01 microseconds |
| apscheduler | 4 | 2700 | the framework under test is 1.07x faster than apscheduler | 2.848 milliseconds |
| apscheduler | 16 | 40 | the framework under test is 1.47x faster than apscheduler | 170.72 microseconds |
| apscheduler | 16 | 300 | the framework under test is 1.14x faster than apscheduler | 439.04 microseconds |
| apscheduler | 64 | 40 | the framework under test is 1.22x faster than apscheduler | 184.98 microseconds |

## Openly published unfavourable data

> Tiers where the framework under test is behind (the two sides differ by more than that tier's band). **A report missing this part is not acceptable**  -  a report that only shows where it wins gives the reader every reason to distrust it. Tiers whose difference is smaller than the band are not here but in the 'no winner' section: that is the verdict's outcome, not a place to hide things.

| baseline | concurrency | body tier (microseconds) | gap | median of the framework under test |
|---|---|---|---|---|
| thread_pool | 1 | 40 | thread_pool is 1.89x faster than zoo | 157.73 microseconds |
| thread_pool | 1 | 300 | thread_pool is 1.31x faster than zoo | 424.91 microseconds |
| asyncio_pool | 1 | 2700 | asyncio_pool is 1.07x faster than zoo | 2.797 milliseconds |
| bare_thread | 1 | 2700 | bare_thread is 1.03x faster than zoo | 2.797 milliseconds |
| thread_pool | 1 | 2700 | thread_pool is 1.05x faster than zoo | 2.797 milliseconds |
| thread_pool | 1 | 10000 | thread_pool is 1.03x faster than zoo | 9.931 milliseconds |
| thread_pool | 4 | 40 | thread_pool is 1.93x faster than zoo | 183.23 microseconds |
| asyncio_pool | 4 | 300 | asyncio_pool is 1.07x faster than zoo | 439.01 microseconds |
| bare_thread | 4 | 300 | bare_thread is 1.07x faster than zoo | 439.01 microseconds |
| process_pool | 4 | 300 | process_pool is 1.49x faster than zoo | 439.01 microseconds |
| thread_pool | 4 | 300 | thread_pool is 1.21x faster than zoo | 439.01 microseconds |
| asyncio_pool | 4 | 2700 | asyncio_pool is 1.04x faster than zoo | 2.848 milliseconds |
| bare_thread | 4 | 2700 | bare_thread is 1.07x faster than zoo | 2.848 milliseconds |
| process_pool | 4 | 2700 | process_pool is 2.04x faster than zoo | 2.848 milliseconds |
| thread_pool | 4 | 2700 | thread_pool is 1.05x faster than zoo | 2.848 milliseconds |
| process_pool | 4 | 10000 | process_pool is 2.19x faster than zoo | 10.050 milliseconds |
| thread_pool | 4 | 10000 | thread_pool is 1.03x faster than zoo | 10.050 milliseconds |
| thread_pool | 16 | 40 | thread_pool is 1.89x faster than zoo | 170.72 microseconds |
| asyncio_pool | 16 | 300 | asyncio_pool is 1.11x faster than zoo | 439.04 microseconds |
| process_pool | 16 | 300 | process_pool is 1.57x faster than zoo | 439.04 microseconds |
| thread_pool | 16 | 300 | thread_pool is 1.25x faster than zoo | 439.04 microseconds |
| asyncio_pool | 16 | 2700 | asyncio_pool is 1.04x faster than zoo | 2.926 milliseconds |
| bare_thread | 16 | 2700 | bare_thread is 1.07x faster than zoo | 2.926 milliseconds |
| process_pool | 16 | 2700 | process_pool is 2.10x faster than zoo | 2.926 milliseconds |
| thread_pool | 16 | 2700 | thread_pool is 1.08x faster than zoo | 2.926 milliseconds |
| asyncio_pool | 16 | 10000 | asyncio_pool is 1.05x faster than zoo | 10.507 milliseconds |
| bare_thread | 16 | 10000 | bare_thread is 1.07x faster than zoo | 10.507 milliseconds |
| process_pool | 16 | 10000 | process_pool is 2.23x faster than zoo | 10.507 milliseconds |
| thread_pool | 16 | 10000 | thread_pool is 1.05x faster than zoo | 10.507 milliseconds |
| thread_pool | 64 | 40 | thread_pool is 2.07x faster than zoo | 184.98 microseconds |
| asyncio_pool | 64 | 300 | asyncio_pool is 1.26x faster than zoo | 472.05 microseconds |
| bare_thread | 64 | 300 | bare_thread is 1.11x faster than zoo | 472.05 microseconds |
| process_pool | 64 | 300 | process_pool is 1.59x faster than zoo | 472.05 microseconds |
| thread_pool | 64 | 300 | thread_pool is 1.35x faster than zoo | 472.05 microseconds |
| asyncio_pool | 64 | 2700 | asyncio_pool is 1.06x faster than zoo | 2.937 milliseconds |
| bare_thread | 64 | 2700 | bare_thread is 1.08x faster than zoo | 2.937 milliseconds |
| process_pool | 64 | 2700 | process_pool is 2.07x faster than zoo | 2.937 milliseconds |
| thread_pool | 64 | 2700 | thread_pool is 1.06x faster than zoo | 2.937 milliseconds |
| bare_thread | 64 | 10000 | bare_thread is 1.07x faster than zoo | 10.515 milliseconds |
| process_pool | 64 | 10000 | process_pool is 2.23x faster than zoo | 10.515 milliseconds |
| thread_pool | 64 | 10000 | thread_pool is 1.05x faster than zoo | 10.515 milliseconds |

## tiers with no winner

> Tiers where the two sides **cannot be told apart**: the difference is smaller than that tier's band. The band comes from the two sides' own across-round spread (stdev / median of the per-task medians) **within the same run**, so it varies with the sample and the reader can recompute it from the archived per-round samples. It also covers every measured tier and states a reason for each  -  **this section is not a place to hide things**: reporting noise-level differences as '1.00x faster' is.

> criterion: band = the framework's own across-round relative spread plus the baseline's; a single side's across-round relative spread = stdev / median of that unit's per-round per-task medians

band this run: 1.0% to 33.3% (it varies with each tier's spread).

| baseline | concurrency | body tier (microseconds) | ratio | reason |
|---|---|---|---|---|
| asyncio_pool | 1 | 40 | 0.89x | the two sides differ by 10.9%, less than the band of 19.3%  -  no winner |
| bare_thread | 1 | 40 | 0.90x | the two sides differ by 10.1%, less than the band of 16.8%  -  no winner |
| asyncio_pool | 1 | 300 | 0.95x | the two sides differ by 4.8%, less than the band of 9.9%  -  no winner |
| bare_thread | 1 | 300 | 0.95x | the two sides differ by 4.6%, less than the band of 9.8%  -  no winner |
| apscheduler | 1 | 10000 | 1.00x | the two sides differ by 0.4%, less than the band of 1.3%  -  no winner |
| asyncio_pool | 1 | 10000 | 0.95x | the two sides differ by 4.6%, less than the band of 6.3%  -  no winner |
| bare_thread | 1 | 10000 | 1.00x | the two sides differ by 0.4%, less than the band of 1.0%  -  no winner |
| process_pool | 1 | 10000 | 1.01x | the two sides differ by 0.8%, less than the band of 1.3%  -  no winner |
| asyncio_pool | 4 | 40 | 0.87x | the two sides differ by 12.8%, less than the band of 17.9%  -  no winner |
| bare_thread | 4 | 40 | 0.88x | the two sides differ by 11.5%, less than the band of 13.1%  -  no winner |
| process_pool | 4 | 40 | 1.03x | the two sides differ by 2.9%, less than the band of 26.2%  -  no winner |
| apscheduler | 4 | 10000 | 1.00x | the two sides differ by 0.3%, less than the band of 2.1%  -  no winner |
| asyncio_pool | 4 | 10000 | 0.99x | the two sides differ by 0.7%, less than the band of 1.6%  -  no winner |
| bare_thread | 4 | 10000 | 0.97x | the two sides differ by 2.6%, less than the band of 3.9%  -  no winner |
| asyncio_pool | 16 | 40 | 0.81x | the two sides differ by 19.1%, less than the band of 22.7%  -  no winner |
| bare_thread | 16 | 40 | 1.02x | the two sides differ by 2.2%, less than the band of 18.1%  -  no winner |
| process_pool | 16 | 40 | 0.94x | the two sides differ by 5.8%, less than the band of 33.0%  -  no winner |
| bare_thread | 16 | 300 | 0.98x | the two sides differ by 1.9%, less than the band of 6.6%  -  no winner |
| apscheduler | 16 | 2700 | 1.01x | the two sides differ by 0.9%, less than the band of 2.0%  -  no winner |
| apscheduler | 16 | 10000 | 0.98x | the two sides differ by 2.2%, less than the band of 4.4%  -  no winner |
| asyncio_pool | 64 | 40 | 0.67x | the two sides differ by 33.0%, less than the band of 33.3%  -  no winner |
| bare_thread | 64 | 40 | 0.92x | the two sides differ by 8.4%, less than the band of 12.7%  -  no winner |
| process_pool | 64 | 40 | 1.06x | the two sides differ by 5.8%, less than the band of 22.3%  -  no winner |
| apscheduler | 64 | 300 | 1.07x | the two sides differ by 7.1%, less than the band of 8.4%  -  no winner |
| apscheduler | 64 | 2700 | 1.01x | the two sides differ by 1.3%, less than the band of 4.2%  -  no winner |
| apscheduler | 64 | 10000 | 0.98x | the two sides differ by 1.9%, less than the band of 5.1%  -  no winner |
| asyncio_pool | 64 | 10000 | 0.99x | the two sides differ by 0.6%, less than the band of 4.5%  -  no winner |

## caliber limitations and sources of bias

- **measured layer**: dispatch layer: drives BaseWaiter.execute_service() directly (thread-pool mode) instead of Master's full run loop - the latter's service loop fires once per second, which would lift the latency scale from microseconds to seconds. **One scheduling round per submitted task**, so that whole round's bookkeeping is charged to this single task; Master's usage serves several workers per round, which would spread that cost over tasks
- **scope note** (apscheduler): APScheduler's scheduled-job model, including job registration and trigger-decision cost; the library swallows exceptions raised inside jobs, so this tier re-raises them during drain. **Caliber note**: the library offers no job-completion notification, so the completion signal comes from this tier's Condition, and the cost of one notification per task counts toward its end-to-end
- **scope note** (asyncio_pool): the asyncio event loop dispatches **synchronous** bodies (equivalent to loop.run_in_executor), not native coroutine concurrency; the difference from a thread pool is the event loop's own cost
- **scope note** (bare_thread): one threading.Thread per task, no pooling and no reuse; thread creation cost counts toward end-to-end. Bodies' self-reported timings are collected with a lock-free list.append (atomic under the GIL) - **instrumentation must not charge one tier a cost other tiers do not pay**, or the comparison becomes a comparison of instrumentation
- **scope note** (process_pool): ProcessPoolExecutor: pickling the body and its return value across processes counts toward end-to-end, so this tier is not directly comparable to the multi-threaded tiers
- **scope note** (thread_pool): concurrent.futures.ThreadPoolExecutor, pool size = concurrency; threads are reused, with no task-queue semantics
- **scope note** (zoo): end-to-end stops at the worker completing and being deregistered from the in-flight table; the framework's later result delivery splits by generation: the previous generation's delivery path is broken (dispatch looks reactors up by reactor_name rather than by topic, so every call raises), and the completion signal can only come from the framework's internal callback; the current generation has that fixed, and the completion signal comes from the framework's own result reactor. **Caliber note**: the completion signal comes from this tier's Condition, and the cost of one notification per task counts toward its end-to-end; a bias in that direction makes the framework under test look slower, i.e. it works against it rather than for it
- **groups with no overhead figure**: for the (adapter, concurrency) groups below, concurrency exceeds the parallel capacity of the machine this run was on: the body's self-reported duration contains scheduling waits from oversubscription, which is not the same thing as 'wall clock / concurrency', so subtracting them is meaningless (measured, it produced negative values and meaningless change factors). This report therefore **gives them no framework overhead and no crossing point**; end-to-end, throughput and the body's self-reported values remain valid and not one field has been dropped from the raw archive. The criterion is the logical core count from the environment self-report, plus one evidential backstop: a computed negative overhead is likewise treated as unreadable  -  a framework can only add time, never remove it
- **native task line: no verdict**: the native task line came out not_applicable (the framework under test has no such capability (symbols not found: adaptive_params, dual_arm_worker, native_params, native_task_worker) - this line does not apply to this version, and that is not a failure) and produced no figures. It is reported here rather than folded into the self-check: one line's failure must not hold back the other versions' data or the main matrix
- **self-check items excluded from the verdict**: for these units the body-tier deviation did not take part in the verdict (concurrency above the lowest, or a tier so short that wall clock is dominated by scheduling granularity). The raw values are still in place, but the wall clock at those two places is not enough to conclude whether the calibration is right  -  **factor this in when reading those tiers' numbers**

> absolute durations: valid only within a single run; do not compare across runs

## bring your own baseline

the baselines of this harness are **open**: implement the four methods of `zoo_bench.adapters.BaseAdapter` (`setup` / `submit` / `drain` / `teardown`), register with `zoo_bench.adapters.registry.register`, or list your module path in `extra_modules` of `matrix.yaml`  -  **no file in this repository needs to change**. Your implementation goes through the same equivalence verification (submit N bodies, assert all ran and each exactly once); failing it means the data is not trusted. **We deliberately write this path into the report**: a report written by the framework's own maintainer, measuring that framework, gives every reason to suspect a slowed baseline or a chosen caliber; letting anyone submit their own baseline is what makes this report challengeable and falsifiable.

## self-check

- overall measurement self-check: passed
- process isolation: {'parent_pid': 56883, 'child_pids': [57338, 57341, 57344, 57347, 57350, 57356, 57362, 57368, 57375, 57393, 57411, 57429, 57447, 57505, 57563, 57620, 57688, 57690, 57692, 57694, 57696, 57701, 57706, 57711, 57716, 57733, 57750, 57768, 57789, 57849, 57908, 57973, 58040, 58074, 58108, 58142, 58176, 58309, 58442, 58575, 58708, 59237, 59766, 60296, 60825, 62938, 65051, 67164, 69280, 69284, 69288, 69292, 69296, 69303, 69310, 69317, 69324, 69343, 69362, 69381, 69400, 69467, 69535, 69603, 69670, 69672, 69674, 69676, 69678, 69683, 69688, 69693, 69698, 69715, 69732, 69749, 69766, 69822, 69879, 69940, 70005, 70007, 70009, 70011, 70013, 70018, 70023, 70028, 70033, 70050, 70067, 70084, 70102, 70167, 70232, 70297], 'all_distinct': True, 'none_equals_parent': True}

## Appendix: all values

> the body carries only pivot tables for readability; this appendix gives **every field as measured before any trimming** per unit (percentiles, spread, measured body durations, ...). **Not one value is dropped**: the details live inside the report body itself, not behind links, folds or attachments  -  'verifiable' that requires a separate lookup is not verifiable

### Appendix: Latency percentiles and jitter

| scheme | concurrency | body tier (microseconds) | median | p95 | p99 | relative spread |
|---|---|---|---|---|---|---|
| apscheduler | 1 | 40 | 288.55 microseconds | 329.21 microseconds | 348.86 microseconds | 0.070 |
| apscheduler | 1 | 300 | 547.57 microseconds | 582.23 microseconds | 591.17 microseconds | 0.031 |
| apscheduler | 1 | 2700 | 3.120 milliseconds | 3.300 milliseconds | 3.323 milliseconds | 0.028 |
| apscheduler | 1 | 10000 | 9.890 milliseconds | 10.004 milliseconds | 10.068 milliseconds | 0.007 |
| apscheduler | 4 | 40 | 253.08 microseconds | 265.53 microseconds | 317.11 microseconds | 0.063 |
| apscheduler | 4 | 300 | 519.95 microseconds | 547.62 microseconds | 560.45 microseconds | 0.024 |
| apscheduler | 4 | 2700 | 3.035 milliseconds | 3.090 milliseconds | 3.115 milliseconds | 0.013 |
| apscheduler | 4 | 10000 | 10.021 milliseconds | 10.167 milliseconds | 10.479 milliseconds | 0.013 |
| apscheduler | 16 | 40 | 251.40 microseconds | 266.50 microseconds | 270.78 microseconds | 0.033 |
| apscheduler | 16 | 300 | 501.69 microseconds | 532.97 microseconds | 561.46 microseconds | 0.034 |
| apscheduler | 16 | 2700 | 2.952 milliseconds | 3.018 milliseconds | 3.039 milliseconds | 0.011 |
| apscheduler | 16 | 10000 | 10.271 milliseconds | 10.573 milliseconds | 11.250 milliseconds | 0.024 |
| apscheduler | 64 | 40 | 224.83 microseconds | 252.58 microseconds | 268.72 microseconds | 0.064 |
| apscheduler | 64 | 300 | 505.68 microseconds | 591.84 microseconds | 596.80 microseconds | 0.064 |
| apscheduler | 64 | 2700 | 2.974 milliseconds | 3.060 milliseconds | 3.248 milliseconds | 0.023 |
| apscheduler | 64 | 10000 | 10.311 milliseconds | 10.844 milliseconds | 10.870 milliseconds | 0.027 |
| asyncio_pool | 1 | 40 | 140.57 microseconds | 171.45 microseconds | 177.08 microseconds | 0.092 |
| asyncio_pool | 1 | 300 | 404.67 microseconds | 456.89 microseconds | 485.11 microseconds | 0.061 |
| asyncio_pool | 1 | 2700 | 2.605 milliseconds | 2.662 milliseconds | 2.662 milliseconds | 0.013 |
| asyncio_pool | 1 | 10000 | 9.476 milliseconds | 9.716 milliseconds | 11.633 milliseconds | 0.057 |
| asyncio_pool | 4 | 40 | 159.80 microseconds | 183.66 microseconds | 186.21 microseconds | 0.098 |
| asyncio_pool | 4 | 300 | 410.40 microseconds | 426.74 microseconds | 442.25 microseconds | 0.026 |
| asyncio_pool | 4 | 2700 | 2.749 milliseconds | 2.791 milliseconds | 2.832 milliseconds | 0.010 |
| asyncio_pool | 4 | 10000 | 9.977 milliseconds | 10.058 milliseconds | 10.063 milliseconds | 0.009 |
| asyncio_pool | 16 | 40 | 138.08 microseconds | 155.96 microseconds | 163.23 microseconds | 0.073 |
| asyncio_pool | 16 | 300 | 394.14 microseconds | 402.93 microseconds | 421.98 microseconds | 0.021 |
| asyncio_pool | 16 | 2700 | 2.822 milliseconds | 2.867 milliseconds | 2.934 milliseconds | 0.012 |
| asyncio_pool | 16 | 10000 | 9.992 milliseconds | 10.044 milliseconds | 10.143 milliseconds | 0.005 |
| asyncio_pool | 64 | 40 | 123.94 microseconds | 210.21 microseconds | 217.01 microseconds | 0.225 |
| asyncio_pool | 64 | 300 | 373.95 microseconds | 381.84 microseconds | 408.52 microseconds | 0.024 |
| asyncio_pool | 64 | 2700 | 2.777 milliseconds | 2.813 milliseconds | 2.815 milliseconds | 0.008 |
| asyncio_pool | 64 | 10000 | 10.449 milliseconds | 10.665 milliseconds | 10.843 milliseconds | 0.021 |
| bare_thread | 1 | 40 | 141.79 microseconds | 164.71 microseconds | 170.32 microseconds | 0.067 |
| bare_thread | 1 | 300 | 405.40 microseconds | 449.66 microseconds | 458.71 microseconds | 0.060 |
| bare_thread | 1 | 2700 | 2.717 milliseconds | 2.750 milliseconds | 2.785 milliseconds | 0.009 |
| bare_thread | 1 | 10000 | 9.972 milliseconds | 10.057 milliseconds | 10.074 milliseconds | 0.004 |
| bare_thread | 4 | 40 | 162.08 microseconds | 176.75 microseconds | 185.28 microseconds | 0.049 |
| bare_thread | 4 | 300 | 409.00 microseconds | 423.72 microseconds | 429.34 microseconds | 0.017 |
| bare_thread | 4 | 2700 | 2.659 milliseconds | 2.692 milliseconds | 2.693 milliseconds | 0.005 |
| bare_thread | 4 | 10000 | 9.793 milliseconds | 10.400 milliseconds | 11.034 milliseconds | 0.031 |
| bare_thread | 16 | 40 | 174.45 microseconds | 182.12 microseconds | 188.43 microseconds | 0.027 |
| bare_thread | 16 | 300 | 430.66 microseconds | 440.99 microseconds | 442.80 microseconds | 0.014 |
| bare_thread | 16 | 2700 | 2.724 milliseconds | 2.770 milliseconds | 2.774 milliseconds | 0.008 |
| bare_thread | 16 | 10000 | 9.834 milliseconds | 9.899 milliseconds | 10.052 milliseconds | 0.006 |
| bare_thread | 64 | 40 | 169.50 microseconds | 176.72 microseconds | 178.09 microseconds | 0.019 |
| bare_thread | 64 | 300 | 424.17 microseconds | 430.10 microseconds | 432.47 microseconds | 0.009 |
| bare_thread | 64 | 2700 | 2.722 milliseconds | 2.741 milliseconds | 2.744 milliseconds | 0.005 |
| bare_thread | 64 | 10000 | 9.812 milliseconds | 9.871 milliseconds | 9.945 milliseconds | 0.004 |
| process_pool | 1 | 40 | 332.18 microseconds | 372.02 microseconds | 385.43 microseconds | 0.057 |
| process_pool | 1 | 300 | 606.50 microseconds | 644.08 microseconds | 651.86 microseconds | 0.035 |
| process_pool | 1 | 2700 | 3.017 milliseconds | 3.063 milliseconds | 3.187 milliseconds | 0.014 |
| process_pool | 1 | 10000 | 10.015 milliseconds | 10.125 milliseconds | 10.215 milliseconds | 0.007 |
| process_pool | 4 | 40 | 188.60 microseconds | 247.46 microseconds | 253.31 microseconds | 0.180 |
| process_pool | 4 | 300 | 295.22 microseconds | 373.91 microseconds | 401.46 microseconds | 0.112 |
| process_pool | 4 | 2700 | 1.398 milliseconds | 1.700 milliseconds | 2.011 milliseconds | 0.107 |
| process_pool | 4 | 10000 | 4.598 milliseconds | 5.151 milliseconds | 5.619 milliseconds | 0.056 |
| process_pool | 16 | 40 | 160.88 microseconds | 220.91 microseconds | 241.56 microseconds | 0.175 |
| process_pool | 16 | 300 | 280.12 microseconds | 321.58 microseconds | 336.26 microseconds | 0.083 |
| process_pool | 16 | 2700 | 1.395 milliseconds | 1.464 milliseconds | 1.476 milliseconds | 0.026 |
| process_pool | 16 | 10000 | 4.711 milliseconds | 4.815 milliseconds | 4.940 milliseconds | 0.015 |
| process_pool | 64 | 40 | 195.74 microseconds | 242.18 microseconds | 253.66 microseconds | 0.115 |
| process_pool | 64 | 300 | 297.44 microseconds | 327.42 microseconds | 331.38 microseconds | 0.056 |
| process_pool | 64 | 2700 | 1.417 milliseconds | 1.538 milliseconds | 1.615 milliseconds | 0.039 |
| process_pool | 64 | 10000 | 4.709 milliseconds | 4.749 milliseconds | 4.799 milliseconds | 0.005 |
| thread_pool | 1 | 40 | 83.56 microseconds | 99.22 microseconds | 104.74 microseconds | 0.104 |
| thread_pool | 1 | 300 | 323.16 microseconds | 358.34 microseconds | 360.98 microseconds | 0.048 |
| thread_pool | 1 | 2700 | 2.667 milliseconds | 2.692 milliseconds | 2.741 milliseconds | 0.008 |
| thread_pool | 1 | 10000 | 9.618 milliseconds | 9.832 milliseconds | 10.008 milliseconds | 0.011 |
| thread_pool | 4 | 40 | 94.96 microseconds | 107.50 microseconds | 111.37 microseconds | 0.072 |
| thread_pool | 4 | 300 | 362.52 microseconds | 370.91 microseconds | 372.59 microseconds | 0.019 |
| thread_pool | 4 | 2700 | 2.706 milliseconds | 2.753 milliseconds | 2.776 milliseconds | 0.013 |
| thread_pool | 4 | 10000 | 9.798 milliseconds | 9.901 milliseconds | 9.981 milliseconds | 0.008 |
| thread_pool | 16 | 40 | 90.52 microseconds | 104.07 microseconds | 112.32 microseconds | 0.070 |
| thread_pool | 16 | 300 | 351.18 microseconds | 363.24 microseconds | 369.43 microseconds | 0.020 |
| thread_pool | 16 | 2700 | 2.718 milliseconds | 2.758 milliseconds | 2.904 milliseconds | 0.019 |
| thread_pool | 16 | 10000 | 10.001 milliseconds | 10.427 milliseconds | 10.716 milliseconds | 0.019 |
| thread_pool | 64 | 40 | 89.55 microseconds | 91.29 microseconds | 91.83 microseconds | 0.017 |
| thread_pool | 64 | 300 | 349.61 microseconds | 355.29 microseconds | 356.04 microseconds | 0.008 |
| thread_pool | 64 | 2700 | 2.768 milliseconds | 2.814 milliseconds | 2.828 milliseconds | 0.007 |
| thread_pool | 64 | 10000 | 10.022 milliseconds | 10.175 milliseconds | 10.246 milliseconds | 0.014 |
| zoo | 1 | 40 | 157.73 microseconds | 190.47 microseconds | 212.68 microseconds | 0.101 |
| zoo | 1 | 300 | 424.91 microseconds | 455.46 microseconds | 458.23 microseconds | 0.038 |
| zoo | 1 | 2700 | 2.797 milliseconds | 2.822 milliseconds | 2.825 milliseconds | 0.006 |
| zoo | 1 | 10000 | 9.931 milliseconds | 10.029 milliseconds | 10.091 milliseconds | 0.006 |
| zoo | 4 | 40 | 183.23 microseconds | 208.27 microseconds | 212.35 microseconds | 0.081 |
| zoo | 4 | 300 | 439.01 microseconds | 459.84 microseconds | 497.54 microseconds | 0.038 |
| zoo | 4 | 2700 | 2.848 milliseconds | 2.913 milliseconds | 2.938 milliseconds | 0.012 |
| zoo | 4 | 10000 | 10.050 milliseconds | 10.172 milliseconds | 10.227 milliseconds | 0.008 |
| zoo | 16 | 40 | 170.72 microseconds | 225.48 microseconds | 232.07 microseconds | 0.154 |
| zoo | 16 | 300 | 439.04 microseconds | 483.56 microseconds | 497.31 microseconds | 0.052 |
| zoo | 16 | 2700 | 2.926 milliseconds | 2.968 milliseconds | 2.985 milliseconds | 0.010 |
| zoo | 16 | 10000 | 10.507 milliseconds | 10.987 milliseconds | 11.238 milliseconds | 0.020 |
| zoo | 64 | 40 | 184.98 microseconds | 222.63 microseconds | 225.59 microseconds | 0.108 |
| zoo | 64 | 300 | 472.05 microseconds | 487.34 microseconds | 496.03 microseconds | 0.020 |
| zoo | 64 | 2700 | 2.937 milliseconds | 3.067 milliseconds | 3.105 milliseconds | 0.019 |
| zoo | 64 | 10000 | 10.515 milliseconds | 11.025 milliseconds | 11.049 milliseconds | 0.024 |

### Appendix: Framework overhead share

| scheme | concurrency | body tier (microseconds) | framework overhead | overhead share | body as measured |
|---|---|---|---|---|---|
| apscheduler | 1 | 40 | 237.02 microseconds | 82.14% | 51.52 microseconds |
| apscheduler | 1 | 300 | 244.40 microseconds | 44.63% | 303.17 microseconds |
| apscheduler | 1 | 2700 | 304.02 microseconds | 9.74% | 2.816 milliseconds |
| apscheduler | 1 | 10000 | 408.51 microseconds | 4.13% | 9.482 milliseconds |
| apscheduler | 4 | 40 | 196.71 microseconds | 77.73% | 56.37 microseconds |
| apscheduler | 4 | 300 | 210.79 microseconds | 40.54% | 309.16 microseconds |
| apscheduler | 4 | 2700 | 293.59 microseconds | 9.67% | 2.741 milliseconds |
| apscheduler | 4 | 10000 | - | - | 14.823 milliseconds |
| apscheduler | 16 | 40 | - | - | 53.43 microseconds |
| apscheduler | 16 | 300 | - | - | 309.91 microseconds |
| apscheduler | 16 | 2700 | - | - | 2.713 milliseconds |
| apscheduler | 16 | 10000 | - | - | 20.657 milliseconds |
| apscheduler | 64 | 40 | - | - | 50.89 microseconds |
| apscheduler | 64 | 300 | - | - | 297.16 microseconds |
| apscheduler | 64 | 2700 | - | - | 2.718 milliseconds |
| apscheduler | 64 | 10000 | - | - | 20.896 milliseconds |
| asyncio_pool | 1 | 40 | 94.13 microseconds | 66.96% | 46.44 microseconds |
| asyncio_pool | 1 | 300 | 105.90 microseconds | 26.17% | 298.76 microseconds |
| asyncio_pool | 1 | 2700 | 117.31 microseconds | 4.50% | 2.487 milliseconds |
| asyncio_pool | 1 | 10000 | 133.41 microseconds | 1.41% | 9.342 milliseconds |
| asyncio_pool | 4 | 40 | 106.51 microseconds | 66.65% | 53.29 microseconds |
| asyncio_pool | 4 | 300 | 104.83 microseconds | 25.54% | 305.58 microseconds |
| asyncio_pool | 4 | 2700 | 128.74 microseconds | 4.68% | 2.620 milliseconds |
| asyncio_pool | 4 | 10000 | - | - | 14.654 milliseconds |
| asyncio_pool | 16 | 40 | - | - | 53.84 microseconds |
| asyncio_pool | 16 | 300 | - | - | 303.55 microseconds |
| asyncio_pool | 16 | 2700 | - | - | 2.698 milliseconds |
| asyncio_pool | 16 | 10000 | - | - | 24.790 milliseconds |
| asyncio_pool | 64 | 40 | - | - | 52.47 microseconds |
| asyncio_pool | 64 | 300 | - | - | 296.86 microseconds |
| asyncio_pool | 64 | 2700 | - | - | 2.643 milliseconds |
| asyncio_pool | 64 | 10000 | - | - | 86.213 milliseconds |
| bare_thread | 1 | 40 | 90.48 microseconds | 63.81% | 51.31 microseconds |
| bare_thread | 1 | 300 | 113.83 microseconds | 28.08% | 291.57 microseconds |
| bare_thread | 1 | 2700 | 114.16 microseconds | 4.20% | 2.603 milliseconds |
| bare_thread | 1 | 10000 | 125.91 microseconds | 1.26% | 9.846 milliseconds |
| bare_thread | 4 | 40 | 110.40 microseconds | 68.11% | 51.69 microseconds |
| bare_thread | 4 | 300 | 113.04 microseconds | 27.64% | 295.95 microseconds |
| bare_thread | 4 | 2700 | 114.79 microseconds | 4.32% | 2.545 milliseconds |
| bare_thread | 4 | 10000 | - | - | 14.784 milliseconds |
| bare_thread | 16 | 40 | - | - | 50.76 microseconds |
| bare_thread | 16 | 300 | - | - | 299.23 microseconds |
| bare_thread | 16 | 2700 | - | - | 2.594 milliseconds |
| bare_thread | 16 | 10000 | - | - | 14.902 milliseconds |
| bare_thread | 64 | 40 | - | - | 50.72 microseconds |
| bare_thread | 64 | 300 | - | - | 296.82 microseconds |
| bare_thread | 64 | 2700 | - | - | 2.584 milliseconds |
| bare_thread | 64 | 10000 | - | - | 14.905 milliseconds |
| process_pool | 1 | 40 | 275.31 microseconds | 82.88% | 56.87 microseconds |
| process_pool | 1 | 300 | 303.41 microseconds | 50.03% | 303.09 microseconds |
| process_pool | 1 | 2700 | 340.55 microseconds | 11.29% | 2.677 milliseconds |
| process_pool | 1 | 10000 | 364.82 microseconds | 3.64% | 9.650 milliseconds |
| process_pool | 4 | 40 | 93.61 microseconds | 49.63% | 94.99 microseconds |
| process_pool | 4 | 300 | - | - | 569.90 microseconds |
| process_pool | 4 | 2700 | - | - | 4.831 milliseconds |
| process_pool | 4 | 10000 | - | - | 17.613 milliseconds |
| process_pool | 16 | 40 | - | - | 90.15 microseconds |
| process_pool | 16 | 300 | - | - | 564.29 microseconds |
| process_pool | 16 | 2700 | - | - | 9.624 milliseconds |
| process_pool | 16 | 10000 | - | - | 53.594 milliseconds |
| process_pool | 64 | 40 | - | - | 95.71 microseconds |
| process_pool | 64 | 300 | - | - | 572.83 microseconds |
| process_pool | 64 | 2700 | - | - | 25.932 milliseconds |
| process_pool | 64 | 10000 | - | - | 173.835 milliseconds |
| thread_pool | 1 | 40 | 32.13 microseconds | 38.45% | 51.44 microseconds |
| thread_pool | 1 | 300 | 34.74 microseconds | 10.75% | 288.43 microseconds |
| thread_pool | 1 | 2700 | 48.01 microseconds | 1.80% | 2.619 milliseconds |
| thread_pool | 1 | 10000 | 57.70 microseconds | 0.60% | 9.560 milliseconds |
| thread_pool | 4 | 40 | 45.01 microseconds | 47.40% | 49.95 microseconds |
| thread_pool | 4 | 300 | 57.72 microseconds | 15.92% | 304.79 microseconds |
| thread_pool | 4 | 2700 | 85.16 microseconds | 3.15% | 2.620 milliseconds |
| thread_pool | 4 | 10000 | - | - | 19.499 milliseconds |
| thread_pool | 16 | 40 | - | - | 48.21 microseconds |
| thread_pool | 16 | 300 | - | - | 297.14 microseconds |
| thread_pool | 16 | 2700 | - | - | 2.632 milliseconds |
| thread_pool | 16 | 10000 | - | - | 20.219 milliseconds |
| thread_pool | 64 | 40 | - | - | 47.21 microseconds |
| thread_pool | 64 | 300 | - | - | 293.01 microseconds |
| thread_pool | 64 | 2700 | - | - | 2.681 milliseconds |
| thread_pool | 64 | 10000 | - | - | 25.119 milliseconds |
| zoo | 1 | 40 | 107.14 microseconds | 67.93% | 50.59 microseconds |
| zoo | 1 | 300 | 121.39 microseconds | 28.57% | 303.52 microseconds |
| zoo | 1 | 2700 | 144.58 microseconds | 5.17% | 2.652 milliseconds |
| zoo | 1 | 10000 | 179.89 microseconds | 1.81% | 9.751 milliseconds |
| zoo | 4 | 40 | 133.53 microseconds | 72.88% | 49.70 microseconds |
| zoo | 4 | 300 | 139.58 microseconds | 31.79% | 299.43 microseconds |
| zoo | 4 | 2700 | 150.03 microseconds | 5.27% | 2.698 milliseconds |
| zoo | 4 | 10000 | - | - | 15.037 milliseconds |
| zoo | 16 | 40 | - | - | 44.68 microseconds |
| zoo | 16 | 300 | - | - | 301.40 microseconds |
| zoo | 16 | 2700 | - | - | 2.756 milliseconds |
| zoo | 16 | 10000 | - | - | 51.225 milliseconds |
| zoo | 64 | 40 | - | - | 47.40 microseconds |
| zoo | 64 | 300 | - | - | 301.28 microseconds |
| zoo | 64 | 2700 | - | - | 2.753 milliseconds |
| zoo | 64 | 10000 | - | - | 82.491 milliseconds |

### Appendix: Throughput and concurrency scaling

| scheme | concurrency | body tier (microseconds) | throughput (tasks/s) |
|---|---|---|---|
| apscheduler | 1 | 40 | 3465.7 |
| apscheduler | 1 | 300 | 1826.2 |
| apscheduler | 1 | 2700 | 320.5 |
| apscheduler | 1 | 10000 | 101.1 |
| apscheduler | 4 | 40 | 15805.1 |
| apscheduler | 4 | 300 | 7693.1 |
| apscheduler | 4 | 2700 | 1317.9 |
| apscheduler | 4 | 10000 | 399.2 |
| apscheduler | 16 | 40 | 63644.3 |
| apscheduler | 16 | 300 | 31892.4 |
| apscheduler | 16 | 2700 | 5419.3 |
| apscheduler | 16 | 10000 | 1557.7 |
| apscheduler | 64 | 40 | 284656.4 |
| apscheduler | 64 | 300 | 126561.3 |
| apscheduler | 64 | 2700 | 21519.8 |
| apscheduler | 64 | 10000 | 6207.0 |
| asyncio_pool | 1 | 40 | 7114.0 |
| asyncio_pool | 1 | 300 | 2471.2 |
| asyncio_pool | 1 | 2700 | 383.9 |
| asyncio_pool | 1 | 10000 | 105.5 |
| asyncio_pool | 4 | 40 | 25031.0 |
| asyncio_pool | 4 | 300 | 9746.5 |
| asyncio_pool | 4 | 2700 | 1455.1 |
| asyncio_pool | 4 | 10000 | 400.9 |
| asyncio_pool | 16 | 40 | 115870.7 |
| asyncio_pool | 16 | 300 | 40594.7 |
| asyncio_pool | 16 | 2700 | 5670.2 |
| asyncio_pool | 16 | 10000 | 1601.3 |
| asyncio_pool | 64 | 40 | 516381.4 |
| asyncio_pool | 64 | 300 | 171144.5 |
| asyncio_pool | 64 | 2700 | 23045.6 |
| asyncio_pool | 64 | 10000 | 6125.1 |
| bare_thread | 1 | 40 | 7052.9 |
| bare_thread | 1 | 300 | 2466.7 |
| bare_thread | 1 | 2700 | 368.0 |
| bare_thread | 1 | 10000 | 100.3 |
| bare_thread | 4 | 40 | 24678.8 |
| bare_thread | 4 | 300 | 9780.0 |
| bare_thread | 4 | 2700 | 1504.0 |
| bare_thread | 4 | 10000 | 408.4 |
| bare_thread | 16 | 40 | 91714.4 |
| bare_thread | 16 | 300 | 37152.4 |
| bare_thread | 16 | 2700 | 5873.4 |
| bare_thread | 16 | 10000 | 1626.9 |
| bare_thread | 64 | 40 | 377578.2 |
| bare_thread | 64 | 300 | 150881.9 |
| bare_thread | 64 | 2700 | 23513.0 |
| bare_thread | 64 | 10000 | 6522.6 |
| process_pool | 1 | 40 | 3010.4 |
| process_pool | 1 | 300 | 1648.8 |
| process_pool | 1 | 2700 | 331.4 |
| process_pool | 1 | 10000 | 99.9 |
| process_pool | 4 | 40 | 21208.8 |
| process_pool | 4 | 300 | 13549.0 |
| process_pool | 4 | 2700 | 2861.8 |
| process_pool | 4 | 10000 | 869.9 |
| process_pool | 16 | 40 | 99452.7 |
| process_pool | 16 | 300 | 57119.0 |
| process_pool | 16 | 2700 | 11467.3 |
| process_pool | 16 | 10000 | 3396.2 |
| process_pool | 64 | 40 | 326969.3 |
| process_pool | 64 | 300 | 215170.2 |
| process_pool | 64 | 2700 | 45160.7 |
| process_pool | 64 | 10000 | 13591.7 |
| thread_pool | 1 | 40 | 11967.2 |
| thread_pool | 1 | 300 | 3094.4 |
| thread_pool | 1 | 2700 | 374.9 |
| thread_pool | 1 | 10000 | 104.0 |
| thread_pool | 4 | 40 | 42122.3 |
| thread_pool | 4 | 300 | 11034.0 |
| thread_pool | 4 | 2700 | 1478.5 |
| thread_pool | 4 | 10000 | 408.3 |
| thread_pool | 16 | 40 | 176759.3 |
| thread_pool | 16 | 300 | 45560.6 |
| thread_pool | 16 | 2700 | 5885.6 |
| thread_pool | 16 | 10000 | 1599.9 |
| thread_pool | 64 | 40 | 714659.3 |
| thread_pool | 64 | 300 | 183061.8 |
| thread_pool | 64 | 2700 | 23119.9 |
| thread_pool | 64 | 10000 | 6385.7 |
| zoo | 1 | 40 | 6340.1 |
| zoo | 1 | 300 | 2353.5 |
| zoo | 1 | 2700 | 357.6 |
| zoo | 1 | 10000 | 100.7 |
| zoo | 4 | 40 | 21830.7 |
| zoo | 4 | 300 | 9111.4 |
| zoo | 4 | 2700 | 1404.6 |
| zoo | 4 | 10000 | 398.0 |
| zoo | 16 | 40 | 93723.0 |
| zoo | 16 | 300 | 36443.4 |
| zoo | 16 | 2700 | 5468.7 |
| zoo | 16 | 10000 | 1522.8 |
| zoo | 64 | 40 | 345987.8 |
| zoo | 64 | 300 | 135579.6 |
| zoo | 64 | 2700 | 21794.5 |
| zoo | 64 | 10000 | 6086.6 |