> ## Documentation Index
> Fetch the complete documentation index at: https://docs.kvant.sbs/jeston/llms.txt
> Use this file to discover all available pages before exploring further.

# Benchmarking

# Benchmarking

***

## Operations · Benchmark

> **Operational objective** — Measure the complete workload honestly, including concurrency, dependencies, memory, and tail behavior.

A benchmark that explains saturation and can be repeated across changes.

### Key concepts

<Columns cols={3}>
  <Card title="p50" icon="sparkles">
    Typical request
  </Card>

  <Card title="p95" icon="shield-check">
    Slow tail
  </Card>

  <Card title="p99" icon="gauge-high">
    Extreme tail
  </Card>
</Columns>

### Evidence model

```mermaid placement="top-right" theme={null}
flowchart LR
    workload[Workload model] --> clients[Concurrent clients]
    clients --> system[Complete topology]
    system --> samples[Latency samples]
    samples --> percentiles[p50 p95 p99]
    percentiles --> decision[Capacity decision]
```

### Evidence over assumption

| Signal   | Healthy interpretation | Action when it degrades                 |
| -------- | ---------------------- | --------------------------------------- |
| p50      | Typical request        | Useful for baseline.                    |
| p95      | Slow tail              | Useful for user experience.             |
| p99      | Extreme tail           | Useful for saturation and incidents.    |
| RSS/heap | Memory behavior        | Useful for leak and pressure detection. |

<Warning>
  Do not compare frameworks using different workloads, caches, connection pools, or failure policies.
</Warning>

### Operational context

A fast isolated handler can still produce a slow product when database, network, serialization, or queue behavior dominates.

> **Operator's principle**
>
> A benchmark is a decision instrument, not a victory number.

### Implementation notes

| Engineering move      | Guidance                                                                 |
| --------------------- | ------------------------------------------------------------------------ |
| **Model real work**   | Choose payloads, reads, writes, streams, tools, and dependency behavior. |
| **Warm and cold run** | Separate startup, cache, connection, and steady-state effects.           |
| **Increase pressure** | Find the point where latency, errors, memory, or queues change slope.    |
| **Record context**    | Keep runtime, hardware, version, topology, and command with the result.  |

### Decision lens

| Mode      | Practical emphasis                              |
| --------- | ----------------------------------------------- |
| **Micro** | Locate a function or serialization cost.        |
| **Load**  | Understand concurrent request behavior.         |
| **Soak**  | Find leaks, queue growth, and slow degradation. |

## Related topics

<Columns cols={3}>
  * [server-stack · **Apply capacity findings**](/jeston/jeston/operations/production) — Read the focused guide for this boundary.
  * [chart-line · **Instrument load**](/jeston/jeston/platform/health-observability) — Read the focused guide for this boundary.
  * [database · **Test database pressure**](/jeston/jeston/platform/sql) — Read the focused guide for this boundary.
</Columns>

## References

[1]: https://github.com/jeffersoncampos12p-dev/jeston "Jeston source repository"

[2]: https://www.npmjs.com/package/@kvantjs/jeston "Jeston package on npm"

[3]: https://nodejs.org/api/http.html "Node.js HTTP API"

[4]: https://developer.mozilla.org/en-US/docs/Web/API/AbortSignal "AbortSignal Web API"

[5]: https://react.dev/reference/react-dom/server "React server rendering APIs"

[6]: https://www.typescriptlang.org/docs/handbook/intro.html "TypeScript handbook"