Understanding the intuition behind `request-rate`

I have conducted an analysis of the `request-rate` and `interval` variables in the `benchmarking_script.py` and would like to ensure that my understanding is correct.

My understanding is that the `request-rate` parameter introduces delays between each request to mimic the `queries-per-second` (`QPS`) rate. For example, if it is set to 5, then 5 requests are sent within a 1-second window for a sufficiently large number of samples.

That being said, the delay implementation generates random delay values with slightly high variance, but the average is fairly consistent:

[`interval = np.random.exponential(1.0 / request_rate)`](https://github.com/google/JetStream/blob/main/benchmarks/benchmark_serving.py#L351)

*The graph of `exponential distribution` from which `1/request-rate` is being sampled*

___

When I plot the `interval` values for a given request-rate (e.g., 5), I get the following plot after running it 1000 times:

With these statistics:

```
Mean: 0.19030281331325313
Variance: 0.03450950095960781
Standard Deviation: 0.18576733017300917
Minimum: 6.344257769502491e-05
Maximum: 1.2152877332855887
Sum: 190.30281331325313
```

Given that the mean is around 0.2, the overall QPS is 5 requests per second (since 5 × 0.2 = 1 second).

___

Here are the statistics for `request-rate = 10`:

```
**Mean: 0.0983383984114717**
Variance: 0.008309511474488432
Standard Deviation: 0.0911565218428634
Minimum: 3.9151506088517006e-05
Maximum: 0.6463796229426464
Sum: 98.3383984114717
```

___

In conclusion, the `request-rate` parameter effectively mimics the QPS (queries per second) metric when the number of samples is large enough.

I would like to confirm that my understanding is correct and document this in the issues section for anyone else who might have the same question.