1. Traffic Generate arrivals with RPS and burst shape.

2. Webserver The front door owns the end-to-end request deadline.

3. Limiter queue Every admitted request waits for an allow or reject decision.

4. Policy Rules produce 429 and keep later queues under control.

5. App pending When app capacity is full, requests remain pending until capacity opens or timeout expires.

6. Dependency capacity Downstream active capacity and latency can still push the webserver deadline over the edge.

Inputs

Traffic

Target RPS sets the baseline offered traffic. Burst factor adds a smooth wave around that baseline. A burst factor of 0 means the expected traffic shape is flat.

Duration seconds Step milliseconds Target RPS requests/sec Burst factor 0 means flat expected traffic. Higher values add a smooth burst wave. Traffic noise Sample arrivals with random noise around the expected traffic shape Off by default. When disabled, arrivals follow the expected shape deterministically.

Expected offered traffic: 40 rps baseline with a 40% burst wave over 15s

Front door

Webserver

Models the high-performance request owner before the limiter. By default it is generous; the important setting is the end-to-end timeout it applies while downstream stages work.

Webserver active capacity open client requests Webserver pending capacity requests waiting for a webserver slot Webserver pending timeout timeout before the request enters the limiter path End-to-end request timeout webserver returns 503 if the downstream path takes longer

Policy

Rate Limiter

Rules, decision latency, limiter capacity, and limiter failure behavior before app admission.

Window algorithm How requests are counted inside each window Decision distribution Limiter failure mode Used when limiter capacity is full or its pending timeout expires Decision latency milliseconds Unused ignored for constant/exponential Limiter active capacity concurrent limiter decisions Limiter pending capacity requests waiting for a limiter slot Limiter pending timeout timeout returns 503 before policy decision

Constant

Rate limit rules Every request must pass all rules. Policy rejects are 429.

Sliding: 30 / 1s

Backend

Service

Models app capacity, app pending behavior, and app service time before a request tries the dependency.

App active capacity max active requests App pending capacity requests allowed to wait for app capacity App pending timeout timeout while waiting for app capacity Distribution Mean latency milliseconds Std deviation milliseconds

Normal / Gaussian

Dependency

Downstream

Models the external dependency the app is trying to protect.

Dependency active capacity max active requests Dependency distribution Mean latency milliseconds Std deviation milliseconds

Normal / Gaussian

Topology

System Pressure Map

Current placement: traffic enters the webserver first; the limiter sits before app capacity.

Result

Analytics

Summary from the latest run.

Time series

Capacity, Admissions, and Rejections

Incoming Traffic/s shows offered load. Expected Traffic/s is the configured shape. 429 means limiter reject. 503 can come from webserver timeout, limiter capacity, app capacity, or dependency capacity.

Distribution

Latency Distribution by Status

Selectable latency distributions for HTTP 200, 429, 503, and overall.

Select HTTP 200, 429, 503, or overall distribution