Execution Profiles

GuideLLM supports multiple load patterns for different testing scenarios.

Profile Types

Sweep Profile

Automatically explore different request rates to find safe operating ranges.

Use case: Discovery - find optimal request rates for your deployment

Configuration:

{
  "parameters": {
    "profile": "sweep",
    "max_seconds": 30,
    "detect_saturation": true
  }
}

Behaviour: Incrementally increases request rate until saturation or limits are reached.

---

Throughput Profile

Maximum capacity testing to identify performance limits.

Use case: Stress testing - find the breaking point

Configuration:

{
  "parameters": {
    "profile": "throughput",
    "max_seconds": 60,
    "max_requests": 1000
  }
}

Behaviour: Sends requests as fast as possible to saturate the server.

---

Concurrent Profile

Simulate parallel users with fixed concurrency level.

Use case: User simulation - test with realistic concurrent load

Configuration:

{
  "parameters": {
    "profile": "concurrent",
    "rate": 10,
    "max_requests": 100
  }
}

Behaviour: Maintains exactly N concurrent requests at all times.

---

Constant Profile

Fixed requests per second for steady-state testing.

Use case: Baseline measurement - consistent, predictable load

Configuration:

{
  "parameters": {
    "profile": "constant",
    "rate": 5,
    "max_seconds": 10,
    "max_requests": 20
  }
}

Behaviour: Sends requests at a fixed rate (e.g., 5 req/s).

---

Poisson Profile

Randomised request rates following Poisson distribution.

Use case: Realistic simulation - natural traffic patterns

Configuration:

{
  "parameters": {
    "profile": "poisson",
    "rate": 5,
    "max_seconds": 30
  }
}

Behaviour: Random intervals averaging to the specified rate.

---

Synchronous Profile

Sequential requests for baseline measurements.

Use case: Single-user testing - minimum latency baseline

Configuration:

{
  "parameters": {
    "profile": "synchronous",
    "max_requests": 50
  }
}

Behaviour: Waits for each request to complete before sending the next.

Profile Selection Guide

Scenario	Recommended Profile	Why
First-time testing	sweep	Automatically finds safe operating range
Load testing	constant	Predictable, repeatable results
Capacity planning	throughput	Find maximum capacity
User simulation	concurrent	Realistic concurrent load
Production-like traffic	poisson	Natural traffic patterns
Baseline latency	synchronous	Minimum possible latency

Common Parameters

All profiles support these common parameters:

Parameter	Description	Default
max_seconds	Maximum duration in seconds	None
max_requests	Maximum number of requests	None
max_errors	Error threshold before stopping	None
warmup	Warmup period to exclude (% or absolute)	None
cooldown	Cooldown period to exclude (% or absolute)	None

Warmup Recommendations

Use "warmup": "5%" or "warmup": "10%" to exclude initial cold-start effects from measurements.

Examples

Quick Test

Fast test with minimal samples:

{
  "parameters": {
    "profile": "constant",
    "rate": 5,
    "max_seconds": 10,
    "max_requests": 20,
    "warmup": "0"
  }
}

Production Load Test

Realistic production simulation:

{
  "parameters": {
    "profile": "poisson",
    "rate": 50,
    "max_seconds": 300,
    "warmup": "5%",
    "detect_saturation": true
  }
}

Capacity Test

Find maximum throughput:

{
  "parameters": {
    "profile": "throughput",
    "max_seconds": 60,
    "max_requests": 5000,
    "warmup": "10%"
  }
}