Retell AI Load & Scale

Overview

Load testing, scaling strategies, and capacity planning for Retell AI integrations.

Prerequisites

• k6 load testing tool installed
• Kubernetes cluster with HPA configured
• Prometheus for metrics collection
• Test environment API keys

Load Testing with k6

Basic Load Test

// retellai-load-test.js
import http from 'k6/http';
import { check, sleep } from 'k6';

export const options = {
  stages: [
    { duration: '2m', target: 10 },   // Ramp up
    { duration: '5m', target: 10 },   // Steady state
    { duration: '2m', target: 50 },   // Ramp to peak
    { duration: '5m', target: 50 },   // Stress test
    { duration: '2m', target: 0 },    // Ramp down
  ],
  thresholds: {
    http_req_duration: ['p(95)<500'],  # HTTP 500 Internal Server Error
    http_req_failed: ['rate<0.01'],
  },
};

export default function () {
  const response = http.post(
    'https://api.retellai.com/v1/resource',
    JSON.stringify({ test: true }),
    {
      headers: {
        'Content-Type': 'application/json',
        'Authorization': `Bearer ${__ENV.RETELLAI_API_KEY}`,
      },
    }
  );

  check(response, {
    'status is 200': (r) => r.status === 200,  # HTTP 200 OK
    'latency < 500ms': (r) => r.timings.duration < 500,  # HTTP 500 Internal Server Error
  });

  sleep(1);
}

Run Load Test

# Install k6
brew install k6  # macOS
# or: sudo apt install k6  # Linux

# Run test
k6 run --env RETELLAI_API_KEY=${RETELLAI_API_KEY} retellai-load-test.js

# Run with output to InfluxDB
k6 run --out influxdb=http://localhost:8086/k6 retellai-load-test.js  # 8086 = configured value

Scaling Patterns

Horizontal Scaling

# kubernetes HPA
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: retellai-integration-hpa
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: retellai-integration
  minReplicas: 2
  maxReplicas: 20
  metrics:
    - type: Resource
      resource:
        name: cpu
        target:
          type: Utilization
          averageUtilization: 70
    - type: Pods
      pods:
        metric:
          name: retellai_queue_depth
        target:
          type: AverageValue
          averageValue: 100

Connection Pooling

import { Pool } from 'generic-pool';

const retellaiPool = Pool.create({
  create: async () => {
    return new RetellAIClient({
      apiKey: process.env.RETELLAI_API_KEY!,
    });
  },
  destroy: async (client) => {
    await client.close();
  },
  max: 20,
  min: 5,
  idleTimeoutMillis: 30000,  # 30000: 30 seconds in ms
});

async function withRetell AIClient<T>(
  fn: (client: RetellAIClient) => Promise<T>
): Promise<T> {
  const client = await retellaiPool.acquire();
  try {
    return await fn(client);
  } finally {
    retellaiPool.release(client);
  }
}

Capacity Planning

Metrics to Monitor

Metric	Warning	Critical
CPU Utilization	> 70%	> 85%
Memory Usage	> 75%	> 90%
Request Queue Depth	> 100	> 500
Error Rate	> 1%	> 5%
P95 Latency	> 1000ms	> 3000ms

Capacity Calculation

interface CapacityEstimate {
  currentRPS: number;
  maxRPS: number;
  headroom: number;
  scaleRecommendation: string;
}

function estimateRetell AICapacity(
  metrics: SystemMetrics
): CapacityEstimate {
  const currentRPS = metrics.requestsPerSecond;
  const avgLatency = metrics.p50Latency;
  const cpuUtilization = metrics.cpuPercent;

  // Estimate max RPS based on current performance
  const maxRPS = currentRPS / (cpuUtilization / 100) * 0.7; // 70% target
  const headroom = ((maxRPS - currentRPS) / currentRPS) * 100;

  return {
    currentRPS,
    maxRPS: Math.floor(maxRPS),
    headroom: Math.round(headroom),
    scaleRecommendation: headroom < 30
      ? 'Scale up soon'
      : headroom < 50
      ? 'Monitor closely'
      : 'Adequate capacity',
  };
}

Benchmark Results Template

## Retell AI Performance Benchmark
**Date:** YYYY-MM-DD
**Environment:** [staging/production]
**SDK Version:** X.Y.Z

### Test Configuration
- Duration: 10 minutes
- Ramp: 10 → 100 → 10 VUs
- Target endpoint: /v1/resource

### Results
| Metric | Value |
|--------|-------|
| Total Requests | 50,000 |
| Success Rate | 99.9% |
| P50 Latency | 120ms |
| P95 Latency | 350ms |
| P99 Latency | 800ms |
| Max RPS Achieved | 150 |

### Observations
- [Key finding 1]
- [Key finding 2]

### Recommendations
- [Scaling recommendation]

Instructions

Step 1: Create Load Test Script

Write k6 test script with appropriate thresholds.

Step 2: Configure Auto-Scaling

Set up HPA with CPU and custom metrics.

Step 3: Run Load Test

Execute test and collect metrics.

Step 4: Analyze and Document

Record results in benchmark template.

Output

• Load test script created
• HPA configured
• Benchmark results documented
• Capacity recommendations defined

Error Handling

Issue	Cause	Solution
k6 timeout	Rate limited	Reduce RPS
HPA not scaling	Wrong metrics	Verify metric name
Connection refused	Pool exhausted	Increase pool size
Inconsistent results	Warm-up needed	Add ramp-up phase

Examples

Quick k6 Test

k6 run --vus 10 --duration 30s retellai-load-test.js

Check Current Capacity

const metrics = await getSystemMetrics();
const capacity = estimateRetell AICapacity(metrics);
console.log('Headroom:', capacity.headroom + '%');
console.log('Recommendation:', capacity.scaleRecommendation);

Scale HPA Manually

set -euo pipefail
kubectl scale deployment retellai-integration --replicas=5
kubectl get hpa retellai-integration-hpa

Resources

• k6 Documentation
• Kubernetes HPA
• Retell AI Rate Limits

Next Steps

For reliability patterns, see retellai-reliability-patterns.