# Load Testing Guidance ## Purpose Provide executable load testing strategies for dispatcher performance validation. ## Load Testing Modes ### AMS Mode (AEM 6.5/AMS) **Prerequisites:** - Accessible AMS dispatcher host - Load testing tool with network access - Permission to monitor host metrics **Approach:** ```text # 1. Establish baseline # Run load test against AMS dispatcher # 2. Monitor host resources (outside MCP) # CPU, memory, disk I/O, network I/O, Apache worker saturation # 3. Analyze dispatcher behavior monitor_metrics({ "log_path": "/var/log/httpd/dispatcher.log", "window_minutes": 30, "breakdown_by": "status_code" }) tail_logs({"log_path":"/var/log/httpd/dispatcher.log","lines":500}) ``` **Key Metrics:** - Throughput (requests/sec) - Response time (p50, p95, p99) - Apache worker utilization - System resource usage - Cache hit ratio during load ## Load Test Scenarios ### Scenario 1: Cache Warming **Goal:** Validate cache efficiency after initial population. ```text 1. Clear cache 2. Run synthetic traffic across key URLs 3. Measure cache hit ratio progression 4. Validate steady-state performance ``` ### Scenario 2: Traffic Spike Simulation **Goal:** Verify behavior under sudden load increase. ```text 1. Start with baseline load (e.g., 10 users) 2. Ramp to 10x load over 2 minutes 3. Sustain peak for 10 minutes 4. Measure response time degradation 5. Check error rate increase ``` ### Scenario 3: Cache Invalidation Under Load **Goal:** Validate invalidation behavior doesn't cause cascade failures. ```text 1. Establish steady-state load 2. Trigger cache invalidation (via flush) 3. Measure cache miss spike 4. Measure backend request surge 5. Validate recovery time ``` ### Scenario 4: Mixed Content Type Load **Goal:** Validate performance across HTML, JSON, assets. ```text 1. Generate mixed traffic: 60% HTML, 30% images, 10% JSON 2. Measure cache hit ratio per content type 3. Identify content type bottlenecks 4. Validate content-type-specific optimizations ``` ## Load Testing Tools ### Apache JMeter ```xml 100 300 1800 ``` ### k6 (Recommended for CI/CD) ```javascript import http from 'k6/http'; import { check } from 'k6'; export let options = { stages: [ { duration: '5m', target: 100 }, // Ramp-up { duration: '30m', target: 100 }, // Sustained { duration: '2m', target: 0 }, // Ramp-down ], thresholds: { 'http_req_duration': ['p(95)<500'], // 95% under 500ms 'http_req_failed': ['rate<0.01'], // <1% errors }, }; export default function () { let res = http.get('http://dispatcher:8080/content/site/en.html'); check(res, { 'status 200': (r) => r.status === 200 }); } ``` ### Gatling (for complex scenarios) ```scala val scn = scenario("Dispatcher Load Test") .exec(http("Homepage").get("/content/site/en.html")) .pause(1) .exec(http("Product Page").get("/content/site/en/products.html")) .pause(2) setUp(scn.inject(rampUsers(100) during (5 minutes))) .protocols(http.baseUrl("http://dispatcher:8080")) ``` ## Capacity Planning Calculations ### Formula: Max Concurrent Users ```text Max Users = (Throughput @ Target Latency) / (Requests per User Session) Example: - Dispatcher handles 1000 req/sec at p95 < 300ms - Average user session: 10 requests - Max Users = 1000 / 10 = 100 concurrent users/sec turnover ``` ### Formula: Cache Hit Ratio Impact ```text Backend Load = Total Requests x (1 - Cache Hit Ratio) Example: - 1000 req/sec to dispatcher - Cache hit ratio: 80% - Backend load: 1000 x (1 - 0.80) = 200 req/sec to AEM ``` ### Bottleneck Identification | Symptom | Likely Bottleneck | Investigation | |---------|-------------------|---------------| | High CPU on dispatcher | Filter/rewrite complexity | Simplify rules, profile Apache | | High memory on dispatcher | Cache size too large | Reduce cache scope, add eviction | | High disk I/O | Cache read/write thrashing | Add cache levels, optimize invalidation | | Backend saturation | Low cache hit ratio | Improve cacheability, extend TTLs | | Network saturation | Large uncached assets | Add compression, optimize assets | ## Pre-Production Load Test Checklist - [ ] Load test environment mirrors production topology - [ ] Dispatcher config is production-identical - [ ] Cache is warmed before sustained load phase - [ ] Backend (AEM) can handle expected cache miss load - [ ] Monitoring is active (metrics, logs) - [ ] Success criteria defined (latency, error rate, throughput) - [ ] Rollback plan ready if performance degrades - [ ] Load ramp is gradual (avoid thundering herd) - [ ] Multiple content types represented in load mix - [ ] Peak load sustained for meaningful duration (>15 min) ## References - Dispatcher caching docs: https://experienceleague.adobe.com/en/docs/experience-manager-dispatcher/using/configuring/dispatcher-configuration#caching-documents - AMS dispatcher overview: https://experienceleague.adobe.com/en/docs/experience-manager-learn/ams/dispatcher/overview