Multi-region distributed load

Problem: Your application serves users on several continents. A test from a single cloud region does not show the latency that users far from your origin see. Multi-region load gives you a realistic mix of latency profiles and exposes region-specific performance differences (CDN miss rates, routing issues, regionalised services).

Test type: Load test or Stress test with geographic coverage.

Prerequisites

• A MaxoPerf workspace with access to two or more cloud load locations (visible in the Locations dropdown).
• A test file (any engine) already validated with a smoke test from a single region.

Step by step in MaxoPerf

1. Configure locations in the MaxoPerf console

MaxoPerf distributes load across locations at the runner level, so you do not change your test file. You set the location split in the Configuration tab.

1. Open the test and switch to the Configuration tab.
2. Under Locations, click Add location.
3. Select a second (and third) region, for example eu-west-1 (Ireland) and us-east-1 (Virginia).
4. Set the weight (percentage of total VUs) for each location. Weights must sum to 100.
5. Save.

Common distributions:

• 50 / 50: equal split for a symmetric test.
• 70 / 20 / 10: most traffic from the primary user region, a smaller share from secondary regions.
• Single secondary at 5 %: a canary region that catches routing anomalies without changing the load shape much.

2. Start the run

1. Click Run now.
2. Open the run overview. A Locations filter sits at the top of the charts.

3. Analyse per-region metrics

1. In the run overview, use the Locations filter to view metrics for a single region at a time.
2. Compare p95 latency between regions. Investigate a large difference (> 20 %).
3. Switch to the Runners tab to see each runner’s location, status, and live VU count.

CDN warm-up effect

The first few minutes of a multi-region test often show higher latency from regions that were not active before, because CDN edge nodes serve cold-cache misses. Allow a 5-minute ramp-up before you read steady-state results.

Verify

• Both (all) regions appear in the Runners tab with Running status and VU counts proportional to their weight.
• The Locations filter in the overview shows distinct per-region throughput and latency charts.
• p95 latency differs between regions in line with geographic distance: a few milliseconds for co-located regions, tens of milliseconds for intercontinental pairs.
• No region shows an unusually high error rate. If one does, look for routing or firewall issues from that source region.

Variations

• Private location + cloud location: mix a private datacenter (BYOC Outpost) location with a managed cloud region. See Private datacenter test.
• Single region targeted spike: add the spike pattern from Spike and recover, but apply it to one region only by adjusting weights live.
• Different scenarios per region: Taurus supports per-executor location assignment when you need different scenarios from different origins. See Taurus fundamentals.

Where to go next

• Private datacenter (Outpost) test: add a location inside your own infrastructure.
• Connect a private datacenter: the how-to for installing the BYOC Outpost agent.
• Load test: the test type this recipe uses.