04-round-robin

Pattern 04: Round-Robin Distribution

Distribute individual test cases evenly across all available pCloudy device slots. The highest execution ratio pattern — no slot sits idle while another finishes early.

When to use this pattern

• You have a large, heterogeneous test suite (100+ tests of varying duration)
• Maximum speed is the priority — you want the shortest possible wall-clock time
• Your tests are independent (no shared state between test cases)

How it works

200 tests → 5 device slots → 40 tests per slot (balanced)

Without round-robin (by class):
  Slot 1: ████████████████████████ 40 tests, 48 min
  Slot 2: ████████████ 20 tests, 22 min  ← idle for 26 min
  Slot 3: ██████████████████ 30 tests, 35 min ← idle for 13 min
  Slot 4: ████████████████████████████ 50 tests, 58 min
  Slot 5: ████████████████ 30 tests, 35 min
  Total wall-clock: 58 min | Slot utilisation: 64%

With round-robin:
  Slot 1: ████████████████████ 40 tests, ~42 min
  Slot 2: ████████████████████ 40 tests, ~42 min
  Slot 3: ████████████████████ 40 tests, ~42 min
  Slot 4: ████████████████████ 40 tests, ~42 min
  Slot 5: ████████████████████ 40 tests, ~42 min
  Total wall-clock: ~42 min | Slot utilisation: 94%

Execution ratio impact

Highest of all four patterns. Round-robin minimises idle slot time and maximises the ratio of tests executed per available execution window. If your execution ratio is below 80%, switching to round-robin from a by-class split typically recovers 15–25 percentage points.

Tradeoffs

Advantage	Disadvantage
Maximum device utilisation	Tests must be fully independent
Shortest total execution time	Harder to attribute failures to a domain
Scales linearly — add a slot, reduce time proportionally	Requires a distribution script (included below)

Test independence requirement

Round-robin requires that tests share no state. Each test must:

• Create its own test data (not depend on a previous test's output)
• Clean up after itself
• Not assume a specific device or session

If your tests share state, use Pattern 01 (By Test Class) or Pattern 02 (By Feature) first, then migrate to round-robin once tests are independent.

---

Implementation

• Appium + TestNG (Java) — uses TestNG's dataprovider + custom distributor
• WebdriverIO (Node.js) — uses spec sharding
• Playwright (Node.js) — uses built-in --shard flag
• Selenium (Java) — uses parallel thread pool with queue