Spike Test

Performance testing requires careful scenario design matching specific criteria. While load, stress, and soak tests are commonly performed, spike testing is an underutilized but valuable approach. A spike test simulates an extreme increment or decrement in load to evaluate recovery behavior and identify application weaknesses.

What is a Spike Test?

A spike test introduces a sudden, dramatic increase in load to observe:

• How the system handles the sudden surge
• Whether the system recovers to normal performance
• What breaks under extreme conditions
• How long recovery takes

LoadRunner Implementation

The approach uses two script groups:

1. Baseline Group: 10 virtual users ramping up for constant 1-hour load
2. Spike Group: 40 users delayed 30 minutes before simultaneous deployment

This creates a visible performance disruption in controller graphs, with the spike group overlaying the steady-state baseline.

LoadRunner Scenario Configuration

Group 1 (Baseline):
  - Virtual Users: 10
  - Ramp Up: Gradual over 5 minutes
  - Duration: 60 minutes
  - Ramp Down: 5 minutes

Group 2 (Spike):
  - Virtual Users: 40
  - Start Delay: 30 minutes
  - Ramp Up: Immediate (all at once)
  - Duration: 5 minutes
  - Ramp Down: Immediate

Apache JMeter Implementation

The Ultimate Thread Group plugin enables complex scenarios with precise control.

Configuration

Row 1 - Baseline Load:

• Threads: 100
• Initial Delay: 0
• Startup Time: 300 seconds (gradual ramp)
• Hold Load: 3600 seconds (1 hour)
• Shutdown Time: 60 seconds

Row 2 - Spike Load:

• Threads: 100
• Initial Delay: 1800 seconds (30 minutes)
• Startup Time: 0 (immediate)
• Hold Load: 300 seconds (5 minutes)
• Shutdown Time: 0 (immediate)

Ultimate Thread Group Settings

Ultimate Thread Group:
  Row 1: Start=100, Initial Delay=0, Startup=300, Hold=3600, Shutdown=60
  Row 2: Start=100, Initial Delay=1800, Startup=0, Hold=300, Shutdown=0

The combined threading visualization displays the spike halfway through when additional threads join the existing load.

Monitoring Considerations

Continuous infrastructure monitoring proves essential during spike tests:

1. Establish baseline - The steady-state period before the spike provides baseline behavior for comparison
2. Monitor during spike - Track resource utilization, response times, and error rates
3. Verify recovery - Confirm the system returns to baseline performance
4. Wait between spikes - If running multiple spikes, ensure full recovery before introducing additional load

Key Metrics to Monitor

• CPU utilization
• Memory usage
• Response times
• Error rates
• Queue depths
• Connection pool usage
• Database performance

Best Practices

1. Single spike first - Start with one spike to understand behavior before adding complexity
2. Document recovery time - Measure how long it takes to return to normal
3. Monitor all tiers - Watch web servers, application servers, and databases
4. Compare to baseline - Always have a steady-state reference point
5. Test in production-like environment - Results are only valid if the environment matches production

Further Reading

• Guru99 - Spike Testing
• Performance Testing Types