Common Test Types
Load Testing
Normal workload to measure response times
Stress Testing
Peak workload to observe stability under extremes
Soak Testing
Extended test to observe memory utilization/leaks
Spike Testing
Short bursts to observe stability under extremes
Scalability Testing
Step up load to monitor capability under increasing load
Volume Testing
Large volumes of data to monitor behavior
Common Metrics
Throughput
TPS
Response Time
Percentiles
Error Rate
Concurrency
Latency
Wait Time
Monitor Types
CPU
Memory
Disk
Network
Database
Web Server
Container
Swap
Typical Process Stages
1
Define Requirements
Understand app under test, define how to test and measure success
2
Test Plan
Map NFRs to business process and tests
3
Test Asset Creation
Create test assets from test plan
4
Test Execution
Execute test scenarios defined in plan
5
Results Analysis
Collate results, logs, monitors and analyze
6
Reporting
Report findings, outcomes, gaps and recommendations
Common Problems
Bottlenecks - Data flow constrained by element
Misconfiguration - Connection pools, JVM, threads
Resource constraints - Network, CPU, RAM, Disk
Scaling - App doesn't scale correctly
Environments - Test differs from production
Common Misconceptions
More hardware will fix the issue
Testing should be left to the end
Test envs easily compare to prod
Minimal data sets are representative
All test types fit every project
Should test all functionality
Testers work alone
Professional Tips
Quality - Deliver excellence in all that you do
Integrity - Be honest and communicate openly
Passion - Love what you do, show it in your work
Commitment - Deliver and exceed customer needs
Innovation - Find the best solutions
Learn - Continue to grow and adapt
Percentiles vs Average
Why Use Percentiles?
Percentile shows the value below which a given percentage of observations fall. This helps understand variations in measurements. For instance, 50th percentile (p50) shows median response time, while p95 shows what 95% of users experience.
Average Can Be Misleading
Due to peaks, average response time can be skewed upwards. A few slow responses can dramatically affect the average. Always analyze percentiles (p50, p90, p95, p99) alongside averages for a complete picture.