Mastering Software Performance Testing: Key Concepts & Calculation Formulas

Software Performance Testing: Core Concepts and Formulas

System throughput metrics are closely linked to request CPU consumption, external interfaces, and I/O. Higher CPU usage per request or slower external I/O reduces overall throughput.

Key parameters include:

QPS (TPS) : number of requests/transactions per second.

Concurrency : number of requests processed simultaneously.

Response time : average time to respond to a request.

The relationship is: QPS (TPS) = Concurrency / AverageResponseTime When either concurrency or QPS reaches its limit, overall throughput cannot increase further and may even drop due to overload, context switching, and memory contention.

Evaluating System Throughput

Beyond QPS and concurrency, daily page views (PV) are often used to estimate capacity. By observing logs, one can derive daily traffic from peak TPS using typical ratios (e.g., for Taobao, PeakTPS : PV ≈ 1 : 11×3600).

Given a measured peak TPS, daily throughput can be approximated as: DailyThroughput = PeakTPS × HoursOfPeakLoad × 3600 Load testing formulas also relate TPS, concurrent virtual users, and think time:

TPS = U_concurrent / (T_response + T_think)

Performance Testing Focus Areas

Different roles care about different metrics:

Users : perceived response time.

Administrators : response time, resource utilization, scalability, maximum concurrent users, bottlenecks, hardware upgrades, 24/7 availability.

Developers/Designers : architecture, database design, code efficiency, memory usage, thread synchronization, resource contention.

Performance Test Engineers : all of the above.

Key Terminology

Response time = NetworkTransmission(N1+N2+N3+N4) + ApplicationServerProcessing(A1+A3) + DatabaseProcessing(A2).

Concurrent users can be estimated by:

ConcurrentUsers = RPS + ConcurrentConnections + AverageThinkTime

Average concurrent users: C = nL / T where n is daily active users, L is average session length, T is observation period.

Peak concurrent users (Poisson approximation): Peak ≈ C + 3×√C Throughput is the number of requests processed per unit time, measured as requests/second, pages/second, users/day, etc., and reflects system load capacity.

When no bottleneck is present, throughput relates to virtual users: F = VU × R / T where F is throughput, VU is virtual users, R is requests per user, T is test duration.

Performance Counters

Metrics such as CPU usage, memory consumption, and process time help monitor and analyze scalability and locate bottlenecks.

Think Time

Think time simulates the pause between user actions. It influences request rate: R = T / TS where TS is think time.

Typical steps to calculate think time:

Determine concurrent users: C = nL / T.

Measure average throughput.

Count average requests per user.

Compute think time: TS = T / R.