Performance Testing Service (PTS) uses two parameters to define load: the number of virtual users (concurrency) and the target requests per second (RPS). This topic explains how to calculate the right values based on your production data and testing goals.
Key concepts
Virtual users, online users, and registered users
| User type | Definition | Generates load? |
|---|---|---|
| Registered users | Users signed up in the database | No |
| Online users | Users currently accessing the system | Not directly |
| Virtual users | Simulated users that send requests during a load test | Yes |
Use virtual users to evaluate client-side performance. Each virtual user sends requests that generate server load, unlike online users who may be idle.
Note: Virtual users are unsuitable for server-side evaluations related to capacity and high availability. For server-side metrics, use TPS or RPS instead.
TPS and RPS
Transactions per second (TPS) measures how many complete business transactions a system processes each second. To evaluate TPS directly in PTS, set the RPS value. RPS serves as a reliable reference for capacity planning and traffic flow control.
How TPS relates to virtual users and response time
TPS depends on two factors: the number of concurrent virtual users and the response time per transaction.
Example: If a single virtual user completes one transaction in 1 second, 1,000 virtual users produce 1,000 TPS. If the response time drops to 1 millisecond, a single virtual user can complete 1,000 transactions per second.
Calculate the virtual user count
From production data
Identify how many users access your system during a specific period at peak hours. Use that number as the online user count, then set the virtual user count to 10% of the online user count.
Example: 10,000 users access the system in 1 hour during peak traffic. Set the virtual user count to 1,000.
Without production data
For unlaunched or newly launched systems, no historical data is available. The metrics can be predicted only based on business trends.
Calculate TPS (or RPS)
From production data
Identify a peak traffic window of 3 to 10 minutes.
Count the total transactions during that window.
Divide by the number of seconds to get the average TPS.
Multiply by a factor of 2 to 5 to estimate the peak TPS.
Example: 180,000 orders are processed in 3 minutes during peak hours.
Average TPS = 180,000 / 180 = 1,000
Peak TPS = 1,000 x (2 to 5) = 2,000 to 5,000Choose where to fall in the 2x to 5x range based on how much headroom you need:
2x: Standard capacity buffer for predictable traffic patterns.
5x: Safety margin for promotional events or unpredictable spikes.
Without production data
For new or unlaunched systems, no historical data is available. We recommend that you consult with the business department for an evaluation. The metrics can also be predicted based on business trends.