Understanding QPS, TPS, Concurrency, and Throughput: Key Metrics for System Performance

QPS and TPS Definitions

QPS ( Queries Per Second) measures the number of queries a server can handle per second, reflecting its maximum throughput capacity.

TPS ( Transactions Per Second) counts the number of complete transactions per second, where a transaction includes a client request, server processing, and response.

Differences Between QPS and TPS

TPS counts each complete request‑response cycle, while QPS can count multiple individual queries generated by a single page request. For example, a page load may generate three server queries, resulting in one TPS but three QPS.

Visiting a page that triggers three server requests creates one transaction (TPS) but three queries (QPS). Similarly, a fast eater handling ten buns per second represents TPS, whereas a slower eater handling one bun per second represents QPS.

Concurrency and Throughput

Concurrency (parallelism) is the number of requests a system can handle simultaneously, indicating load capacity. It can be derived from analyzing request logs within a one‑second interval.

System throughput is typically determined by two factors: QPS/TPS and concurrency. When either factor reaches its limit, overall throughput stops increasing and may even decline due to overload, context switching, memory pressure, etc.

Related Traffic Metrics

PV (Page View) : Number of page accesses; each refresh counts as one view.

UV (Unique Visitor) : Count of distinct users visiting within a day, deduplicated by unique identifiers.

DAU (Daily Active Users) : Number of users who log in or use the product in a single day, similar to UV.

MAU (Monthly Active Users) : Deduplicated active users over a month.

Key Performance Testing Concepts

From a user perspective, the critical metric is response time—the interval from initiating an action to the result being displayed. Short response times improve user experience.

Administrators should monitor:

Response time

Reasonable server resource usage

Reasonable application server and database resource usage

Scalability of the system

Maximum supported concurrent users and business processing capacity

Potential performance bottlenecks

Hardware upgrades that could improve performance

24/7 availability support

Developers should consider:

Reasonable architecture design

Reasonable database design

Code performance issues

Inefficient memory usage

Poor thread synchronization

Resource contention problems

System Throughput Evaluation

When designing a system, account for CPU, I/O, and external service latency to estimate performance. Besides QPS and concurrency, daily PV is another dimension for capacity planning.

Typical technical approach:

Identify the system's peak TPS and daily PV, noting that they have a relatively stable relationship except for holidays or seasonal effects.

Use stress testing or experience to determine the highest TPS, then calculate the maximum daily throughput based on the TPS‑PV relationship.

Important parameters influencing throughput include QPS/TPS, concurrency, and average response time, with the relationship:

QPS (or TPS) = Concurrency ÷ Average Response Time

When either concurrency or response time reaches its limit, overall throughput plateaus or declines.