Ktor Benchmarks

This repository contains benchmark tests for various Ktor performance metrics including memory allocation, throughput, I/O operations, and utilities.

Prerequisites

• JDK 17 or higher
• Gradle (wrapper included)

Project Structure

allocation-benchmark

Tests memory allocation on a given request to ensure Ktor maintains a low memory profile by avoiding erroneous allocations.

Tests:

• ServerCallAllocationTest - Measures memory allocated per request for server engines (Jetty, Tomcat, Netty, CIO)
• ClientCallAllocationTest - Measures memory allocated per request for client engines (CIO, Apache, OkHttp, Java)

Running Tests:

cd allocation-benchmark
./gradlew test                # Run all allocation tests
./gradlew serverTests         # Run only server allocation tests
./gradlew dumpAllocations     # Generate new allocation baselines
./gradlew reportServer        # Start web server to view allocation reports

Viewing Reports: After running reportServer, open your browser to view:

• previewClasses.html - Largest memory consumers by type
• previewSites.html - Code sites that allocate the most memory

TeamCity: Allocation Tests

server-benchmarks

JMH-based benchmarks for testing Ktor server performance under various scenarios including integration tests and specific feature benchmarks.

Running Benchmarks:

cd server-benchmarks
./gradlew jmh                 # Run benchmarks (configure includes in build.gradle.kts)

Note: Edit build.gradle.kts to configure which benchmarks to run via the includes list.

client-benchmarks

Throughput benchmarks comparing different Ktor client engines (CIO, Apache, OkHttp, Java) against a local server.

Running Benchmarks:

cd client-benchmarks
./gradlew run                 # Run client throughput benchmarks

This will:

1. Start a Netty server on port 8080
2. Run concurrent requests using different client engines
3. Report requests per second for each engine

io-benchmarks

JMH benchmarks for I/O operations including file reading and socket operations across different dispatchers and implementations.

Running Benchmarks:

cd io-benchmarks
./gradlew jmh                 # Run all I/O benchmarks

Benchmarks include:

• File reading with various dispatchers (IO, Hot, Fixed, Blocking, etc.)
• Socket read/write operations
• Comparison between JVM native and Ktor implementations

See io-benchmarks/README.md for latest results.

utils-benchmarks

JMH benchmarks for Ktor utility components, primarily focusing on pipeline execution performance.

Running Benchmarks:

cd utils-benchmarks
./gradlew jmh                 # Run pipeline benchmarks

See utils-benchmarks/README.md for latest results.

benchmark-server

Gradle plugin providing test server infrastructure used by other benchmark modules. Contains reusable server configurations and routes.

Quick Start

To run all benchmarks:

# Allocation tests
cd allocation-benchmark && ./gradlew test && cd ..

# Server benchmarks
cd server-benchmarks && ./gradlew jmh && cd ..

# Client benchmarks
cd client-benchmarks && ./gradlew run && cd ..

# I/O benchmarks
cd io-benchmarks && ./gradlew jmh && cd ..

# Utils benchmarks
cd utils-benchmarks && ./gradlew jmh && cd ..

Benchmark Types

• Allocation Benchmarks - Memory allocation testing using instrumentation agent
• JMH Benchmarks - Standard JMH microbenchmarks for throughput and average time measurements
• Throughput Benchmarks - Concurrent request testing for client/server performance

Configuration

• JVM arguments and benchmark parameters are configured in each subproject's build.gradle.kts
• Allocation baselines are stored in allocations/ directories
• JMH results may be output to CSV files in respective directories