axboe/fio

Features

• Disk I/O Benchmarking - The Distributed I/O Stress Tester measures storage throughput and latency by running configurable job files or command line parameters.
• Storage Performance Benchmarking - Executes controlled read and write workloads to measure throughput and latency for storage devices and file systems.
• Synthetic I/O Workload Generators - Provides a framework for simulating complex storage patterns including random reads, writes, and data deduplication scenarios.
• Distributed Socket Execution - Manages remote benchmarking clients via network sockets to execute coordinated workloads across multiple machines.
• Remote Socket Management - Controls the connectivity and execution of remote test clients across a cluster using network sockets.
• Block Size Specifications - The Distributed I/O Stress Tester sets fixed or ranged block sizes for I/O units and allows weighted distribution of sizes within a workload.
• High Performance I/O Engines - Provides a common API abstraction layer to support diverse I/O engines and various operating systems.
• I/O Trace Replay Engines - Reproduces real-world storage traffic by replaying recorded block-level I/O logs against a device.
• Disk I/O Performance Analysis - Tracks I/O operations and sector counts by reading system block statistics to evaluate storage performance.
• Storage Device Profilers - Analyzes disk utilization and identifies performance bottlenecks through detailed latency histograms and bandwidth logs.
• System Statistics Collection - Reads system-level storage metrics such as IO ticks and queue time to track device performance.
• Distributed Storage Testing - Runs coordinated performance benchmarks across multiple remote servers using a single controller to evaluate network storage.
• Configurable I/O Workloads - Runs storage tests based on workload specifications defined in configuration files or command line arguments.
• I/O Access Pattern Simulation - The Distributed I/O Stress Tester selects which file to service next using strategies such as uniform random, round-robin, or sequential distributions.
• Queue Depth Control - The Distributed I/O Stress Tester manages the number of I/O units kept in flight and defines batch limits to test queue performance.
• I/O Job Specifications - Uses job files to specify global and thread-specific parameters required to simulate a complex I/O load.
• I/O Trace Replay - Provides an I/O trace replay engine to reproduce real-world storage traffic from recorded logs.
• Workload Stress Testers - Triggers complex synthetic I/O workloads and stress tests using direct command line arguments.
• Steady State Detection - Tracks performance over a rolling window and terminates the job once IOPS or latency stabilizes.
• Storage Deduplication Simulation - Generates unique data buffers and repeats them across a workload to test how systems handle redundant data.
• Synthetic I/O Workload Generation - Generates synthetic random or sequential data patterns to test how storage hardware handles specific traffic profiles.
• System Performance Benchmarking Tools - Measures the throughput and latency of storage devices and file systems using synthetic workloads.
• Distributed Benchmark Suites - Manages a controller client that orchestrates multiple server backends to execute benchmarks across remote machines.
• Storage Deduplication - Generates redundant data buffers to test the efficiency and performance of storage deduplication subsystems.
• I/O Rate Limiting - The Distributed I/O Stress Tester caps the maximum bandwidth or IOPS and exits the job if minimum performance thresholds are not met.
• Resource Group Constraints - The Distributed I/O Stress Tester assigns threads to specific control groups to constrain or prioritize disk I/O performance during benchmarks.
• Performance Metric Aggregators - The Distributed I/O Stress Tester combines output from multiple connections into a single report to simplify analysis across grouped workloads.
• Benchmark Thread Synchronization - Coordinates execution between threads using semaphores to ensure accurate parallel measurement of storage performance.
• Automation Job Coordination - Schedules jobs using start delays and runtime caps to ensure tasks finish in a specific, coordinated order.
• Configuration File Loading - Provides mechanisms to load and parse complex benchmark configurations from external files or standard input.
• CPU Core Timing Offloading - Offloads clock updates to a dedicated core to minimize timing overhead and maximize measurement accuracy.
• Buffer Pattern Generation - Fills buffers with patterns or random data and controls deduplication levels to test storage efficiency.
• I/O Trace Analysis - Reads and merges block-level I/O trace files to verify storage performance against device behavior.
• File System Access Patterns - The Distributed I/O Stress Tester informs the operating system about intended access patterns of a file to improve read-ahead and caching behavior.
• Deduplication Data Generators - Generates repeatable data buffers with specific deduplication levels to test storage compression and efficiency.
• Precision Benchmarking Timers - The Distributed I/O Stress Tester tracks elapsed time using the most precise available clock source to ensure accurate performance benchmarking.
• CPU Idleness Analysis - Measures CPU overhead by running idle-priority threads to derive the actual idleness of each processor.
• CPU Idleness Tracking - Tracks the time CPUs spend idle during execution to differentiate between storage latency and system inactivity.
• Rolling Window Analysis - Monitors IOPS and latency over a moving time window to detect when the storage device reaches a steady state.
• Metric and Performance Monitors - Generates graphical representations of storage data using axes and labels to analyze benchmark results.
• Performance Visualization - Generates bar and line graphs from collected data using custom labels and scaling to represent metrics.
• Batch Workload Execution - Runs multiple performance tests in sequence by processing a series of job definitions from a file.
• Physical Data Placement Mapping - The Distributed I/O Stress Tester maps data to physical storage locations using stream IDs or predefined schemes to benchmark specialized drives.
• Warm-up Period Management - The Distributed I/O Stress Tester controls the initial ramp-up phase of a workload to ensure measurements occur only after reaching a steady state.
• System Idle Time Measurement - Calculates the time required to complete units of work across multiple CPUs to profile system idle time.
• Data Integrity Checksums - Writes checksums or patterns to blocks and verifies them during reads to ensure data has not been corrupted.
• Data Placement Strategy Testing - The Distributed I/O Stress Tester determines which placement identifier to use for I/O operations using random, round-robin, or scheme-based selection.
• I/O Flow Control - The Distributed I/O Stress Tester regulates the volume of data moving through a job by monitoring thresholds and weights to prevent saturation.
• Performance Metric Exports - Outputs results in human-readable, JSON, or semicolon-separated formats including latency histograms and bandwidth logs.
• Flow Throttling - The Distributed I/O Stress Tester stalls threads or processes based on a weight-to-counter ratio to maintain a proportional distribution of IO.
• Network Testing - Flexible I/O tester for storage performance.

Star history