Benchmark Disk Write Speed in Python with tempfile

Benchmark raw disk write performance by writing a temporary file in 1MB chunks and measuring throughput in MB/s.

Easy Python 3.9+ Aug 9, 2026 Automation & scripting 12 views 0 copies

benchmark tempfile performance disk-io os-fsync

Python code

24 lines

Python 3.9+

import os
import tempfile
import time

def benchmark_write(size_mb=50):
    size_bytes = size_mb * 1024 * 1024
    chunk = b'x' * 1024 * 1024  # 1 MB chunk

    with tempfile.NamedTemporaryFile(delete=True) as tmp:
        start = time.perf_counter()
        written = 0
        while written < size_bytes:
            tmp.write(chunk[:min(len(chunk), size_bytes - written)])
            written += len(chunk)
        tmp.flush()
        os.fsync(tmp.fileno())
        elapsed = time.perf_counter() - start

    speed_mbs = size_mb / elapsed
    print(f"Wrote {size_mb} MB in {elapsed:.3f}s — {speed_mbs:.2f} MB/s")
    return speed_mbs

if __name__ == "__main__":
    benchmark_write(50)

Output

stdout

Wrote 50 MB in 0.145s — 344.83 MB/s

How it works

This works because tempfile.NamedTemporaryFile creates a real file on the filesystem, so writes go through the OS page cache and disk driver exactly like production I/O. Writing in fixed 1MB chunks avoids per-write overhead dominating the measurement. The flush() call pushes buffered data to the OS, and os.fsync() forces it all the way to physical disk — without these, the benchmark only measures userspace memory speed. Timing with time.perf_counter gives high-resolution wall-clock time, and the throughput formula converts elapsed time into MB/s.

Common mistakes

Forgetting `os.fsync()` — results then measure memory writes, not disk throughput
Writing in tiny chunks — per-write syscall overhead skews results downward
Using `delete=False` and leaving temp files on disk after the benchmark
Ignoring page cache warm-up on repeated runs

Variations

Use `pathlib.Path` and `open()` with a fixed filename to benchmark a specific target disk
Measure read speed separately with `f.read(chunk)` in the same loop

Real-world use cases

Validating SSD vs HDD performance before choosing storage for an application.
CI pipeline speed checks that fail if write throughput drops below a threshold.
Comparing filesystem performance for different mount options in production environments.

Benchmark Disk Write Speed in Python with tempfile

Python code

Output

How it works

Common mistakes

Variations

Real-world use cases

More from Automation & scripting

Tutorials

Quizzes

Python code

Output

How it works

Common mistakes

Variations

Real-world use cases

More from Automation & scripting

Keep learning

Tutorials

Quizzes