@flowrdesk/test-silo-engine

Independent benchmark suite for the Flowrdesk Silo Series.
Test and compare Silo, Pino, and Winston on your own hardware.

Why This Package Exists

We believe benchmark claims should be verifiable. Every number published about the Silo Series was generated using the exact tests in this package — on real hardware, with real disk I/O, with GC forced before every measurement.

Install this package, run the tests on your own machine, and see the results for yourself.

Install

npm install @flowrdesk/test-silo-engine

Pino and Winston are included as dependencies. No additional installs required.

Required: Always run your test file with node --expose-gc your_file.js for accurate memory readings.
For large log counts (100M+), add --max-old-space-size=12288 to increase available heap.

Quick Start

import { battleRoyale } from "@flowrdesk/test-silo-engine";

// Run all three engines head-to-head
// Default: 1,000,000 logs per engine
battleRoyale();

Run it:

node --expose-gc your_file.js

API

battleRoyale(logCount?)

Runs Silo, Pino, and Winston back-to-back under identical conditions. Same log payload, same flush guarantee, GC forced between each engine. Reports LPS, memory, and timing for each.

import { battleRoyale } from "@flowrdesk/test-silo-engine";

battleRoyale(); // 1,000,000 logs (default)
battleRoyale(10_000_000); // 10M logs

testSilo(logCount?)

Runs Silo in isolation.

import { testSilo } from "@flowrdesk/test-silo-engine";

testSilo(); // 1,000,000 logs (default)
testSilo(1_000_000_000); // 1B logs — allow ~15-28 minutes

testPino(logCount?)

Runs Pino in isolation.

import { testPino } from "@flowrdesk/test-silo-engine";

testPino(); // 1,000,000 logs (default)
testPino(25_000_000); // 25M logs

Note: Pino requires significant heap at scale. Use --max-old-space-size=12288 for runs above 10M logs.

testWinston(logCount?)

Runs Winston in isolation.

import { testWinston } from "@flowrdesk/test-silo-engine";

testWinston(); // 1,000,000 logs (default)
testWinston(25_000_000); // 25M logs

Note: Winston is significantly slower than Silo and Pino. A 25M log run takes approximately 8 minutes.

Example: Run All Four Tests

import {
  battleRoyale,
  testSilo,
  testPino,
  testWinston,
} from "@flowrdesk/test-silo-engine";

const LOG_COUNT = 10_000_000;

console.log("--- BATTLE ROYALE ---");
await battleRoyale(LOG_COUNT);

console.log("--- SILO SOLO ---");
await testSilo(LOG_COUNT);

console.log("--- PINO SOLO ---");
await testPino(LOG_COUNT);

console.log("--- WINSTON SOLO ---");
await testWinston(LOG_COUNT);

What Gets Measured

The Battle Royale test reports:

| Metric | Description | | ------------ | --------------------------------------------- | | lps - avg | The average logs per seconds over 5 runs | | best | The best logs per second out of 5 Runs | | worst | The worst Logs per second out of 5 Runs | | Mem - avg | The average memory used over 5 runs | | Total Time | The total time it took to complete all 5 runs |

The isolation test reports:

| Metric | Description | | ------ | -------------------------------------- | | time | Total elapsed time in seconds | | lps | Logs per second — throughput | | cpu | CPU usage during the run | | mem | Heap growth in MB from start to finish |

All measurements use process.hrtime.bigint() for timing and process.memoryUsage().heapUsed for memory. GC is forced before each run when --expose-gc is available.

Published Benchmark Results

Results from our reference hardware. Your numbers will vary based on CPU, disk speed, and available RAM.

Battle Royale — 100,000 Logs

| Engine | LPS - Avg | Best | Worst | Avg Memory | Total Time | | -------- | ----------- | ----------- | ----------- | ------------ | ---------- | | Silo | 506,538 | 606,405 | 296,617 | 86.80 MB | 1.20 secs | | Pino | 176,145 | 235,772 | 143,143 | 69.17 MB | 3.06 secs | | Winston | 71,443 | 73,157 | 69,766 | 22.83 MB | 7.10 secs |

Battle Royale — 1,000,000 Logs

| Engine | LPS - Avg | Best | Worst | Avg Memory | Total Time | | -------- | ----------- | ----------- | ----------- | ------------- | ---------- | | Silo | 457,754 | 585,949 | 377,351 | 317.86 MB | 11.33 secs | | Pino | 150,843 | 166,224 | 144,342 | 215.08 MB | 33.68 secs | | Winston | 75,505 | 78,067 | 73,560 | 137.74 MB | 66.44 secs |

Battle Royale — 10,000,000 Logs

| Engine | LPS - Avg | Best | Worst | Avg Memory | Total Time | | -------- | ----------- | ----------- | ----------- | --------------- | ----------- | | Silo | 403,827 | 411,095 | 395,535 | 308.89 MB | 123.91 secs | | Pino | 143,928 | 148,514 | 141,799 | 1,696.81 MB | 350.51 secs | | Winston | 80,044 | 80,864 | 79,468 | 1,301.36 MB | 625.37 secs |

Battle Royale — 25,000,000 Logs

| Engine | LPS - Avg | Best | Worst | Avg Memory | Total Time | | -------- | ----------- | ----------- | ----------- | --------------- | ------------- | | Silo | 420,836 | 445,849 | 393,658 | 296.18 MB | 297.77 secs | | Pino | 117,573 | 126,362 | 94,856 | 3,937.00 MB | 1,091.88 secs | | Winston | 54,294 | 64,584 | 40,603 | 3,244.94 MB | 2,376.68 secs |

Silo 100M and 1B results are available in the @flowrdesk/silo package benchmarks.

Log Files

Each engine writes to its own folder during testing:

• Silo → tests_silo/
• Pino → tests_pino/
• Winston → tests_winston/

These folders are safe to delete after testing.

Part of the Silo Series

This package is maintained by Flowrdesk LLC as part of the Silo Series ecosystem.

| Package | Description | | ------------------------------------------------------------------ | ----------------------- | | @flowrdesk/silo | The core logging engine | | @flowrdesk/test-silo-engine | This benchmark suite |

License

Apache-2.0 — see LICENSE for full text.
Copyright 2026 John Spriggs (Flowrdesk LLC)