🚀 fast-thread

fast-thread is a high-performance module designed for ultra-fast object transmission between threads in Node.js, optimizing parallel processing with SharedArrayBuffer, Atomics, and fast-json-stringify.

🔥 Why fast-thread?

In Node.js, thread communication has inherent limitations:

• No native memory sharing like in other languages.
• All data must be serialized and sent via messaging, adding overhead.
• Complex objects cannot be shared natively, requiring costly serialization.

For high-performance applications, reducing serialization time directly improves parallel processing speed.

🚀 The Challenge: Finding the Fastest Serialization Method

We tested multiple serialization methods to find the fastest approach for inter-thread communication in Node.js.

| Serialization Method | Pros | Cons | |---------------------------|-------------------------------------------|---------------------------------------------| | JSON (JSON.stringify) | Built-in, no dependencies | Slower serialization, larger payload | | msgpack-lite | Compact binary format | Slower than JSON in V8 | | CBOR | Compact and structured binary format | Still slower than JSON | | BSON | Optimized for MongoDB, fast parsing | High overhead for small objects | | Protobuf.js | Compact, schema-based, high performance | Requires pre-defined schema | | fast-json-stringify | Faster than JSON.stringify with schema | Still requires copying in postMessage() | | fast-thread 🚀 | SharedArrayBuffer + Atomics (No Copy) | Requires structured memory handling |

After extensive benchmarking, the fastest approach was a combination of:

• ✅ SharedArrayBuffer for direct memory access
• ✅ Atomics.wait()/Atomics.notify() for ultra-fast synchronization
• ✅ fast-json-stringify for zero-copy, schema-based serialization

📊 Benchmark Results

Our tests measured message throughput (msg/sec) and bandwidth (MB/sec) over a 10s test period.

| Name | Messages | Messages Per Second | MB Per Second | |--------------------------|----------|---------------------|--------------| | fast-thread | 617,483 | 61,748.30 | 65.34 | | JSON | 524,235 | 52,423.50 | 55.48 | | fast-json-stringify | 500,024 | 50,002.40 | 52.10 | | BSON | 420,946 | 42,094.60 | 44.19 | | Protobuf.js | 296,340 | 29,634.00 | 29.75 | | msgpack-lite | 288,180 | 28,818.00 | 29.86 | | CBOR | 223,945 | 22,394.50 | 23.20 |

🚀 fast-thread achieved the best performance with a throughput of ~61,748 messages per second and 65.34 MB/sec.

⚡ Solution: SharedArrayBuffer + Atomics + fast-json-stringify

How It Works

• SharedArrayBuffer is used for zero-copy memory sharing.
• Atomics.wait()/Atomics.notify() provide fast synchronization between threads.
• fast-json-stringify eliminates JSON parsing overhead.

📌 Installation

pnpm install fast-thread

🛠 Example Usage

Worker Thread (worker_fast.js)

const { workerData } = require("worker_threads");
const fastJson = require("fast-json-stringify");
const { unpackObject, packObject } = require("fast-thread");

const sharedBuffer = workerData;
const stringify = fastJson({
    title: "Example",
    type: "object",
    properties: {
        id: { type: "integer" },
        name: { type: "string" },
        timestamp: { type: "integer" },
        data: { type: "string" }
    }
});

async function processData() {
    while (true) {
        Atomics.wait(sharedBuffer.signal, 0, 0);

        let obj = unpackObject(sharedBuffer);
        if (!obj) continue;

        obj.processed = true;
        packObject(stringify(obj), sharedBuffer, 1);
    }
}

processData();

Main Thread (test.js)

const { FastThread } = require("fast-thread");

(async () => {
    const fastThread = new FastThread("./worker_fast.js", 1024 * 1024)

    fastThread.on("message", (data) => {
        console.log("[Main] Processed data received:", data);
    });

    fastThread.on("terminated", () => {
        console.log("Thread closed");
    });

    for(let i = 0; i < 100; i++){
        fastThread.send({
            id: i,
            name: "User " + i,
            timestamp: Date.now(),
            data: "x".repeat(512)
        });

        await fastThread.awaitThreadReponse();
    }

    setTimeout(() => fastThread.terminate(), 2000);
})();

Since SharedArrayBuffer has a fixed size, it's essential to send messages sequentially, waiting for a response before sending the next job. This prevents buffer overwriting, ensuring each message is processed correctly.

Currently, no parallel message handling system has been implemented, meaning multiple messages cannot be processed simultaneously within the same thread instance.

To handle this limitation, the await fastThread.awaitThreadResponse(); call ensures that each message is processed before sending the next one. Without this mechanism, sending multiple messages at once could result in data loss or overwritten responses.

Main Thread (test-alternative.js)

const { Worker } = require("worker_threads");
const fastJson = require("fast-json-stringify");

const { 
    createSharedBuffer, packObject, 
    unpackObject 
} = require("fast-thread");

const sharedBuffer = createSharedBuffer();
const worker = new Worker("./worker_fast.js", { workerData: sharedBuffer });

const stringify = fastJson({
    title: "Example",
    type: "object",
    properties: {
        id: { type: "integer" },
        name: { type: "string" },
        timestamp: { type: "integer" },
        data: { type: "string" }
    }
});

packObject(stringify({ 
    id: 1, 
    name: "User A", 
    timestamp: Date.now(), 
    data: "x".repeat(512) 
}), sharedBuffer);

const checkResponses = () => {
    Atomics.wait(sharedBuffer.signal, 1, 0);
    const processedData = unpackObject(sharedBuffer, 1);
    console.log("[Main] Processed data received:", processedData);
    setImmediate(checkResponses);
};

📌 Conclusion

fast-thread enables blazing-fast thread communication in Node.js using SharedArrayBuffer, Atomics, and fast-json-stringify. This approach eliminates message serialization overhead, delivering unparalleled performance in high-throughput parallel workloads.