Runtime Manual/Core Systems/Modern Web Serving & Concurrency

Modern Web Serving & Concurrency

Standard Web Fetch API (bee serve), node:http, and Lockless Multi-Isolate Thread Pools

Beejs offers two high-performance paradigms for building web services:

Modern Standard Web Serving (bee serve [file]): Built on W3C / WinterCG standard Request / Response and export default { fetch(req) } model;
Node.js Compatible Serving (bee run server.ts): Built on node:http backed by a lockless multi-Worker thread pool in Rust Tokio.

1. Modern Web Application Serving (`bee serve`)

bee serve is the recommended, zero-overhead entrypoint for modern web applications, aligned with Cloudflare Workers, Deno, and Bun, featuring native TypeScript and JSX execution.

1.1 Writing Your First Web Service

Simply export an object with a fetch handler, or export the function directly:

code

// app.ts
export default {
  async fetch(req: Request): Promise<Response> {
    const url = new URL(req.url);

    // Route matching
    if (url.pathname === "/") {
      return new Response("🚀 Welcome to Beejs Web Server!");
    }

    if (url.pathname === "/api/echo" && req.method === "POST") {
      const data = await req.json();
      return new Response(JSON.stringify({ received: data, time: Date.now() }), {
        status: 200,
        headers: { "Content-Type": "application/json" }
      });
    }

    if (url.pathname === "/api/info") {
      return new Response(JSON.stringify({
        runtime: "beejs",
        version: "v1.16.0",
        arch: process.arch,
        platform: process.platform
      }), {
        headers: { "Content-Type": "application/json", "X-Powered-By": "beejs" }
      });
    }

    return new Response("Not Found", { status: 404 });
  }
};

1.2 Starting the Server

code

# Automatically detects and runs app.ts, app.js, server.ts, index.ts, etc.
$ bee serve

# Or specify a custom file, port, and host
$ bee serve app.ts --port 8080 --host 0.0.0.0

Console output:

code

🚀 Starting Beejs Web Server on http://0.0.0.0:8080
📄 Serving application: app.ts
✅ Listening on http://0.0.0.0:8080 (Ctrl+C to stop)

1.3 Key Highlights

Zero Glue Overhead: Directly bridges incoming HTTP packets to standard Request instances without unnecessary wrapper streams;
Async & Promise Support: Supports asynchronous handlers and drains microtasks automatically in the event loop;
Full Body Mixin: Request and Response fully implement req.text(), req.json(), and req.arrayBuffer();
Sandbox Integration: Compatible with --max-memory <MB> and --sandbox permissions to bound untrusted execution.

2. Classic Node.js Compatible Serving (`node:http`)

For legacy services or Express-style architectures, node:http works out of the box:

code

// server.ts - Classic Node.js style
import http from 'node:http';

const server = http.createServer(async (req, res) => {
  const { method, url } = req;
  const parsedUrl = new URL(url || '/', `http://${req.headers.host}`);

  res.setHeader('Content-Type', 'application/json; charset=utf-8');
  res.setHeader('Access-Control-Allow-Origin', '*');

  if (method === 'GET' && parsedUrl.pathname === '/api/users') {
    res.writeHead(200);
    res.end(JSON.stringify({ code: 0, data: ['Alice', 'Bob'] }));
    return;
  }

  res.writeHead(404);
  res.end(JSON.stringify({ error: 'Not Found' }));
});

server.listen(3000, () => {
  console.log('🚀 HTTP Server running: http://localhost:3000');
});

Run command:

code

bee run server.ts

3. Multi-Worker Thread Pool Architecture (`--workers`)

The Single-Thread Bottleneck

In conventional single-threaded runtimes, when a request triggers heavy JSON serialization, cryptography, or tensor inference, the event loop stalls and stalls all incoming requests.

Beejs Lockless Multi-Isolate Model

Beejs features a built-in multi-Worker thread pool in Rust:

code

                        Concurrent TCP Traffic
                                  │
                                  ▼
                 +──────────────────────────────────+
                 │      Main Dispatch Thread (Tokio)│
                 │   - TCP Listener & SO_REUSEPORT  │
                 │   - Lockless Cross-Thread Queue  │
                 +──────────────────────────────────+
                        │         │         │
            ┌───────────┘         │         └───────────┐
            ▼                     ▼                     ▼
+───────────────────+ +───────────────────+ +───────────────────+
| Worker 1 (Isolate)| | Worker 2 (Isolate)| | Worker N (Isolate)|
| - Dedicated Heap  | | - Dedicated Heap  | | - Dedicated Heap  |
| - Independent GC  | | - Independent GC  | | - Independent GC  |
+───────────────────+ +───────────────────+ +───────────────────+

Launching Workers

Specify the number of worker isolates with -W or --workers:

code

bee run --workers 8 server.ts

Or via environment variable:

code

export BEE_WORKERS=8
bee run server.ts

Benefits:

True Multi-Core Parallelism: Isolates execute independently in parallel on separate OS threads without blocking one another;
Zero Startup Penalty: Workers are pre-warmed at boot time, eliminating thread creation overhead per request.

4. Production Benchmarks & Best Practices

Benchmark with tools like autocannon or wrk:

code

# Launch with 8 workers
bee run --workers 8 server.ts

# Benchmark 100 concurrent connections for 10 seconds
npx autocannon -c 100 -d 10 http://localhost:3000/api/users

Optimization Tips

Prefer bee serve for Microservices: The Fetch API model avoids EventEmitter and streaming buffer wrapper overhead, yielding higher RPS;
Calibrate Workers: For I/O services, set workers to cores ~ 2 * cores; for CPU/tensor-heavy workloads, match the physical core count;
Enforce Resource Quotas: For public-facing endpoints, combine with --sandbox and --max-memory 512 to prevent memory leaks and unauthorized disk access.

PreviousNative Text Embeddings & Vectors NextSIMD & Memory Model