impress

Enterprise application server for Node.js: secure, lightweight, interactive, and scalable.

Description

First Node.js server scaled with multithreading and extra thin workload isolation. Optimized for high-intensive data exchange, rapid development, and clean architecture. Provides everything you need out of the box for reliable and efficient backend, network communication with web and mobile clients, protocol-agnostic API, run-time type validation, real-time and in-memory data processing, and reliable stateful services.

Weak sides: not a good choice for content publishing including blogs and online stores, server-side rendering, serving static content and stateless services.

Strong sides: security and architecture for enterprise-level applications, long-lived connections over websocket to minimize overhead for cryptographic handshake, no third-party dependencies.

Quick start

• See project template: metarhia/Example
• Start server with node server.js
• See documentation and specifications

API endpoint example: application/api/example.1/citiesByCountry.js

async ({ countryId }) => {
  const fields = ['cityId', 'name'];
  const where = { countryId };
  const data = await domain.db.select('City', fields, where);
  return { result: 'success', data };
};

You can call it from client-side:

const res = await metacom.api.example.citiesByCountry({ countryId: 3 });

Metarhia and impress application server way

• Applied code needs to be simple and secure, so we use sandboxing with v8 isolated contexts, worker threads and javascript closures;
• Domain code should be separated from system code; so we use DDD, layered (onion) architecture, DI, SOLID and GRASP principles, contract-based approach;
• Impress supports stateful applications with RPC and client-session sticky to servers; microservices, centralized or distributed architecture;
• No I/O is faster even than async I/O, so we hold state in memory, share it among multiple threads and use lazy I/O for persistent storage;
• We use just internal trusted dependencies, no third-party npm packages; total Metarhia technology stack size is less than 2mb.

Features

• Auto API routing, just create endpoint files as an async function;
• Code live reload with file system watch (when files change on disk);
• Graceful shutdown and application state recovery after reload;
• Minimal dependencies and code size;
• Can scale with multiple threads and servers;
• Code sandboxing for security and context isolation;
• Auto module loader with dependency injection for namespaces;
• Layered architecture out of the box: core, domain, API, client;
• Utilize multiple CPU cores and serve multiple ports with worker threads;
• Inter-process communication and shared memory used for state management;
• State synchronization mechanism with transactions and subscription;
• Cache server-side executable JavaScript in memory;
• Rapid API development support: AJAX RPC and Websocket;
• Serve static files from memory cache;
• Application configuration (for different named environments);
• Database access layer for PostgreSQL and relational db schemas;
• Persistent sessions support with authentication, groups, and anonymous;
• Multiple protocols: HTTP, HTTPS, WS, WSS;
• Logging with buffering (lazy write) and rotation (keep logs N days);
• File utilities: upload, download, streaming;
• Built-in simple testing framework;
• Server health monitoring;
• Built-in data structures validation and preprocessing library;
• Task scheduling (interval or certain time);
• Concurrency control: request queue with timeout and size;
• Execution timeout and error handling;

Requirements

• Node.js v12.9.0 or later (v14 preferred)
• Linux (tested on Fedora 30, Ubuntu 16, 18, 19 and 20, CentOS 7 and 8)
• Postgresql 9.5 or later (v11.8 preferred)
• OpenSSL v1.1.1 or later (optional, for https & wss)
• certbot (recommended but optional)

License & Contributors

Copyright (c) 2012-2021 Metarhia contributors. See github for full contributors list. Impress Application Server is MIT licensed. Project coordinator: <[e-mail]>

Changelog

[2.1.1][] - 2021-03-30

• Use metawatch instead of internal watcher
• Remove resources from memory on metawatch 'delete'
• Decompose class Application into Cache, Modules, Interfaces, Resources
• Fix domain and lib live reload bugs