Technical Requirements
This page gives a brief overview of the Metro Retro architecture and system requirements.

Introduction

Metro Retro is a dynamic, real time application providing users a rich interactive environment to work in collaboratively. It is primarily deployed into a virtual private cloud operated by Deqo Software Ltd and provided in the form of Software as a Service.
This article describes the new On Premises version of Metro Retro currently under pilot and it's technical requirements for installation.

Logical Topology

Metro Retro is exposed to customers as a single page web application. Behind this, there are a number of services working in synergy to provide a working service. From a logical perspective (ignoring hardware), these are:
Service
Product / Provider
Role
Web Gateway
Caddy
Acts as reverse proxy and load balancer.
Content Delivery Network
BunnyCDN
Provides edge cache for static assets & SPA build files.
App Server
NodeJS Application
Application API (HTTP based) and Auth/Login UI.
Podopolis
Elixir Application
Socket server that enables the real time collaboration functionality.
Main Database
Postgres
Stores all system data except for board contents.
Object Database
MongoDB
Stores board contents.
Cache
Redis
Provides a cache and short term storage solution.
Message Queue
Rabbit MQ
Enables inter-service communication between components.
File Store
S3 Compatible Storage
Stores static assets & user uploaded data.
App Worker
NodeJS Application
Background job processor for various system tasks/functionality.
Snapper
NodeJS Application
Generates board preview images via Puppeteer.
The diagram below shows the logical topology layout including the data transfer routes. Blue lines are one way transfers, purple lines are bi-directional transfers.

Physical Topology

Cloud Deployment

From a physical perspective, there are flexible options for deployment. Within the main cloud deployment, we utilize Digital Ocean with the following:
  • Ubuntu VMs for App Server, App Worker, Snapper, Podopolis, Rabbit MQ, Redis & Caddy
  • Digital Ocean Managed Postgres & MongoDB instances
  • Digital Ocean Spaces (S3 Compatible Storage)
  • Bunny CDN as the external CDN provider
Each component runs directly on a dedicated VM of a size applicable for the current system load.

Deployment On Your Infrastructure

The on-premises version of Metro Retro is intended to be run as an unmanaged service on private infrastructure. E.g. beyond initial installation we do not provide day to day operations or servers. The exact compute requirements will depend on the expected peak user count, which can be discussed as part of your engagement.
That aside, it is possible to run Metro Retro on a single machine, utilizing Linux containers, in a deployment that can easily support a few hundred users, potentially more. We use such a deployment for our test system and have an ongoing pilot at an institution with a similar setup.
We use mostly open Open Source Software for external dependencies (e.g. Rabbit, Redis, etc). The components that are exceptions to this are the CDN and S3 Compatible Storage. For CDN, a private CDN can be used, or alternatively the system is able to run without one. For S3, we recommend using MinIO, though a purely file based solution is possible (but not recommended).
We support OpenStack. We don't currently support Kubernetes, but are open to doing so in the future.

​

Last modified 4mo ago