System Design Introduction
Torus is a learning engineering platform for authoring, delivering and improving online courses.
- Automated Course Project Lifecyle Authors can make edits and "push a button" to have those edits made available to end users
- Tight LMS Integration Torus is accessible from any LTI 1.3 compliant LMSes, and Torus defers to those LMSes roster management, gradebook and other functionality
- Run-time extensibility The system can have new content types and learning activities added to it at run-time by third party developers
- Simon Ecosystem Integration Torus will enable deep integrations with existing and future Simon Ecosystem projects
- Learning Engineering Centric UX Learning engineering science and best practies embedded in the UI/UX guide users through effective course creation
The following are the key non-functional requirements and considerations that fuel Torus technology and design choices:
- Usability Torus must be understandable and approachable by new users, especially those unfamiliar with learning engineering concepts
- Testability Quality assurance must be baked into Torus from the beginning and at multiple levels from unit testing, integration testing and performance testing
- Scalability Torus must scale to 10x to 100x the user load of the current OLI system, which at peak usage levels sees around 1,000 reqs/min
- Developer Productivity Torus developers must have access to a powerful, modern software development platform to maximum their productivity
All server side code in Torus is implemented in Elixir using the Phoenix web application framework.
Torus also powers client-side UIs with Elixir code via Phoenix LiveView.
The limited amount of pure client-side code in Torus is written in TypeScript and utilises React for UI.
Primary data storage uses Postgres RDBMS.
Torus is implemented in an architecture that resembles that of a traditional monolithic web-based application. There is a single application server that handles end user requests from web browser clients through a layered set of services that ultimately access the persistence layer.
A diagram of this architecture at a conceptual level is as follows:
Torus is designed to take advantage of Phoenix clustering support and Distributed Erlang to meet scalability and performance requirements. The various "Applications" that comprise Torus can be replicated across and run on any number of Erlang nodes that are part of the cluster. This allows Torus to mimic modern, scalable service based architectures by varying its deployed configuration.