documentation/content/docs/architecture/index.md
2023-08-19 10:50:16 +03:00

48 lines
1.6 KiB
Markdown

---
title: "Architecture"
weight: 4
summary: How is Piped architectured?
---
## How is Piped's architecture?
Piped has 3 components:
- [A frontend in VueJS](https://github.com/TeamPiped/Piped)
- [A backend in Java](https://github.com/TeamPiped/Piped-Backend) which uses [NewPipeExtractor](https://github.com/TeamNewPipe/NewPipeExtractor)
- [A proxy in Rust](https://github.com/TeamPiped/piped-proxy)
## Frontend
- Uses shaka-player for streaming.
- Uses a router for a single page application.
## Backend
- Uses Java 17
- Uses a JNI wrapper around [Reqwest](https://github.com/seanmonstar/reqwest), a Rust HTTP client.
- Uses ActiveJ to achieve maximum performance. Which is [really fast](https://web-frameworks-benchmark.netlify.app/result)
- Supports OpenJ9, and Hotspot
- Each running instance should configure their own proxy, thus allowing multi-gigabit content delivery.
- Uses ~70-130 MB of ram. (on OpenJ9)
## Database
- We currently support PostgreSQL, CockroachDB and YugabyteDB for high availability deployments.
# Proxy
- Uses Rust.
- Has HTTP/2 support.
- Uses [actix-web](https://github.com/actix/actix-web) and [reqwest](https://github.com/seanmonstar/reqwest) for maximum performance.
- Low memory footprint and high throughput.
- Converts `jpeg` images to `webp` on the fly to reduce bandwidth usage.
# Server-Side Caching
Caching is done at a Reverse-Proxy/CDN level to reduce the load to the backend. This also makes it more scalable.
# LBRY
LBRY streams are automatically used to stream content via LBRY if the same video is available there.