Multi-level Distributed Cache

Project description

This distributed cache system is designed to support multiple clients that read and write data items stored in a database. There are 2 levels of cache nodes, arranged in a tree structure.

The system is implemented leveraging the Akka framework: clients, caches, and the main database are all modeled as actors within the Akka actor system.

Clients interact with the system through the cache nodes, which are responsible for processing read and write requests. These requests include basic operations namely Read and Write, as well as critical variants, Critical Read and Critical Write, each with specific guarantees.

Additionally, the system considers the possibility of cache crashes and implements a crash detection algorithm based on timeouts. Arecovery procedure is also implemented to restore the system to a consistent state after a crash.

The goal of the system is to maintain eventual consistency between the database and the cache nodes, even in the presence of crashes.

A web server was created to interact with the system, firing: client operations, cache crashes and recoveries, system consistency check.

Artifacts

Team and role

Team size: 2

• Designed protocols for each operation (Read, Write, Critical Read, Critical Write) along with the other team member.
• I proposed and implemented the request and response message mechanisms to correctly route messages through the system.
• I implemented Critical Write, Crash Detection and Recovery algorithms.
• I proposed and implemented the web server to interact with the system.

Tech stack