Distributed Matrix Multiplication

This project involves the use of REST interface and gRPC based services to allow matrix multiplication to be split among multiple machines. Block Matrix multiplication is a method for multiplying matrices which is useful in distributed systems as it can easily split the calculations among mutiple machines.

Architecture

• This project involves 3 components
  • a REST interface to upload matrix as input files. I'm using FastAPI for this.
  • Async gRPC Servers to do the matrix multiplication and addition
  • Async gRPC Clients

Setup with Docker

• Run docker-compose build to build images
• Running docker-compose up should run 1 REST container and 3 gRPC server containers.
• Visit http://0.0.0.0:8010/docs to access the API

Setup without Docker

• Install the dependencies
  • pip install -r gRPC/requirements.txt
  • pip install -r REST/requirements.txt
• Run 3 gRPC Servers on different ports
  • Sample: python async_grpcServer.py --port=8000
• Modify the client listeners in REST/async_grpcClient.py file to point to the running gRPC servers.
• Run FastAPI app
  • uvicorn async_app:app --port=8010
• Visit http://0.0.0.0:8010/docs to access the API