Distributed Filtering

A project for the Distributed System Enviroment course (2023) at BUT FIT introducing distributed computing using Microsoft Orleans framework.

The application provides a means to distribute work - image filtering - across a cluster. A cluster consists of workers (clients) connected to a server. The server handles worker disconnection (via heartbeat every 2 seconds) and automatically redistributes unfinished work to other available workers. The client will be assigned work on connection, if available.

The server provides an API to:

• initiate a job (apply bilateral filter/add gausian noise) - divide an image into batches of work and distribute them to workers.
• retrive status/progress of the current job and currently working clients
• cancel current job

Note: server runs swagger for API manipulation.

Running

Premise: the application has been build (dotnet build DistributedFiltering.sln).

Run server with the Orleans cluster:

dotnet DistributedFiltering.Server.dll --Cluster:Address=192.168.0.1

Server will initiate silo (cluster) at the address 192.168.0.1, when unspecified, server will use address from the appsettings.json (by default 127.0.0.1). This will also run ASP.NET Core Web API with metioned endpoints.

Run client:

dotnet DistributedFiltering.Client.dll --Cluster:Address=192.168.0.1

Client tries to connect to the Orleans cluster at the address 192.168.0.1, when unspecified, client will use address from the appsettings.json (by default 127.0.0.1). If the client cannot connect to the server or the server crashes, the client crashes.