Message Broker

As the final project in Bale Camp, We were asked to implement a fast and simple gRPC based broker. the main purpose of this project was to gain hands-on experience with golang and some common tools such as :

• Kubernetes:
  • Gained practical experience in utilizing Kubernetes for orchestrating workflows, ensuring smooth deployment, scaling, and management of containerized applications. -Docker: -Acquired hands-on knowledge of Docker for streamlined containerization, facilitating consistent deployment across diverse environments.
• Envoy:
  • Worked extensively with Envoy as a proxy server for load balancing and rate limiting, elevating the reliability and performance of applications.
• Prometheus:
  • Utilized Prometheus for in-depth monitoring, enabling real-time tracking of system metrics and proactive identification of potential issues.
• Grafana:
  • Developed proficiency in Grafana to craft insightful dashboards, visually representing system performance for effective decision-making.
• Jaeger:
  • Gained valuable experience with Jaeger for distributed tracing, offering insights into request flow across the system and aiding in debugging and optimization.
• gRPC:
  • Implemented gRPC for seamless communication between components, leveraging its high-performance RPC capabilities to amplify system speed and responsiveness.
• k6:
  • Applied k6 for rigorous performance testi
• Concurrency in golang:
  • Gained practical experience with concurrency features in golang, such as goroutines and channels.
• Data Storages:
  • Used different datasources like Postgres (SQL), Cassandra/Scylla (noSQL) to persist the messages. We compared the performance of each and decided the best use case for each one.
• Caching and Batching:
  • In order to overcome the performance bottlenecks, we took advantage of different methods such as caching and batching.

RPCs Interface

• Publish Requst

message PublishRequest {
  string subject = 1;
  bytes body = 2;
  int32 expirationSeconds = 3;
}

• Fetch Request

message FetchRequest {
  string subject = 1;
  int32 id = 2;
}

• Subscribe Request

message SubscribeRequest {
  string subject = 1;
}

• RPC Service

service Broker {
  rpc Publish (PublishRequest) returns (PublishResponse);
  rpc Subscribe(SubscribeRequest) returns (stream MessageResponse);
  rpc Fetch(FetchRequest) returns (MessageResponse);
}

Performance Report

In order to test our performance, We had several load testing phases on broker's publish service using K6 and custom-built golang clients.

RAM

Postgres

Cassandra