A Serverless Publish/Subscribe System

TL;DR

This paper presents a novel serverless publish/subscribe system on IBM Bluemix that supports multiple matching types, including a new customizable function-based approach, demonstrating practical deployment despite scalability constraints.

Contribution

It introduces a serverless pub/sub architecture with a unique function-based matching method, enhancing flexibility and deployment simplicity on cloud platforms.

Findings

System supports topic, content, and function-based matchings

Function-based matching allows highly customizable subscriptions

Scalability limited by vendor-specific resource constraints

Abstract

Operating a scalable and reliable server application, such as publish/subscribe (pub/sub) systems, requires tremendous development efforts and resources. The emerging serverless paradigm simplifies the development and deployment of highly available applications by delegating most operational concerns to the cloud providers. The serverless paradigm describes a programming model where the developers break the application downs into smaller microservices that run on the cloud in response to events. This paper proposes designing a serverless pub/sub system based on the IBM Bluemix cloud platform. Our pub/sub system performs topic-based, content-based, and function-based matchings. The function-based matching is a novel matching approach where the subscribers can define a highly customizable subscription function that the broker applies to the publications in the cloud. The evaluations of the designed system verify the practicality of the designed system. However, the vendor-specific constraints of the IBM Bluemix resources are a bottleneck to the scalability of the broker.