Docker and kubernetes : deploying speech recognition system for scalability

Abstract

This project aims to provide a robust solution for deploying an existing automatic speech recognition (ASR) system. The deployment solutions will enable the system to handle multiple requests concurrently and support dynamically changing workload. The deployment solution is constructed using containerisation deployment technology: Docker and Kubernetes. The solution is implemented on Amazon Web Service (AWS) cloud platform and mainly use Elastic Kubernetes Service (EKS) on the platform. The solution ensures the auto-scaling of the system and even load balancing between the requests to computational resources to provide higher availability, and ensures the security of the application and the computational resources. This report will present the solution in terms of the designed architecture diagram, detailed illustration of important components and the implementation steps. The report will also demonstrate the experiments result that showing the performance of the solution and evaluate the cost to implement the solution.