WAE: Workload Automation Engine for CDN-specialized Container Orchestration

TL;DR

This paper introduces WAE, a modular workload automation engine for CDN container orchestration that optimizes resource management, reduces latency and costs, and improves efficiency based on real CDN data.

Contribution

We propose a novel modular system with an algorithm for optimal CDN function assignment, utilizing containerization to enhance performance and reduce costs.

Findings

Latency reduced by 45%

Deployment cost decreased by 66%

CPU efficiency increased by 20%

Abstract

Content Delivery Network (CDN) has been emerged as a compelling technology to provide efficient and scalable web services even under high client request. However, this leads to a dilemma between minimum deployment cost and robust service under heavy loads. To solve this problem, we propose the Workload Automation Engine (WAE) which enables dynamic resource management, automated scaling and rapid service deployment with least cost for CDN providers. Our modular design uses an algorithm to calculate the optimal assignment of virtual CDN functions such as streaming, progressive delivering and load balancer. In particular, we study on real CDN data which belongs to Medianova CDN Company in Turkey. Also we use Docker containerization as an underlying system. The results reveal that our containerized design reduces the latency and deployment cost by 45% and by 66%, respectively. Moreover, we obtain roughly 20% more CPU efficiency and 35% more utilized network.