Persistent Memory for Content Delivery Networks
Timothy Miskell, Tushar Gohad, Ai Bee Lim, Felipe Pastor, Jeremy Payne, Greg Smith, Chris Cavigioli · NOMS 2022-2022 IEEE/IFIP Network Operations and Management Symposium · 2022
A Content Delivery Network (CDN) refers to a geographically distributed network of caching servers at the network edge that provides fast and efficient delivery of popular content. Video makes up at least 80% of the current Internet traffic. Consumer trends, such as high-definition video streaming across devices, growing IoT traffic, and data-intensive use cases requiring ultra-low latency transit are expected to make high-throughput content delivery increasingly important. As this content continues to get richer, more localized, personalized, interactive, and immersive, the volume of data required grows substantially. Therefore, the pressure is on CDN providers to deliver more content at a lower cost. CDN providers are increasingly looking for ways to optimize platform cost and performance to deliver to customer expectations for an exceptional experience without exorbitant capital and operational expenditures. Fortunately, advances in processor, memory and storage architectures are available to make the cost over efficiency equation easier to solve. In this paper, we show how AT&T* and Intel® have collaborated to demonstrate a foundational architecture for the next generation of open standards-based high performance CDNs; this includes benchmarks on 2nd and 3rd Generation Intel® Xeon® Scalable Processors with Intel® Optane™ Persistent Memory (PMem) technology as the elements to increase video stream density per CDN node, while lowering Total Cost of Ownership (TCO).