CSR: Medium: Collaborative Research: Foundations of Cache Network Operations for Content Delivery

Carnegie Mellon University, Pittsburgh PA

Investigators

Abstract

When a user goes online to read the news, watch videos, buy products, play games, or use a social network, the content accessed by the user is delivered to his/her device by a large Internet-scale network called a content delivery network (or, CDN for short). A CDN may consist of hundreds of thousands of servers located in thousands of data centers around the world. Operating a large CDN is challenging due to the sheer scale and diversity of the traffic and due to the vastness of their distributed server infrastructure. This research focuses on enabling highly efficient CDN operations. The key goals of this project are minimizing the operating cost and maximizing the robustness of the network. This project develops techniques for mapping different traffic classes to servers, so as to optimize resource usage, increase hit rates, and lower bandwidth costs. This project also develops adaptive and self-tuning systems that are robust to common sources of CDN variability, including object popularity/size, load, latency, and availability. Since CDNs are the key infrastructure enabling web, video, and other online applications, this project has a significant positive impact on the quality-of-experience of Internet users, while lowering the cost of Internet services. The project will broaden participation by involving undergraduates in the research. Further, the project will help attract women to STEM through outreach activities, including "Tech Nights" and "Roadshows" for Middle-School girls, and including semi-annual "Career Advice talks" to students from high school through grad school. The project will also incorporate topics in large-scale network operations in the curriculum. Finally, the project will invest significantly in outreach and tech transfer to the CDN industry by incorporating our research in production systems and enabling students to gain real-world experience in the CDN industry through internships. All products of the project, including research papers, software, and data, will be made available to the public. In particular, the software and data produced by the project will be archived for at least five years, and even longer when feasible. There will be two software and data repositories: one at UMass Amherst (https://www.cs.umass.edu/~ramesh/Site/CODE_DATA.html) and the other at CMU (http://www.cs.cmu.edu/~harchol/SoftwareDataRepository.html). This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.