Content Placement as a Key to Leveraging Geo-Distributed Infrastructures

The evolution of computing infrastructure points to a future with massively geo-distributed infrastructures deployed at hundreds or even thousands of locations. In parallel, the evolution of users’ traffic demands is making two challenges increasingly prominent on the Internet. First, a rapid growth in traffic, especially video traffic, is increasing the cost for Internet service providers (ISPs) who carry this traffic on their network backbone. Second, the Internet’s poor infrastructure support for mobility is increasingly visible at a time when end-users are predominantly mobile.

In this talk, I will show that leveraging geo-distributed infrastructures and more importantly, the flexibility of placing content on them, can help reduce the cost of carrying network traffic in ISPs as well as enhance support for mobility on the Internet. In the first part of the talk, we consider an ISP that also deploys a content delivery network across its point-of-presence locations. In such ISPs, we show that the flexibility to place content is so powerful that even simple placement schemes can singlehandedly achieve network cost and end-user performance close to an ideal scheme with future knowledge. Moreover, simple placement schemes provide these benefits even with a simple, static network routing, potentially simplifying the task of route optimization for such ISPs. In the next part of the talk, we present a global name service Auspice that supports secure, name-based communication between mobile entities on the Internet. Auspice addresses the challenge of rapidly resolving names to network locations under high mobility. To this end, its demand-aware replica placement engine intelligently replicates name records to provide low lookup latency, low update cost, and high availability. Auspice easily outperforms several commercial managed DNS providers as well as state-of-the-art research alternatives in our experiments.

Speaker Details

Abhigyan Sharma is a Ph.D. candidate in Computer Science at the University of Massachusetts Amherst. The recent and current topics of his research are network traffic engineering, content delivery, distributed data stores and datacenter energy efficiency. His work has appeared at prestigious conference venues such as SIGCOMM, SIGMETRICS and INFOCOM. He completed his B. Tech. in Computer Science and Engineering at Indian Institute of Technology Kharagpur in 2008.

Date:
Speakers:
Abhigyan Sharma
Affiliation:
University of Massachusetts Amherst
    • Portrait of Jeff Running

      Jeff Running