3R: Ensuring Sustainable Scalability in Globally Distributed Systems

  An appealing model for building large-scale distributed systems is a cooperative one in which nodes are 
expected to contribute resources in exchange for using the supported services. Beyond its potential for natural 
scalability and high performance, this cooperative approach can deliver the robustness of self-organization, as 
some nodes autonomously compensate for the failures of others. We are experimenting with this approach in a 
number of domains and applications, including data-sharing services and overlay multicast systems.

  The past few years have witnessed a growing number of cooperative, globally-distributed systems. Most of these
 systems are built following an overlay approach and under the assumption that no information is available about
 the underlying network. Thus, each of these systems regularly and independently probes its environment as it 
attempts, for instance, to route around problematic links or find good hosts to replicate content to ensure 
availability or durability. As this class of systems grows in popularity, the do-it-yourself approach will 
result in an unsustainable degree of network monitoring (and poor measurement accuracy) and restrict the 
variety, number and span of large-scale distributed services.

  In this talk, I will present some of my group's efforts in highly scalable, cooperative systems and introduce 
our newest project aimed at ensuring sustainable scalability across distributed systems. Our 3R approach is 
based on the thesis that a large fraction of globally-distributed systems can reduce their aggregated control 
and administrative overhead by strategically reusing/recycling the view of the network gathered by long-running,
 ubiquitous services such as CDNs and P2P systems.  In a recent SIGCOMM publication we show, for instance, that 
redirections by the Akamai CDN are primarily driven by network conditions, and we illustrate how these 
redirections can be used as hints for network detouring.

Bio:

Fabian Bustamante is an assistant professor in the EECS Department at Northwestern University. He joined 
Northwestern in 2002, after receiving his Ph.D. from the College of Computing at Georgia Tech. Fabian is a 
recipient of the National Science Foundation CAREER award. He is the head of the AquaLab group at Northwestern,
 which researches systems issues in large-scale distributed computing. Fabian’s research has been funded from 
various sources, including the NSF, Sun Microsystems, Microsoft and Motorola. He frequently serves as part of 
program committees in leading conferences including ICDCS, ICAC and WWW, and is the founder and co-chair for the
 new Workshop on Hot-Topics in Autonomic Computing Systems. For a list of publications and more detailed 
information, please visit: http://www.aqualab.cs.northwestern.edu.