Structured Superpeers: Leveraging Heterogeneity to Provide Constant-Time Lookup

  • Authors:
  • Alper Tugay Mýzrak;Yuchung Cheng;Vineet Kumar;Stefan Savage

  • Affiliations:
  • -;-;-;-

  • Venue:
  • WIAPP '03 Proceedings of the The Third IEEE Workshop on Internet Applications
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

Peer-to-peer (P2P) systems are typically divided intothose that centralize lookup functionality in a single locationand those that distribute the lookup operation acrossthe set of participating hosts. The former approach can offerconstant time lookup latency, but is more expensive toscale and suffers from single points of failure. In contrast,the fully distributed approach is easier to scale and can bemore resilient to failures, but the lookup latency scales asa function of the total number of participants. While theresearch community has made great progress in improvingthe latency of distributed lookup, these systems, exemplifiedby Chord [17], typically require O(log N) hops to locate anobject in a system with N hosts.In this paper, we explore the costs and benefits of a newhybrid approach that partially distributes lookup informationamong a dynamically adjusted set of high-capacity"superpeers". This design exploits the resource heterogeneityinherent in existing P2P systems to provide many ofthe advantages of a centralized system, even while avoidingmost of the problems associated with such systems. Lookupis performed using superpeers in constant-time, and the systemperforms well even in the event of simultaneous super-peerfailures. Finally, while our gain in performance is potentiallyat the expense of scalability, we will show that astraightforward implementation should be able to scale toover one million peers with reasonable lookup rates.