Papers and Talks

Papers

	A Practical Approach to Peer-to-Peer Publish-Subscribe. Ryan Peterson, Venugopalan Ramasubramanian, and Emin Gun Sirer. In ;login: 31(4):42-46, New York, New York, August 2006. PDF
	Corona: A High Performance Publish-Subscribe System for the World Wide Web. Venugopalan Ramasubramanian, Ryan Peterson and Emin Gun Sirer. In Proceedings of Networked System Design and Implementation (NSDI), San Jose, California, May 2006. PDF
	Optimal Resource Utilization in Content Distribution Networks. Yee Jiun Song, Venugopalan Ramasubramanian and Emin Gun Sirer. Cornell University, Computing and Information Science Technical Report TR2005-2004, Ithaca, New York, November 2005. PS PDF
	Perils of Transitive Trust in the Domain Name System. Venugopalan Ramasubramanian and Emin Gun Sirer. In Proceedings of Internet Measurement Conference (IMC), Berkeley, California, October 2005. PDF
	Client Behavior and Feed Characteristics of RSS, A Publish-Subscribe System for Web Micronews. Hongzhou Liu, Venugopalan Ramasubramanian and Emin Gun Sirer. In Proceedings of Internet Measurement Conference (IMC), Berkeley, California, October 2005. PDF
	The Design and Implementation of a Next Generation Name Service for the Internet. Venugopalan Ramasubramanian and Emin Gun Sirer. In Proceedings of SIGCOMM, Portland, Oregon, August 2004. PDF
	Beehive: O(1) Lookup Performance for Power-Law Query Distributions in Peer-to-Peer Overlays. Venugopalan Ramasubramanian and Emin Gun Sirer. In Proceedings of Networked System Design and Implementation (NSDI), San Francisco, California, March 2004. PDF
	Proactive Caching for Better than Single-Hop Lookup Performance. Venugopalan Ramasubramanian and Emin Gun Sirer. Cornell University, Computing and Information Science Technical Report TR2004-1931, Ithaca, New York, February 2004. PDF

Talks

• A Peer-to-Peer DNS.
  Operations Analysis and Research Center for the Internet Workshop, Seattle, November 2006.
  
  Using CoDoNS as an authoritative name service as well as a safety net for legacy DNS with provably good performance, stability, and failure resilience in face of flash crowds and denial of service attacks.

• CoDoNS: Replacing the DNS Hierarchy with Peers.
  DNS Operations Pre-NANOG Workshop, OARCI, San Jose, June 2006.
  
  How CoDoNS may be used to supplant the existing Domain Name System, DNS, to provide high performance, improve failure resilience and thwart denial of service attacks.

• Perils of Transitive Trust in the Domain Name System.
  Invited Plenary Talk: RIPE-52, Istanbul, Turkey, April 2006.
  ACM SIGCOMM Internet Measurement Conference (IMC), Berkeley, California, October 2005.
  
  The Domain Name System, DNS, is based on nameserver delegations, which introduce complex and subtle dependencies between names and nameservers. In this talk, I present results from a large scale survey of DNS, and show that these dependencies lead to a highly insecure naming system. This talk focuses specifically on three aspects of DNS security: the properties of the DNS trusted computing base, the extent and impact of existing vulnerabilities in the DNS infrastructure, and the ease with which attacks against DNS can be launched.

• Self-Organizing Systems for Robust, Large-Scale Infrastructure Services.
  University of Maryland, College Park, Maryland, July 2005.
  
  In this talk, I will present some recent work on building new, robust infrastructure services based on self-organizing, peer-to-peer architectures. In particular, I will describe Herbivore, a system based on Dining Cryptographer Networks for anonymous communication that provides strong anonymity, scalability, and performance. In addition, I will touch upon some recent work in the Honeycomb project, where we are building new infrastructure applications, such as a replacement for DNS, a new CDN and a new pub-sub system, that provide high-performance in the presence of highly-skewed distributions, traditionally considered to pose difficult problems for networked services.

• Building High Performance Infrastructure Services through Analysis-driven Caching.
  Microsoft Research Cambridge, Cambridge, United Kingdom, May 2005.
  Microsoft Research Silicon Valley, Mountain View, California, May 2005.
  Brown University, Providence, Rhode Island, April 2005.
  University of Chicago, Chicago, Illinois, April 2005.
  Rutgers University, Piscataway, New Jersey, April 2005.
  
  Caching in distributed systems has been traditionally performed in an ad-hoc manner; nodes opportunistically cache objects that they encounter using heuristics for cache management. In this talk, I will describe a novel form of caching, where objects are proactively replicated based on an analysis of performance trade-offs. Analysis-driven caching poses the fundamental trade-off between performance and overhead as an optimization problem, determines the optimal caching strategy through numerical algorithms, and replicates objects using inexpensive distributed protocols. I will present Honeycomb, an analysis-driven caching framework for structured overlays, which provides an order of magnitude improvement in lookup performance, from O (log N) to O (1), with minimal network and storage overhead.

  Honeycomb has been used to build three infrastructure services, a name service to replace the current Domain Name System (DNS), a content distribution network, and a publish-subscribe based aggregator for Web MicroNews. In this talk, I will present performance evaluations of these services deployed on Planet-Lab to show that they provide better lookup latency than legacy services, adapt quickly to sudden upheavals in query distribution, and quickly disseminate updates to objects.

• CoDoNS: Fast, Resilient, and Scalable Name Service for the Internet.
  NEC Laboratories America, Princeton, New Jersey, December 2004.
  Invited Speaker: IEEE Computer Communications Workshop, Bonita Springs, Florida, October 2004.
  
  Name services are critical for mapping logical resource names to physical resources in large-scale distributed systems. The Domain Name System (DNS) used on the Internet, however, is slow, vulnerable to denial of service attacks, and does not support proactive updates. These problems stem fundamentally from the structure of the legacy DNS.

  In this talk, I will describe the design and implementation of the Cooperative Domain Name System (CoDoNS), a novel name service, which provides high lookup performance through proactive caching, resilience to denial of service attacks through automatic load-balancing, and fast propagation of updates. CoDoNS derives its scalability, decentralization, self-organization, and failure resilience from peer-to-peer overlays, while it achieves high performance using the Beehive replication framework. Cryptographic delegation, instead of host-based physical delegation, limits potential malfeasance by namespace operators and creates a competitive market for namespace management. Backwards compatibility with existing protocols and wire formats enables CoDoNS to serve as a backup for legacy DNS, as well as a complete replacement.

  I will present performance measurements from a real-life deployment of the system in PlanetLab to show that CoDoNS provides fast lookups, automatically reconfigures around faults without manual involvement and thwarts distributed denial of service attacks by promptly redistributing load across nodes.

• CoDoNS: A High-Performance, Fault-Resilient Replacement for the Legacy Domain Name Service.
  Carnegie Mellon University, Computer Science Department, November 2004.
  
  The Domain Name Service (DNS) is based on a static, manually constructed hierarchy of distributed nameservers. This organization makes the legacy DNS slow, vulnerable to denial-of-service attacks, and prone to load imbalance.

  In this talk, I will describe an alternative architecture for a new Domain Name Service that provides high performance, DoS-resilience, and automatic load balancing in the presence of flash crowds and the "slashdot effect." This talk will focus on the core contribution, proactive caching, which enables scalable peer-to-peer overlays to respond to queries drawn from a Zipf-distribution in under a single hop with a minimal number of object replicas, and illustrate its application to DNS. The resulting system, CoDoNS, has been deployed throughout the globe over Planetlab and can serve either as a high-performance replacement for legacy DNS or as a backup in case legacy DNS comes under attack.

• Self-Organizing Networks.
  NSF STC: TRUST Center, University of California, Berkeley, September 2004.
  
  Self-organization is a required property for the types of large-scale, complex networked systems we are likely to encounter in the future. In this presentation to the NSF site review panel, I outline MagnetOS, a self-organizing operating system for ad hoc and sensor networks, and CoDoNS, a self-organizing name service for wide area networks.

• The Design and Implementation of a Next Generation Name Service for the Internet.
  SIGCOMM, Portland, Oregon, August 2004.
  
  CoDoNS is a safety net and replacement for the legacy DNS system. It provides high performance by intelligently utilizing bandwidth and storage to replicate popular objects on top of a peer-to-peer, self-organizing overlay.

• CoDoNS: A Next Generation Name Service.
  DNSSEC Steering Committee, Washington, DC, August 2004.
  
  In this talk, I describe CoDoNS, and how it can help speed up the adoption of DNSSEC. DNSSEC has been still-born; few people use it, despite its obvious advantages. The US government is busy deploying DNSSEC-enables nameservers for names in its jurisdiction. My thesis is that CoDoNS, which attaches DNSSEC signatures to names obtained via legacy DNS, can help provide signatures even for those domains that were not signed in the first place. This enables clients to enjoy the benefits of secure, tamperproof name service, and provides a stepping stone towards wider adoption of DNSSEC.

• Operating Services 7/24: Domain Name Service for the Next Generation Internet.
  Planetlab Workshop, Palo Alto, California, April 2004.
  
  In this talk, I provide an overview of the CoDoNS system, how it is deployed on PlanetLab, and how it can drastically change the way we think of and implement name services by liberating names from the servers that provide them. This breaks the monopoly that namespace operators enjoy over top level domains, and makes possible a new kind of service.

• CoDoNS: A High-Performance, Fault-Resilient Replacement for the Legacy Domain Name Service.
  University of Pennsylvania, Philadelphia, Pennsylvania, March 2004.
  Microsoft Research Silicon Valley, Palo Alto, California, March 2004.
  
  The Domain Name Service (DNS) is based on a static, manually constructed hierarchy of distributed nameservers. This organization makes the legacy DNS slow, vulnerable to denial-of-service attacks, and prone to load imbalance.

  In this talk, I will describe an alternative architecture for a new Domain Name Service that provides high performance, DoS-resilience, and automatic load balancing in the presence of flash crowds and the "slashdot effect." This talk will focus on the core contribution, proactive caching, which enables scalable peer-to-peer overlays to respond to queries drawn from a Zipf-distribution in under a single hop with a minimal number of object replicas, and illustrate its application to DNS. The resulting system, CoDoNS, has been deployed throughout the globe over Planetlab and can serve either as a high-performance replacement for legacy DNS or as a backup in case legacy DNS comes under attack.

• Beehive: O(1) Lookup Performance for Power-Law Query Distributions in Peer-to-Peer Overlays.
  Symposium on Networked Systems Design and Implementation (NSDI), San Francisco, California, March 2004.
  
  Structured peer-to-peer hash tables provide decentralization, self-organization, failure-resilience, and good worst-case lookup performance for applications, but suffer from high latencies (O($log N$)) in the average case. Such high latencies prohibit them from being used in many relevant, demanding applications such as DNS. In this paper, we present a proactive replication framework that can provide constant lookup performance for common Zipf-like query distributions. This framework is based around a closed-form optimal solution that achieves O($1$) lookup performance with low storage requirements, bandwidth overhead and network load. Simulations show that this replication framework can realistically achieve good latencies, outperform passive caching, and adapt efficiently to sudden changes in object popularity, also known as flash crowds. This framework provides a feasible substrate for high-performance, low-latency applications, such as peer-to-peer domain name service.

Beehive

Computer Science Department
Cornell University