Brocade: Landmark Routing on Overlay Networks

Abstract

Abstract. Recent work such as Tapestry, Pastry, Chord and CAN provide efficient location utilities in the form of overlay infrastructures. These systems treat nodes as if they possessed uniform resources, such as network bandwidth and connectivity. In this paper, we propose a systemic design for a secondaryoverlay of super-nodes which can be used to deliver messages directly to the destination's local network, thus improving route efficiency. We demonstrate the potential performance benefits by proposing a name mapping scheme for a Tapestry-Tapestry secondary overlay, and show preliminary simulation results demonstrating significant routing performance improvement.

Constructing a proximity-aware power law overlay network

“…Second, our algorithm builds "scale-free" power law topologies and assigns peers with different number of P2P connections according to their capacities. Compared to structured P2P systems [14,13,16] and their optimizations [19,21], our system is more resilient to network dynamic and is easier to implement. Our algorithm is fully distributed and based on only local information.…”

“…For unstructured P2P network, KaZaA [4] uses the notion of "supernode" and Gnutella v.0.6 [3] has "ultrapeer". In structured P2P network, such peers are referred to as "supernodes" in [21] and are organized into another layer of overlay called "expressway" [19] to accelerate the routing services. Those powerful peers are assigned with more workloads and serve as the "hubs" in the overlay network.…”

“…Those powerful peers are assigned with more workloads and serve as the "hubs" in the overlay network. In most of the cases [3,21], ordinary peers have only connections to the supernodes. Sometimes [19], ordinary peers may resort to other ordinary peers when the service of "expressway" is not available.…”

Distributed Hashtable on Pre-structured Overlay Networks

“…A number of recent studies have proposed substrateaware techniques to enhance particular DHT protocols. Zhao et al proposed to construct a secondary overlay (called Brocade) on top of existing DHT structures to exploit unique network resources available at each overlay node [35]. Ratnasamy et al introduced a distributed binning scheme for CAN such that the overlay topology resembles the underlying IP network [21].…”

“…In other words, they place protocolspecific requirements on overlay connectivity structures. As a result, substrate-aware link-selection enhancements built for these DHT protocols [3,21,35,34] cannot benefit other services.…”