An Implementation and Its Evaluation of a Framework for Managing States of Nodes among Structured Overlay Networks Kimihiro Mizutani * , Satoshi Matsuura *† , Shinichi Doi * , Kazutoshi Fujikawa *‡ and Hideki Sunahara *§ Email: kimihiro-m@is.naist.jp * Graduate School of Information Science, Nara Institute of Science and Technology † National Institute of Information and Communications Technology ‡ Graduate School of Information Science, Osaka University § Graduate School of Media Design, Keio University Abstract—Structured overlay networks are widely used as platforms for distributed systems. However, in a high churn situation, it is highly costly to maintain the structure and service availability of an overlay network because there is a limitation in churn handling. Several cooperation mechanisms among overlay networks are proposed to improve service availability of overlay networks in a high-churn situation. Several cooperation mech- anisms are proposed among overlay networks. The cooperation mechanisms can group structured overlay networks to simplify routing functions. However, the existing mechanisms do not consider managing churn information among overlay networks. In this paper, we propose a new framework that can share states of the nodes and spread churn information to neighbor nodes among different overlay networks. Using our framework, we are able to maintain each overlay network effectively and improve the reachability of messages under churn. We evaluate our framework with Chord, Kademlia, and Pastry. Our frame- work improves the detection time of churn by about 30% and decreases the stabilization messages by about 30-40%. We discuss combinatorial effects of the sharing state of the nodes among different overlay networks. Index Terms: Multiple Overlay Networks; Peer To Peer Networks; Churn Handling Management I. I NTRODUCTION Structured overlay networks have many famous service primitives and are applied to many peer-to-peer (P2P) ap- plications. For example, distributed hash tables (DHT) [1], decentralized object location and routing (DOLR) [2] and anycast and multicast (CAST) [3] are famous primitives. The popularity of P2P applications ranges from file sharing to conferencing and content distribution. When a user starts a P2P application, the user’s node joins an overlay network. The node contributes some resources while making use of the resources provided by others. The node leaves an overlay network when the user exits the application. Churn is the action of a node joining or leaving an overlay network. If churn occurs frequently, an overlay network will not maintain its structure and service availability. Consistently, a user of a P2P application cannot get target content. Churn Handling is one of the issues in implementing an overlay network. In typical churn handling, a source node sends a stabilization message to some target node and confirms whether the target node is alive or dead, by checking its response. When the target node does not reply to a message within timeout, the source node can detect the target node leaving. That is, the detection time of churn depends on the interval of stabilization messages. These days, a lot of overlay networks work concurrently under high churn rates on a real network. In such a situation, the overlay networks cannot maintain its structure because there is a limitation on churn handling by a single overlay network. To improve service availability of overlay networks, we focus attention on mechanisms of cooperation among overlay networks. Mechanisms of cooperation among overlay networks are used widely to improve the quality of P2P services [4]. A traffic management mechanism is a famous example [5], [6]. The management mechanism provides effective bandwidth management and loss assurance for overlay networks. As a result, the mechanism enables overlay networks to share network resources effectively and prevents network traffic from concentrating at a particular node. However, many mech- anisms of cooperation among overlay networks do not consider improving service availability of overlay networks against a churn situation. This is because each overlay network has its own message protocol and ID-space for managing churn information. For these reasons, there are many redundancies of churn detection among different overlay networks. As an example, figure 1 illustrates the redundancies in churn detection. Node-A and Node-B join overlay network-P and overlay network-Q. Then Node-A has two IDs: A’ and A”, which are published by each overlay network. Also, Node-B has two IDs: B’ and B”. A” sends a stabilization message to B” on overlay network-P at Time: T. A’ sends a stabilization message to B’ on overlay network-Q regardless of whether similar operation is done to the same node in overlay network- P within a short time: t. Such operation is redundant to detect the churn on a real network. In this paper, we propose a new framework integrating ID- space and churn handling functions among overlay networks for eliminating the churn detection redundancies. Our frame- work introduces Identifier Space Manager to general overlay networks. Identifier Space Manager links each ID-space to an Internet Protocol (IP) address as a common ID and provides functions of churn handling for overlay networks as common functions. Such functions can share states of the nodes among 2010 Sixth International Conference on Networking and Services 978-0-7695-3969-0/10 $26.00 © 2010 IEEE DOI 10.1109/ICNS.2010.46 282