Biologically Inspired Adaptation of Routing Tables in Pastry DHT using Collaboration among Peers

摘要

本稿では，分散ハッシュテーブル(DHT）における経路表サイズの自己適応的な設定について議論する．経路選択においてクエリトラヒックと保守トラヒックはトレードオフの関係にあるため，経路表サイズの適切な設定は重要である．そこで本稿では，オーバレイネットワークのトラヒック圭全体を最小化するために，生物に着想を得た手法であるアトラクター選択を用いる．ピア間での協調した動作方法を提案し，各ピアの経路表サイズを進化的な方法で自己適応的に設定する．シミュレーションの結果，提案手法が優れた自己組織性をもたらすことが示された．%In this paper we discuss the self-adaptation of the routing table sizes of each peer in a DHT network with a biologically inspired method to obtain a better performance of the overlay network in terms of lower query traffic and maintenance traffic. Pastry, a well-known DHT algorithm, is used to construct the overlay topology and to provide the routing service. Our goal is to adaptively set the routing table sizes of each peer in a distributed manner such that the overall traffic of the overlay network is minimized. An adequate setting of the routing table sizes is crucial, since there is a tradeoff between query traffic at the cost of maintenance traffic. In order to find appropriate settings, we use the biologically inspired attractor selection scheme and extend our previous work by introducing collaboration among peers to control the adaptation in an evolutionary way. We show by numerical simulations that our proposed mechanism provides self-adaptability to changes in different parameter settings.