A Large Scale Key-Value Store Based on Range-Key Skip Graph and Its Applications Susumu Takeuchi 1 , Jun Shinomiya 2 , Toru Shiraki 3 , Yoshimasa Ishi 3 , Yuuichi Teranishi 1,3 , Mikio Yoshida 4 , and Shinji Shimojo 1 1 National Institute of Information and Communications Technology, Japan, {stakeuti,sshinji}@nict.go.jp 2 School of Engineering, Osaka University, Japan shinomiya.jun@ise.eng.osaka-u.ac.jp 3 Graduate School of Information Science and Technology, Osaka University, Japan {shiraki.toru,ishi.yoshimasa,teranisi}@ist.osaka-u.ac.jp 4 BBR Inc., Japan, yos@bbr.jp Abstract. An overlay network called Range-key Skip Graph (RKSG) has been proposed that can perform range-to-range search on peer-to- peer network. In this paper, we propose a large scale key-value store realized on RKSG, which is possible to scale-out a database with range query. Moreover, applications that utilize the store are implemented and will be demonstrated through private PlanetLab on JGN2plus. Keywords: P2P Overlay Network, Range-key Skip Graph, Key-value Store. 1 Introduction Recently, key-value store that allows a massive number of computers to have a large scale fast database is adopted in many systems. However, existing key-value stores only support to retrieve data by a single key, i.e., only exact matching is provided. In the ubiquitous environment, range retrieval that can extract data within a specific range of a key is required as well as exact matching. For example, a weather sensor application would need to retrieve the current observed data in each sensor located in a specific region, and a navigation application would need to retrieve movement histories of people in a specific region to recommend a route to go to a certain place. In order to realize key-value store that supports exact matching, DHT (Dis- tributed Hash Table), which is a structured P2P (Peer-to-Peer) overlay network such as Chord [1] and Kademlia [2], is effective for archiving high scalable one. Windows Azure 1 is one of the available implementations of DHT-based key-value store. However, DHT-based architecture cannot support range retrieval so that realizing key-value store which supports range query is difficult. Accordingly, a structured overlay network Skip Graph (SG) [3] is proposed to support range 1 Windows Azure Platform: http://www.microsoft.com/azure/windowsazure.mspx H. Kitagawa et al. (Eds.): DASFAA 2010, Part II, LNCS 5982, pp. 432–435, 2010. c  Springer-Verlag Berlin Heidelberg 2010