Elsevier

Computer Networks

Volume 62, 7 April 2014, Pages 1-11
Computer Networks

An ant colony model based replica consistency maintenance strategy in unstructured P2P networks

https://doi.org/10.1016/j.bjp.2014.01.001Get rights and content

Abstract

To improve file availability and peers’ retrieval efficiency, many P2P networks adopt file replication techniques to distribute files and their replicas. Thus how to maintain the replica consistency has been an important research focus, especially in the P2P networks with writable and dynamically changeable files. Existing researches suffer from the shortcomings of either needing more update messages or having lower update success rates. This paper proposes a novel replica consistency maintenance strategy based on an ant colony model. We first define an ant’s actions and states, and then we use ants to update replicas in their walking forward and update file’s pheromones in their returning process. We discuss how to tackle the churn problem when an ant returns. The simulation results show that our strategy could reduce the update cost and the impacts caused by churn in replica consistency maintenance in unstructured P2P networks.

Introduction

P2P is a highly distributed and autonomous network. It eliminates the problem of single point of failure existed in traditional client/server systems. Any node could join and leave P2P network at any time, and also each node in a P2P network plays the both roles of receiving services as a client and providing services as a server. Generally speaking, P2P networks can be classified into two kinds: structured P2P network and unstructured P2P network. A structured P2P network, represented by DHT network, strictly defines the mapping relationship between a peer and its managed file indices, and thus it could achieve a high retrieval success rate. However, the structured P2P network could not adapt well to the environment with high churn rate. An unstructured P2P network is the one that all the nodes are loosely connected and no rigidly organized structure is needed. Unstructured P2P networks could well cope with the churn problem and thus are being widely applied, especially in the file sharing systems.

With the growth of P2P file sharing networks, file replication techniques are usually applied to improve the file availability and reduce the retrieval overhead. Since the replicas of a file are distributed in different nodes, how to maintain the replica consistency has been an attractive research topic [1]. Comparing with the structured P2P file sharing systems, the replica consistency maintenance in unstructured P2P networks is more difficult because there are no index nodes to manage the files’ indices.

The ant colony model is an optimization technique inspired by the foraging behavior of ant colonies [2], [3], [4], [5]. In the ant colonies, ants mark their paths with pheromone traces when they are walking between their nests and the food sources. Foraging ants make use of these traces to guide them to the food sources. A lot of researchers have found that the ant colony algorithm has the ability to achieve optimized solutions and it could be combined with other methods. Particularly, the ant colony algorithm could better adapt to the P2P networks with high churn rates [6].

In this paper, we propose a replica consistency maintenance strategy by employing an ant colony model in unstructured P2P networks. Our main work is the following.

  • (1)

    In order to make the best use of the ant colony model to resolve the replica consistency maintenance problem in unstructured P2P networks, we first demonstrate several definitions, such as the structure of an ant, the action of an ant and the state of an ant. These definitions could contribute to the design of the replica consistency maintenance strategy.

  • (2)

    We model the pheromone update process and present the algorithms for ant creation, ant sending, ant processing and ant returning, based on which a complete replica consistency maintenance strategy is proposed. In the strategy, we take different types of paths into account to make the update ants find out all the replicas to be updated as far as possible with a fewer number of update messages and a higher update speed.

  • (3)

    Since churn is a basic and inherent problem in P2P networks, our strategy takes churn problem into account to design the ant returning process so as to reduce the impact of churn on the update of file’s pheromone.

The remainder of the paper is organized as follows: Section 2 presents the related work on replica consistency maintenance problem in P2P networks; Section 3 demonstrates some important definitions. The replica consistency maintenance strategy based on the ant colony model is detailed in Section 4, and Section 5 describes the anti-churn strategy in the process of an ant returning. Section 6 evaluates the effectiveness of our strategy with simulations. Finally, Section 7 concludes the paper and gives our future research focus.

Section snippets

Related work

In order to improve file availability and reduce bandwidth consumption, researchers have presented many file replication schemes for P2P networks. Since the replicas of a file are distributed in different nodes in the network, for the applications with mutually writable files or the collaboration systems the file consistency maintenance problem must be considered seriously. For this reason, many replica consistency maintenance strategies have been proposed in P2P networks.

Because a structured

The ant colony model based definitions

As mentioned before, the ant colony algorithm is a soft computing technique inspired by the foraging behavior of ant colonies. According to the researches, the ant colony algorithm is suitable to be applied to search paths as well as to search resources [2], [3], [4], [5]. Though a single ant is simple in behavior, an ant colony could complete a complicated task. Also, since all the ants in an ant colony are cooperated to carry out a task, the loss of a single ant will not bring much impact on

An ant colony model for replica consistency maintenance in unstructured P2P networks

In this section, we present an ant colony model for replica consistency maintenance in unstructured P2P networks, which includes the ant’s creating, sending, processing and returning.

  • (1)

    Creating of an ant

When a file, say File, is updated in a node, it creates an ant to be used to propagate the update message. The algorithm, createAnt, is defined as follows:

createAnt(Peer, File, Updatecontent, TTL)
(1) Ant.Antheader.FileID = File.ID;
(2) Ant.Antheader.PeerID = Peer.ID;
(3) Ant.Antheader.Timestamp = 

Replica consistency maintenance strategy under churn

Churn is a basic and inherent problem in P2P networks, and it is caused mainly by node joining and leaving. Existing related researches [18], [19], [20], [21] usually make use of the ant’s foraging ability to achieve the optimized solutions, while pay less attentions to the churn problem, and thus they do not well fit P2P networks with high dynamicity. Due to the dynamicity of the P2P network, an ant bringing the update message may encounter a disconnected path. If the ant is walking forward, a

Simulation setup

We use the cycle-driven mode of Peersim [22], an open source P2P systems simulator, to conduct the simulations. In our simulations, the number of nodes in the network is 500 and TTL is set to 5. There are 1000 different files in the network. Each file and its replicas are randomly distributed in the network, and each node in the network could at most provide 40 shared files and there are no duplicated files in a node. Each simulation cycle creates and sends 50 update ants to maintain the

Conclusions

In this paper, we mainly discussed how to maintain the replica consistency in unstructured P2P networks based on an ant colony model. We first demonstrated several definitions as well as an ant’s actions and states, based on which we presented the replica consistency maintenance strategy. Then, we proposed the calculations of a file’s pheromone used to guide the update ant to the nodes with that file. Considering the fact that churn is a basic and inherent problem in P2P networks, we discussed

Xianfu Meng received his B.E. and M.E. degrees in Computer Science and Technology from Dalian University of Technology, China in 1983 and 1986, respectively. His current research interests include the peer-to-peer computing and the distributed systems.

References (22)

  • Y. Hu, M. Feng, L.N. Bhuyan, A balanced consistency maintenance protocol for structured P2P systems, in: INFOCOM, 2010,...
  • Cited by (9)

    View all citing articles on Scopus

    Xianfu Meng received his B.E. and M.E. degrees in Computer Science and Technology from Dalian University of Technology, China in 1983 and 1986, respectively. His current research interests include the peer-to-peer computing and the distributed systems.

    Changyuan Zhang is a graduated student and received the M.E. degree from Dalian University of Technology, China in 2013. His research interests focus on replica consistency maintenance in P2P networks.

    View full text