Elsevier

Computer Communications

Volume 70, 1 October 2015, Pages 15-27
Computer Communications

Hierarchy-aware skip graph for sensing resource discoveries on large-scale sensor overlay networks

https://doi.org/10.1016/j.comcom.2015.06.021Get rights and content

Abstract

The rapid development of sensor networks has made it possible to build a large-scale sensor overlay network by integrating separated sensor networks all over the world. In our previous research, we proposed to realize sensor overlay networks by connecting their gateways with skip graph (SG) overlay, which is distributed and self-organized, and supports retrieving sensing resources using their properties. Although this method has proven to be effective in some scenarios, the lack of proximity awareness of an SG makes it difficult to build large-scale sensor overlay systems efficiently. In this paper, we present the hierarchy-aware extension of SG (HSG), which uses a hierarchy-aware neighbor selection mechanism that integrates the hierarchy of the physical Internet into overlay construction and routing to improve latency and traffic locality. To construct an HSG efficiently under various network environments, we develop a hierarchy-aware sampling mechanism for nodes to identify better neighbors than SG. To verify and validate the feasibility of HSG as a building block of large-scale sensor overlay networks, we carried out extensive simulations in near practical scenarios, which were obtained by analyzing an Internet measurement dataset. We showed that HSG could improve the routing latency and locality significantly with little overhead.

Introduction

Until now, research in the sensor network domain has mainly focused on routing, data aggregation, and energy conservation inside a single sensor network, whereas researchers have studied the integration of multiple sensor networks only to a limited extent. Owing to the recent technological advances on small, high performance, low power consumption CPU chips, memories with large storage and low cost, and high-speed mobile networks, we can soon expect the deployment of large numbers of autonomous sensor networks. To exploit fully the potential of sensing resources requires a feasible architecture integrating separated sensor network, managed by different organizations.

As the infrastructure of the integrated sensor networks, researchers widely accept connecting the gateways with a flexible overlay middleware that can hide the differences among different sensor networks [2], [3], [4], [5], [6], [7], [8], [9], [10], [11]. This kind of infrastructure can also be referred to as sensor overlay network. A sensor network may internally use arbitrary multi-hop, ad-hoc routing algorithms. Broadly speaking, a sensor can be any kind of data producer, for example, a real sensor, wireless camera, desktop computer running Twitter, or a combination of sensors. The gateway may be a powerful computer itself, or it could be connected to the cloud service. The gateways hide the differences among different sensor networks to applications. With the advent of generic programmable gateways, sensor network gateways have become considerably powerful and intelligent, which makes it easy to realize a sensor overlay network.

Although gateway technologies are developing fast, resource management and sensing resource retrieval of large-scale integrated sensor networks is still a challenging task because the number of sensor networks can be very large, and the owners of the sensor networks can decide to join or leave the network at any time independently. Therefore, it would take an enormous cost to deploy and maintain a centralized management system. To solve this problem, in [6], we proposed PIAX (P2P Interactive Agent eXtensions), which is a framework for realizing large-scale sensor overlay networks. It hides differences among different physical networks, and can connect various sensor networks using overlay technologies. One of the most important functionalities of PIAX is to discover sensing resources by their properties. The property could be the location of the gateway, the IP address of the gateway, or even the average readings for each day. The value of the property could be of any type as long as it is comparable. Realizing such functionality requires a range queriable P2P. In PIAX, SG [12] provides efficient range queries with low cost (Fig. 1).

The upper part of Fig. 1 shows such an overlay network developed by integrating the gateways of separated sensor networks. The numbers near them are the properties. As shown in the lower part of Fig. 1, all gateways are connected with an SG network. SG is a structured P2P overlay network for efficient key retrieval. It is different from DHTs, and enables both exact query and range query. Range query is a very important functionality as it allows users to find sensor networks without knowing their exact unique ID or address. For example, if an earthquake occurs in a certain area, users can find out the required sensor networks needed with a query condition such as “sensors in the area within 100 km from the epicenter of the earthquake”. Section 3.1 explained the functionality of SG.

Similar to most P2P technologies, SG is efficient in sensing resource retrieval; however, it suffers from a mismatch between the logical and physical topologies. Therefore, it is difficult to build large-scale sensor overlay networks that require low retrieval latency and high traffic locality with SG. In large-scale P2P systems, one hop between peers may cross multiple ASes (autonomous systems), countries, or even continents, which would dramatically decrease latency and locality, thus affecting system performance. In fact, nodes in the Internet have natural hierarchical properties such as the ISP networks that they access and the countries in which they are located. The hierarchical properties can estimate latency among nodes with coarse granularity. For example, the latency between nodes accessing the same ISP network is considered to be smaller, while the latency between nodes accessing different ISP networks is larger. Therefore, forwarding messages to nodes in the same ISP or AS can decrease the latency significantly. Furthermore, previous studies indicate that network connectivity failures are primarily due to border gateway protocol (BGP) faults [13]. As a result, node failures in overlay networks are not independent. Instead, nodes in the same ISP or AS tend to fail together. Therefore, improving routing locality will also help in fault isolation. Although introducing proximity into overlay routing has been extensively researched for DHTs, there is little research for range queriable P2Ps, such as SG.

To improve the SG for building a large-scale sensor overlay network, we propose a hierarchical neighbor selection scheme by considering the hierarchical properties and SG features of the nodes. The contribution of this work is as follows:

  • 1.

    We propose HSG, the hierarchy-aware extension of SG. In HSG, we extended the routing tables of SG to include an extra entry called H-Entry, which can reflect Internet hierarchical properties. In the routing process, H-Entry is used with high priority for improving latency and traffic locality.

  • 2.

    For HSG nodes to learn other node information for fulfilling the H-Entry, we propose a hierarchy-aware sampling method, which can work well in both a network with uniform key distribution and one with non-uniform key distribution.

  • 3.

    Aiming at realizing the building blocks of large-scale sensor overlays, we perform extensive simulation with near practical scenarios. The results show that HSG can improve both latency and locality with less overhead than SG.

The rest of this paper is organized as follows. In Section 2, we make a brief introduction of the related research. In Section 3.1, we give a brief instruction of SG. In Section 3.2, we propose HSG, which is the hierarchy-aware extension of SG, and discuss the exact-match query and range query algorithm in detail. In order to verify that we can use HSG for the sensing resource retrieval in large-scale sensor overlay networks, in Section 4, we conduct extensive simulations in a near practical scenario to validate and evaluate the performance and cost of HSG. The final section includes our conclusion and future works.

Section snippets

Related research

Broadly speaking, the related research includes three directions: sensor network integration technologies, proximity awareness in global-scale P2Ps, and various extensions of SG.

As for sensor network integration, IrisNet [2] was one of the first approaches that attempted to integrate different sensing resources for various applications. IrisNet is composed of sensing agents (SAs) and organizing agents (OAs). SAs collect and process data from their attached sensing devices, while OAs provide

Introduction of SG

SG is a distributed data structure for P2P applications to search for keys. Each node in SG has two fields: key and membership vector. Let m(u) denote the membership vector of node u. The elements of m(u) belong to a finite alphabet set Σ. We denote by p the inverse of the size of the alphabet; i.e., p=|Σ|1. We think of m(u) as an infinite word over Σ; but in practice, only an O(logN) length prefix of m(u) is needed on average, where N is the total node number.

There are multiple levels in SG,

Evaluation

Although an evaluation can also be carried out on a distributed testbed or by modifying the client software of a deployed P2P network, these approaches have limitations in our context. The limitations are exemplified by the PlanetLab testbed [42], which consists of about 800 nodes – too few for the analysis of large-scale sensor overlay network. Furthermore, the node distribution of PlanetLab may not appear to be representative for large-scale sensor overlay. Similarly, although many P2P

Conclusion

Rapid development of sensor networks requires novel technologies for building large-scale integrated sensor networks. Aiming at providing range queriable retrievals of sensing resource mechanisms, in this study, we proposed HSG, the hierarchy-aware extension of SG. HSG introduces physical Internet hierarchical properties into range queriable overlay construction and routing it in a simple and low cost manner. The results indicated that HSG offers a suitable solution for global sensing resource

References (44)

  • X. Shao et al.

    A low cost hierarchy-awareness extension of skip graph for world-wide range retrievals

    Proc. COMPSAC’14

    (2014)
  • T. Schütt et al.

    Range queries on structured overlay networks

    Comput. Commun.

    (2008)
  • C. Dannewitz et al.

    Hierarchical DHT-based name resolution for information-centric networks

    Comput. Commun.

    (2013)
  • J.B. Gibbons et al.

    IrisNet: an architecture for a worldwide sensor web

    IEEE Pervasive Comput.

    (2003)
  • P. Desnoyers et al.

    TSAR: a two tier sensor storage architecture using interval skip graphs

    Proc. ACM SenSys’05

    (2005)
  • K. Aberer et al.

    Infrastructure for data processing in large-scale interconnected sensor networks

    Proc. MDM’07

    (2007)
  • A. Antoniou et al.

    A peer-to-peer environment for monitoring multiple wireless sensor networks

    Proc. ACM/IEEE MSWiM’07

    (2007)
  • Y. Teranishi

    PIAX: toward a framework for sensor overlay network

    Proc. CCNC’09

    (2009)
  • Y. Ishi et al.

    Design and implementation of sensor data sharing platform for virtualized wide area sensor networks

    Proc. 3PGCIC’12

    (2012)
  • S. Clayman et al.

    INOX: a managed service platform for inter-connected smart objects

    Proc. IoTSP’11

    (2011)
  • Y.J. Kim et al.

    Service composition using new DSON platform architecture for M2M service

    Proc. ICOIN’12

    (2012)
  • L. Foschini et al.

    M2M-based metropolitan platform for IMS-enabled road traffic management in IoT

    IEEE Commun. Mag.

    (2011)
  • J. Kim et al.

    M2M service platforms: survey, issues, and enabling technologies

    IEEE Commun. Surv. Tutor.

    (2014)
  • J. Aspnes et al.

    Skip graphs

    Proc. ACM-SIAM SODA’03

    (2003)
  • D. Oppenheimer et al.

    Why do internet services fail, and what can be done about it?

    Proc. USITS’03

    (2003)
  • I. Stoica et al.

    Chord: a scalable peer-to-peer lookup service for internet applications

    Proc. SIGCOMM’01

    (2001)
  • A. Rowstron et al.

    Pastry: scalable, distributed object location and routing for large-scale peer-to-peer systems

    Proc. ICDSP’01

    (2001)
  • P. Maymounkov et al.

    Kademlia: a peer-to-peer information system based on the XOR metric

    Proc. IPTPS’02

    (2002)
  • A.R. Bharambe et al.

    Mercury: supporting scalable multi-attribute range queries

    Proc. SIGCOMM’04

    (2004)
  • J. Aspnes et al.

    Skip graphs

    ACM Trans. Algorithms

    (2007)
  • P. Ganesan et al.

    Canon in G major: designing DHTs with hierarchical structure

    Proc. ICDCS’04

    (2004)
  • M.S. Artigas et al.

    Cyclone: a novel design schema for hierarchical DHTs

    Proc. IEEE P2P’05

    (2005)
  • Cited by (4)

    • A virtual replica node-based flash crowds alleviation method for sensor overlay networks

      2016, Journal of Network and Computer Applications
      Citation Excerpt :

      Recently, sensor overlay networks with range queriable P2Ps as an overlay substrate have attracted interest because range queriable P2Ps such as Skip Graph (Aspnes et al., 2007) enable the retrieval of sensing resources using their properties (Desnoyers et al., 2005; Teranishi, 2009; Takeuchi et al., 2014; Shao et al., 2014, 2015a,b). In our previous work, we developed a sensor overlay network based on Skip Graph as presented in Fig. 1 (Teranishi, 2009; Shao et al., 2014, 2015a,b). The right side of the figure illustrates the architecture of one overlay node.

    • A sensor data stream collection scheme considering phase differences for load balancing

      2021, International Journal of Mobile Computing and Multimedia Communications
    • A bottom-up design for spatial search in large networks and clouds

      2018, International Journal of Network Management

    This paper is an extended version of the work originally to be presented at IEEE COMPSAC 2014 [1].

    View full text