Elsevier

Computer Communications

Volume 52, 1 October 2014, Pages 60-70
Computer Communications

An economic replica placement mechanism for streaming content distribution in Hybrid CDN-P2P networks

https://doi.org/10.1016/j.comcom.2014.06.007Get rights and content

Highlights

  • Economic model of streaming content distribution in HCDNs is introduced.

  • A novel replica placement mechanism is introduced for streaming content in HCDN.

  • Asymptotic resource consumption of the solution is studied.

  • Performance of the solution is analyzed in comparison to other strategies.

Abstract

Hybrid CDN-P2P networks blend Content Delivery Networks (CDN) and Peer-to-Peer (P2P) networks to overcome their shortcomings. Replica placement in these networks is still an open problem. Hierarchical structure of these networks makes it inefficient to use available replica placement strategies specialized to CDN or P2P network domains. In this work, we introduce a novel replication strategy which enables traditional CDNs to offer Hybrid CDN-P2P streaming content delivery services. The proposed solution relies on economic model of the Hybrid CDN and employs a dynamic mechanism to optimize the number and places of replicas for P2P service. In particular, processing and communication complexities of the proposed solution are studied and its performance is analyzed in a typical Hybrid CDN-P2P network in comparison to related strategies.

Introduction

Hybrid CDN-P2P networks (HCDN) blend Content Delivery Network (CDN) and Peer-to-Peer (P2P) technologies. They benefit from the advantages of both architectures. CDNs benefit from the ease of management and are able to guarantee high Quality of Service (QoS) [1], but compared to P2P networks, they have higher costs and lower scalability [2]. P2P networks have low running costs and are more scalable, but it is not easy to guarantee their QoS and their management is more challenging [3]. Hybrid CDN-P2P networks combine these technologies to offer low priced high quality services [4].

In HCDNs, users can receive the service in either of Client–Server (CS) or Peer-to-Peer (P2P) approaches. In P2P mode, users need to contribute to content delivery using their upload bandwidth. In order to incentivize the end-users to contribute to P2P content delivery, HCDNs offer P2P service with higher quality compared with CS service [3]. Therefore, for streaming content distribution, an HCDN needs to replicate the stream with different qualities for CS and P2P end-users. In this work, by introducing a novel Replica Placement (RP) mechanism, a traditional CDN has been enabled to offer hybrid streaming content delivery services. It is assumed that the CDN employs its own replica placement strategy for distribution of the content in CS mode. Based on this assumption, a dynamic replica placement mechanism is designed to replicate the streaming content in P2P mode. This RP mechanism requires cooperation of edge servers and relies on communication of a few small messages to determine the best location of replicas for higher quality content distribution in P2P mode.

HCDNs receive the content from content providers and deliver it to end-users either in CS or P2P modes. They charge the content providers based on the delivered traffic and the Service Level Agreement (SLA), similar to traditional CDNs. In HCDNs, higher QoS in P2P mode encourages users to contribute in P2P content delivery using their upload bandwidth. Therefore, in P2P content delivery approach, the HCDN spends less bandwidth to deliver the content to end-users. As a result, in the SLA between the content provider and HCDN provider, price of content delivery in P2P mode must be kept less than traditional CS service. This is necessary to prevent the HCDN from reducing to a pure P2P service provider and to convince the content provider to request for the hybrid content delivery service. HCDN spends different amount of resources for CS and P2P end-users and charges the content provider differently for CS and P2P services. Therefore, in order to select the best number and places of replicas of P2P content, not only the content delivery costs and popularity of the content (like traditional CDNs) but also revenue of the HCDN must be taken into account.

CDNs, P2P networks and consequently HCDNs consist of thousands of servers and peers. This raises concern over the resource consumption of the RP mechanism. Moreover, high expected availability makes it mandatory to implement the RP as a distributed algorithm. The RP problem in this domain has many varying parameters like popularity of the content, prices of bandwidth and storage, and structure of the network. Therefore, the solution must consider many effective parameters but it must not spend much of valuable network bandwidth for transferring these parameters to and from the agents. It is also important to consume the minimum processing resources to solve the problem.

Existing studies of replica placement in HCDNs [6], [7], are confined to replication of static content and employ non-cooperative pull-based approach. In this work, we introduce a cooperative push-based replication strategy that extends the use of the HCDN to streaming content distribution, which is called Hybrid Replica Placement (HPR). To answer the resource consumption concerns, the economic model of the system is introduced. Then a hierarchical tree-based cooperative replica placement is proposed based on the Economic Mechanism Design Theory [5]. Experiments show that using HPR, it is possible to find the optimum placement of replicas. The major contribution of this paper is proposing an economic RP mechanism for replication of streaming content on a traditional CDN to offer Hybrid CDN-P2P streaming content delivery service.

The rest of the paper is organized as follows. We explore related works and compare the problem with similar problems in Section 2. In Section 3, the economic model of the HCDN is introduced. Section 4 is devoted to formal definition of the economic replica placement problem. In Section 5 the economic push based cooperative replica placement mechanism is introduced. The solution is studied through simulation and its outcome is compared with some other strategies in Section 6. Practical implications are discussed in Section 7. Finally the paper is concluded in Section 8.

Section snippets

Background and related work

Effectiveness of combining CDNs and P2P networks has been covered in earlier literature [8], [9]. In [10] the cost savings in Hybrid CDN-P2P architecture is studied. A survey of different architectures and models of Hybrid CDN-P2P networks is reported in [11]. In some works, e.g. [12], [13], [8], the architecture of the HCDN is inspired from P2P systems and the CDN part is employed to fill the missing blocks of the playback buffer. Authors in [14] have tried to introduce a resource scheduling

Economic model of the HCDN

Table 1 summarizes the notations used in this paper to model economics and performance of streaming content distribution in HCDN. Fig. 1 illustrates architecture of a Hybrid CDN-P2P network employed to define the replica placement problem. In these networks, the content provider signs an SLA with the HCDN provider and supplies the content i having two different qualities (BWCS,i and BWP2P,i) to distribution servers of the HCDN. The revenue in the contracted SLAi for distribution of content i

Optimal replication of P2P content over HCDN

We aim to add an RP mechanism to a traditional CDN and enable it to perform as an HCDN. Therefore, it is assumed that the set RepCS,i is determined by the existing RP mechanism of the CDN and it is considered as an input variable of our problem. We focus on optimizing economics of the system considering RepP2P,i as optimization variable. Consequently, we have:RepCostES,P2P,i=TransCostES,P2P,i+StrCostES,P2P,i=BWP2P,i·PBWSource,ES+SizeP2P,i·PSTRES,where BWP2P,i and SizeP2P,i denote the bandwidth

The proposed mechanism (HRP)

In order to design the RP mechanism of the HCDN (HRP), the Economic Mechanism Design Theory is employed. In this section, economic model of the problem which is called Hybrid Replica Placement Model (HRPM) is formalized mathematically. Subsequently, HRP is introduced formally using the popular notations of the Economic Mechanism Design Theory [5]. The major advantages of HRPM over popular cost models used in the literature of content distribution are twofold. First, it computes the complex net

Experimental evaluation

HRP is evaluated in two ways. Relying on a simulation test-bed, performance of HRP is studied. Additionally, to perform a comparative study on the performance and economics of HRP, five other replica placement strategies are implemented. These replication strategies include Greedy Global (GG), Single Greedy (SG), Hot-Spot (HS), Random (RND) and pure CDN. GG algorithm is conceived to yield near optimal outcome. SG solution differs from GG only in that SG just calculates the sorted list of

Practical implications

The performance evaluation of the HCDN architecture with HRP replica placement strategy reveals some features of the system. Experiments verify that in this architecture when a light load of requests is submitted to the system, performance and economics of the content distribution system is similar to a traditional CDN. When the load of the system increases significantly, advantages of hybrid content delivery becomes visible. Edge servers begin to replicate P2P content through HRP

Conclusion and future works

In this paper, we have presented an economic replica placement mechanism designed for streaming content distribution in HCDN environment. The proposed mechanism makes it possible to offer Hybrid CDN-P2P streaming content delivery service using a traditional CDN. In order to design the mechanism, economic model of the system (HRPM) has been introduced and a recursive hierarchical push-based cooperative replica placement mechanism is proposed. The mechanism has been formalized based on the

Acknowledgement

This research was supported by Iran Telecommunication Research Center (ITRC), Contract Number, 19251/500.

References (24)

  • Jussi Kangasharju et al.

    Object replication strategies in content distribution networks

    Comput. Commun.

    (2002)
  • Mukaddim Pathan et al.

    Content delivery networks: state of the art, insights, and imperatives

    (2008)
  • Imsook Ha et al.

    P2P, CDNs, and hybrid networks: the economics of Internet video distribution

    Int. Telecommun. Policy Rev.

    (2010)
  • Heather Yu et al.
    (2010)
  • Jiang Hai, Jun Li, Zhongcheng Li, Xiangyu Bai, Performance evaluation of content distribution in hybrid CDN-P2P...
  • Leonid Hurwicz et al.
    (2006)
  • Wang Zhan, et al., A k-coordinated decentralized replica placement algorithm for the ring-based CDN-P2P architecture,...
  • Fatemeh Keymasi Khalaji, Morteza Analoui, Replica placement algorithms in hybrid CDN-P2P architectures, in: 2012 Sixth...
  • F. Bronzino, et al., An adaptive hybrid CDN/P2P solution for content delivery networks, in: 2012 IEEE Visual...
  • Hao, Xuening Liu, Tongyu Zhan, Vyas Sekar, Feng Qiu, Chuang Lin, Hui Zhang, Bo Li Yin, Livesky: enhancing cdn with p2p,...
  • Cheng Huang, Angela Wang, Jin Li, Keith W. Ross, Understanding hybrid CDN-P2P: why limelight needs its own red swoosh,...
  • ZhiHui Lu et al.

    An analysis and comparison of CDN-P2P-hybrid content delivery system and model

    J. Commun.

    (2012)
  • Cited by (19)

    • Flexible replica placement for enhancing the availability in edge computing environment

      2019, Computer Communications
      Citation Excerpt :

      In the common traditional environment, these are many optimization schemes about replica placement have been proposed. Garmehi et al. [14] proposed a novel replication strategy which enables traditional CDNs to offer Hybrid CDN-P2P streaming content delivery services and employs a dynamic mechanism to optimize the number and places of replicas of P2P service. Ansari et al. [15] proposed the Latency Aware Replica Placement (LEARN) algorithm to place a number of replicas of each Avatar’s virtual disk into suitable cloudlets, which can be handed off among its cloudlets without migrating its virtual disk and reduce the average E2E delay.

    • An approach for designing cognitive engines in cognitive peer-to-peer networks

      2016, Journal of Network and Computer Applications
      Citation Excerpt :

      In recent years, this problem has become more challenging because the peer-to-peer networks have been merged into other systems such as cloud computing, content distribution networks, and social networks. It should be note that, in a hybrid system such as peer-to-peer cloud (Rajiv and Zhao, 2013; Amoretti et al., 2015), peer-to-peer content distribution network(Ryota and Fujita, 2013; Garmehi and Analoui, 2014; Garmehi et al., 2015), and peer-to-peer social networks (Kim and Lee, 2013; Kim and Lee, 2014) the number of problems in designing the management algorithms of peer-to-peer network will be increased. Therefore a new approach such as cognitive networking for solving the management problems in peer-to-peer networks can be useful.

    • Optimal design of Information Centric Networks

      2015, Computer Networks
      Citation Excerpt :

      In this way, given the fact that the content popularity is very skewed (i.e: few objects generate most of the traffic [12,13]), the operator can experience significant economic savings accountable to traffic offloading [14]. While CDNs may also be used to efficiently serve clients’ requests, they are usually regarded as an expensive solution, since they demand to centrally orchestrate replica placement and request routing [15], while in ICN each network device will autonomously perform these choices, thus reducing the overall management costs. To pave the way for a potential paradigm shift from a TCP/IP network to ICN, we specifically consider the migration step to the ICN architecture and we formulate a novel content-aware network planning model that we use to compute the optimal migration strategy for the operator.

    View all citing articles on Scopus
    View full text