Elsevier

Computer Communications

Volume 76, 15 February 2016, Pages 54-66
Computer Communications

Network delay guarantee for differentiated services in content-centric networking

https://doi.org/10.1016/j.comcom.2015.09.009Get rights and content

Abstract

The newly adopted built-in caching mechanism guarantees efficient content delivery for content-centric networking (CCN) as compared to the existing IP-based networks such as the Internet. However, it is a challenge at the same time for CCN to meet QoS requirements due to content caching. In this paper, we investigate the problem of providing network delay guaranteed services in CCN. More specifically, we study the problem of meeting network delay requirements for differentiated services (content providers) in CCN while at the same time optimizing the overall content delivery performance.

To support delay guarantee, we first present a simple and holistic network model which characterizes network delays of routing content to clients at different locations. By aligning network locations with content popularity, we ensure that each content provider has an optimized network delay of routing content to clients. We then derive analytical network delays for content providers by incorporating their content distribution models into the proposed holistic network model, and further formulate the delay guarantee task as a nonlinear integer programming (NIP) problem under the given network resources and traffic access patterns. We evaluate our mechanism and investigate the optimized network performance using different real/synthetic network topologies. With numerical studies, we analyze the process of competing for the network resources by different content providers, and investigate how various factors (e.g., content popularity, traffic volume, router storage capacity) affect this competition process. Our models and results presented in this paper provide guidance in designing resource provisioning and QoS mechanisms for CCN.

Introduction

Driven by the huge volume of content (e.g., video, audio, images), the usage of the Internet is increasingly focused around content delivery. Today users tend not to care where and how to obtain the content, but are more interested in fast and reliable content delivery. Moreover, content over the Internet is expected to grow even faster, i.e., it is believed that global IP traffic will increase threefold over the next 5 years [1]. This poses significant challenges for the Internet due to the mismatch between its host-to-host communication paradigm and the current content-oriented usage.

To address these challenges, content-centric networking (CCN) [2], [3], [4], [5], [6], [7], [8] as a clean-slate approach is proposed. CCN tackles the challenges by adopting two new mechanisms, namely, name-based routing and systematic in-network caching. Name-based routing refers to the mechanism that every piece of content is identified by an addressable name and requests for the content can be routed by network. As a result, users of CCN issue requests for the content (expressed as interests), and the network takes care of locating and retrieving the data. This naturally realizes the so-called location-independent (or location-oblivious) content delivery.

Meanwhile, to provide users with efficient content delivery, CCN employs systematic in-network caching. Each CCN router can store the requested content in its local cache and then use the previously forwarded data to satisfy future requests. By typically storing popular content objects at the router, in-network caching guarantees CCN to have lower bandwidth consumption, less congestion and faster response time to content fetching.

However, in-network content caching also potentially raises many new challenges in both understanding and utilizing the built-in network caching capability. Typical research problems include modeling and analysis of system dynamics under different caching hierarchies and with different cache replacement policies (e.g., LRU, RND, FIFO) [9], [10], [11], [12], [13], provisioning en-router content storage for network performance optimization [14], [15], [16], [17], etc. While most previous work focus on these topics, in this paper we go one step further to explore QoS (Quality of Service) guarantee in CCN. More specifically, we investigate the problem of guaranteeing network delays for differentiated services (content providers) in CCN while at the same time optimizing the overall content delivery performance. This is a significant task for both network administrators/operators and service providers as network delay is a key metric of QoS due to the nature that different kind of content has different network delay requirements. For example, voice and videos are far more sensitive to long network latency than web and emails.

Guaranteeing network delays for differentiated services (content providers) in CCN is a new research problem. While most of existing mechanisms for supporting delay sensitive traffic in network are designed as end-to-end semantics [18], [19], [20], in CCN the concept of “end-to-end flows” or “connections” do not even exist. As a result, existing mechanisms for guaranteeing network delays are no longer applicable in the context of CCN.

Meeting network delay requirements for differentiated services in CCN is also challenging, mostly due to the following factors. First, end-users are generally distributed across network at different locations and have different delays of fetching content objects from content providers. For example, in a network with a tree-like topology, users located at lower-layer nodes often have longer delays of fetching content than those connected at upper-layers. To meet delay requirements for end-users with different locations, the network topology information should be taken into account and the delay guarantee mechanism needs to properly handle this user location diversity. Second, the request access profiles of end-users (e.g., request rate, content distributions) are not always consistent and are changing over time. This also raises significant challenges as long-term and stable access pattern is often required in resource allocation and content assignment.

Another challenge faced when one designs the delay guarantee mechanism in CCN is the huge computational cost. Existing models or approximate algorithms [11], [12], [13] for analyzing caching performance (e.g., cache hit/miss ratio) for a network of caches often require per-content state tracking and analysis, i.e., by adopting Markov models [23]. As a result, significant amount of computation are involved when there is a large number of content objects or routers/nodes in the underlying system, as in the real network. This also implies that most of existing models or approximate algorithms are no longer applicable to the task of network delay guarantee in CCN. The required mechanism or models, on the other hand, needs to be computationally feasible and scalable.

To address these challenges and achieve delay guarantee in CCN, in this paper we make the following contributions:

  • 1.

    We present a simple and holistic network model which characterizes network delays of routing content objects to clients for content of all kinds, namely, locally cached, remotely cached and uncached, based on their locations. By assigning the same top ranked content objects in customer-facing routers as locally cached, and popular objects in peer routers as remotely cached, we ensure that end-users at different locations have a unified content access pattern. And this content access pattern is long-term and stable since the number of top ranked content objects cached in network is rather small as compared to the number of content objects delivered by the network.

  • 2.

    In order for each content provider to have an optimized network delay for its content dissemination, we align network locations with their content popularities by assigning the top most ranked content objects in customer-facing routers and the popular objects in peer routers. We then combine the content distribution model with the proposed holistic network model, and derive an analytical optimized network delay for each content provider.

  • 3.

    With the analytical network delay for each content provider, we further formally formulate the network delay guarantee task in CCN as a nonlinear integer programming (NIP) problem under the given network resources and traffic patterns of the underlying competing content providers. Rather than calculating the exact location for each content object, we approach the problem by specifying the number of top most ranked content objects that are cached locally and that are cached remotely. This significantly reduces the computational cost as compared to the existing models or approximate algorithms to content placement.

  • 4.

    We evaluate our models and investigate the optimized network performance through numerical studies. Using different network topologies, we study how content providers compete for the network resources and how various factors (e.g., content popularity, traffic volume, router storage capacity) affect this competition process. Our results reveal interesting and important phenomena, for example, increasing content population does not significantly influence the competition process, but it degrades the overall network delivery performance; similarly, it is observed that increasing network storage improves the overall content delivery performance, but it almost does not affect the competition process, etc. We believe these results are highly valuable as they provide insights into designing QoS mechanisms for CCN.

The rest of the paper is organized as follows. Section 2 reviews related work. Section 3 gives a detailed description of our models (network model, content distribution model and delay model) as well as the problem formulation. Section 4 presents our numerical studies and evaluation results. We conclude the paper in Section 5.

Section snippets

Related work

Network architectures with built-in storage [2], [3], [4], [5], [6] have received increasingly attention, and there is a large body of research in this field. In this section, we review some of the most well-known work.

One of the most important topics in this area is modeling and analysis of caching mechanisms. Researchers have proposed models and algorithms for analyzing caching effectiveness and characterizing caching dynamics. In [21], Busari and Williamson adopted both synthetic workload

Models and problem formulation

In this section, we present in details our models (network model, content distribution model and delay model) for the delay guarantee task in CCN. We then give mathematical problem formulation and further discuss some related issues (e.g., computational cost, implementation) of our mechanism.

Note that the network model was originally proposed in [29], and in this work we extend it by considering networks with both end-routers and transit routers, and for completely different purposes.

Numerical studies and evaluation

In this section, we evaluate our delay guarantee mechanism through numerical studies. We mainly focus on the following two objectives: (1) illustrate how our delay guarantee model can be adopted to provide delay guaranteed services for different content providers; (2) based on the numerical results, study how different content providers (services) compete for the network resources and how various factors (e.g., content distribution, traffic volumes, router storage capacity) affect the

Conclusion

QoS guarantee for content-centric networks is a new but challenging research area due to the newly introduced built-in caching mechanism. Network delay is a key metric of QoS. In this paper, we investigate the problem of guaranteeing network delays for differentiated services (content providers) in CCN while at the same time optimizing the overall content delivery performance. To address the key challenges, in particular, the high computational cost incurred by conventional solutions such as

Acknowledgment

The authors would like to thank the anonymous reviewers for their valuable comments and suggestions which will certainly improves the quality of this paper.

References (39)

  • M. Gallo et al.

    Performance evaluation of the random replacement policy for networks of caches

    ACM SIGMETRICS Perform. Eval. Rev.

    (2012)
  • P.R. Jelenkovic

    Asymptotic approximation of the move-to-front search cost distribution and least-recently-used caching fault probabilities

    Annal Appl. Probab.

    (1999)
  • E.J. Rosensweig et al.

    Approximate models for general cache networks

    Proceedings of IEEE INFOCOM

    (2010)
  • L. Muscariello et al.

    Bandwidth and storage sharing performance in information centric networking

    Proceedings of ACM SIGCOMM Workshop on Information-centric Networking. ACM

    (2011)
  • G. Carofiglio et al.

    Modeling data transfer in content-centric networking

    Proceedings of the 23rd International Teletraffic Congress

    (2011)
  • A. Jiang et al.

    Optimal content placement for en-route web caching

    Second IEEE International Symposium on Network Computing and Applications

    (2003)
  • M. Korupolu et al.

    Coordinated placement and replacement for large-scale distributed caches

    IEEE Trans. Knowl. Data Eng.

    (2002)
  • S. Borst et al.

    Distributed caching algorithms for content distribution networks

    Proceedings of IEEE Infocom

    (2010)
  • J. Li et al.

    Popularity-driven coordinated caching in named data networking

    Proceedings of the Eighth ACM/IEEE Symposium on Architectures for Networking and Communications Systems

    (2012)
  • Cited by (15)

    • QoSA-ICN: An information-centric approach to QoS in vehicular environments

      2021, Vehicular Communications
      Citation Excerpt :

      M.F. Al-Naday et al. [1] use QoS naming prefixes built on PURSUIT. Weibo Chu et al. [60] employ optimized caching polices where data popularity is considered when making decisions on content placement. Xing et al. [61] focus on energy efficiency within the network, with the approach tightly coupled to the communication network links.

    • Energy-efficient ICN routing mechanism with QoS support

      2018, Computer Networks
      Citation Excerpt :

      The first one was to retrieve the content from the multiple sources simultaneously while maximizing aggregate bandwidth, the second one was to select the highest bandwidth path to a single source for high bandwidth application, and the last one was to minimize per-packet delay for real-time traffic. In [36], the problem of meeting network delay requirements for the differentiated services was investigated. It solved delay guarantee task as a nonlinear integer programming problem.

    • Potentials, trends, and prospects in edge technologies: Fog, cloudlet, mobile edge, and micro data centers

      2018, Computer Networks
      Citation Excerpt :

      Considering the gigantic amount of data flowing from Internet to users, various solutions have been employed, such as CDN and caching to minimize the data and delay from SP to user [145]. For instance, cacheable contents are cached at ISP caches or CDN networks to minimize transit network data flow and delay [146–148]. However, the advent of new technologies, gadget proliferations, smart environments, and IoE are changing the data flow paradigm and patterns.

    • Resource reservation in information centric networking

      2022, ICN 2022 - Proceedings of the 2022 9th ACM Conference on Information-Centric Networking
    • Reducing Video Transmission Cost of the Cloud Service Provider with QoS-Guaranteed

      2022, Communications in Computer and Information Science
    • Priority Service for Paying Content Providers through Dedicated Cache Leasing in Information-Centric Networking

      2019, 2019 2nd IEEE International Conference on Hot Information-Centric Networking, HotICN 2019
    View all citing articles on Scopus

    The research is supported by the National Natural Science Foundation of China (61502393) and the Fundamental Research Funds for the Central Universities (3102014JSJ0016). Zhang was supported in part by US NSF grant CNS-1411636.

    View full text