Elsevier

Computer Networks

Volume 53, Issue 11, 28 July 2009, Pages 1771-1783
Computer Networks

Threshold configuration and routing optimization for PCN-based resilient admission control

https://doi.org/10.1016/j.comnet.2009.01.013Get rights and content

Abstract

Pre-Congestion Notification (PCN) in IP networks uses packet metering and marking within a PCN domain to notify its egress nodes whether link-specific admissible or supportable rate thresholds have been exceeded by high-priority traffic. Based on this information simple admission control and flow termination is implemented. The latter is a new flow control function and useful in case of overload through high priority traffic which can occur in spite of admission control, e.g., when traffic is rerouted in failure cases. Resilient admission control admits only so much traffic that admitted traffic can be rerouted without causing congestion on backup paths in case of a likely failures, e.g., single link failures.

We propose algorithms to configure the link-specific PCN rate thresholds such that resources are utilized efficiently and fairly by competing traffic aggregates while meeting resilience constraints. This is done for the single and dual marking PCN architecture whereby the single marking case is more demanding since it requires that the supportable rate is a fixed multiple of the admissible rate on all links within a single PCN domain. Furthermore, we derive objective functions to optimize the underlying routing system for both cases. Our performance results for various network types show that the dual marking PCN architecture leads to significantly better resource efficiency than the single marking PCN architecture.

Introduction

Internet service providers (ISPs) recently offer increased access speeds, e.g., by digital subscriber lines (DSL), cable TV (CATV), and fiber to the home (FTTH). These technologies significantly increased the traffic volume in carrier networks and in 2005, the major traffic in Japan was already produced by residential users [1]. Popular video services like YouTube produce large traffic volumes, but are only weak precursors of high-quality IP-TV services. They present a challenge for ISPs which need to offer triple play, i.e., the integration of the transport of data, voice, and video. However, the resource management for triple play becomes more and more difficult due to the emerging interactive Web 2.0 since residential users also become content providers. In particular, [2] has shown that some normal users get accustomed with new services, change access technologies, and become “heavy hitters” and hence the majority of the overall traffic is produced by a minority of residential users.

Today, ISPs rely on capacity overprovisioning (CO) to support quality of service (QoS) in terms of packet loss and delay. In [3] admission control (AC) was proposed for IP networks, but so far such techniques are applied only locally, they are rarely in use, and not deployed in core networks. However, there is a firm belief that next generation networks require some form of QoS assurance such as AC to enable services that cannot be provided with CO [4]. Conventional AC prevents overload due to increased user activity. If congestion occurs in core networks, it is mainly caused by failures and redirected traffic, and only to a minor degree by increased user activity [5]. Thus resilient AC is required that admitted traffic can be rerouted in likely failure scenarios without causing congestion on backup paths [6]. In other words, both AC and CO require backup capacity to prevent QoS violations due to backup traffic in case of failures. In case of CO, this backup capacity can be used to accommodate both moderate fluctuations of the traffic matrix and backup traffic. As a consequence, there are no significant bandwidth savings when AC is used instead of CO for QoS provisioning [7]. However, the dynamic behavior of users and services sketched above leads to an unpredictability of future demands such that QoS provisioning remains difficult. Therefore, ISPs see the need for AC to offer premium services over integrated IP networks in the future.

The Internet Engineering Task Force (IETF) currently works on “Congestion and Pre-Congestion Notification” (PCN) [8] with the objective to standardize feedback-based admission control (AC) and flow termination (FT) for high-priority PCN traffic for single DiffServ domains [9]. Each link l of a so-called PCN domain is associated with an admissible and a supportable rate threshold (AR(l),SR(l)) and the egress nodes of the domain are notified via appropriately marked packets if these thresholds are exceeded by high-priority PCN traffic. This feedback is used to implement AC and FT. Various packet marking schemes as well as AC and FT methods are proposed [10]. Some proposals provide two metering and marking schemes [11], [12], to control the admissible and the supportable rate independently of each other (dual marking PCN architecture, DM-PCN). Others provide only a single metering and marking scheme [13], [14] that controls only the admissible rate (single marking PCN architecture, SM-PCN). They implicitly assume that the supportable rate SR(l) of a link l is a fixed multiple b of its admissible rate AR(l) in the entire PCN domainSR(l)=b·AR(l).As a consequence, egress nodes can infer from the ratio of marked and unmarked traffic whether only the admissible or also the implicit supportable rate is exceeded on some link. We call the parameter b the “backup factor” as it controls the relation of primary and backup capacity on the links. The advantage of SM-PCN is that it needs fewer codepoints in the IP header for packet marking and less metering and marking support by routers. Its disadvantage is that Constraint (1) limits traffic engineering capabilities and makes the configuration of the rate thresholds harder when resource efficiency is an objective. In addition, it does not work well with multipath routing and when single edge-to-edge aggregates carry only little traffic [10].

This work investigates the rate threshold setting problem for PCN-based AC and FT. Furthermore, it proposes objective functions for routing optimization in resilient PCN networks. Performance results compare the resource efficiency of DM-PCN and SM-PCN with and without routing optimization for a large set of sample networks. The algorithms presented in this study also serve to configure and optimize PCN networks in practice.

The paper is structured as follows. Section 2 reviews related work showing the historic roots of PCN and similar AC approaches. Section 3 introduces PCN and explains how AC and FT work in the single and dual marking PCN architecture (SM-PCN, DM-PCN). Section 4 proposes algorithms to set the admissible and supportable rate thresholds appropriately for resilient AC. Section 5 provides objective functions to optimize IP routing in order to maximize the admissible protected traffic. Section 6 compares the resource efficiency of SM-PCN and DM-PCN for a large set of networks with different characteristics. Finally, Section 7 summarizes this work and draws conclusions.

Section snippets

Related work

We review related work regarding random early detection (RED), explicit congestion notification (ECN), and stateless core concepts for AC as they can be viewed as historic roots of PCN.

PCN-based flow control

This section illustrates the basic idea of PCN-based admission control (AC) and flow termination (FT) using the nomenclature of [10]. An example illustrates how PCN-based AC and FT fit into the overall Internet structure. We review how AC can be implemented based on appropriate metering and marking schemes. FT methods may reuse the marking scheme for AC or require their own. This leads to the definition of a single and dual marking PCN architecture (SM-PCN, DM-PCN). We show how PCN-based AC and

Threshold configuration for PCN-based flow control

In this section we propose simple and improved algorithms for the configuration of the AR- and SR-thresholds for SM- and DM-PCN. The simple algorithms set thresholds in such a way that the same fraction of all expected ingress–egress aggregates can be admitted as high-priority traffic. This possibly leaves some of the link capacities unused. Therefore, the improved algorithms strive for a higher resource utilization while implementing max–min fairness [39] among ingress–egress aggregates with

Routing optimization for PCN-based flow control

In this section we derive objective functions for routing optimization to maximize the protected throughput of high-priority traffic for both SM- and DM-PCN. We illustrate the effect of the algorithms by numerical results.

Efficiency of SM- and DM-PCN: a parametric study

In this section, we study the ability of SM- and DM-PCN to carry as much protected high-priority traffic as possible. We investigate the impact of simple and improved threshold assignment as well as routing optimization in networks of different size and with different node degree to generalize the results of Sections 4 Threshold configuration for PCN-based flow control, 5 Routing optimization for PCN-based flow control. We first describe the experiment setup and the exact performance measure

Conclusion

Pre-Congestion Notification (PCN) essentially marks packets when PCN traffic exceeds configured admissible or supportable rate thresholds (AR,SR) on a link of the PCN domain. The IETF attempts to use this feedback for simple and scalable admission control (AC) and flow termination (FT) in IP networks. Currently, there are many different options having benefits and drawbacks [10] that need to be understood. One class of methods requires two different marking mechanisms (dual marking PCN

Michael Menth studied computer science and mathematics at the University of Wuerzburg/Germany and Austin/Texas. He worked at the University of Ulm/Germany and Wuerzburg and obtained his Ph.D. in 2004. Currently, he is assistant professor and heading the research group “Next Generation Networks” at the Institute of Computer Science in Wuerzburg. His special interests are performance analysis, optimization of communication networks, resource management, resilience issues, and Future Internet. He

References (47)

  • W. Almesberger et al.

    SRP: a scalable resource reservation for the Internet

    Computer Communications

    (1998)
  • M. Menth et al.

    Resilient network admission control

    Computer Networks

    (2008)
  • G. Dueck et al.

    Threshold accepting; a general purpose optimization algorithm

    Journal of Computational Physics

    (1990)
  • K. Fukuda et al.

    The impact of residential broadband traffic on Japanese ISP backbones

    ACM SIGCOMM Computer Communications Review

    (2005)
  • K. Cho, K. Fukuda, H. Esaki, A. Kato, The impact and implications of the growth in residential user-to-user traffic,...
  • S. Shenker

    Fundamental design issues for the future Internet

    IEEE Journal on Selected Areas in Communications

    (1995)
  • D.M. Johnson

    QoS control versus generous dimensioning

    British Telecom Technology Journal

    (2005)
  • S. Iyer, S. Bhattacharyya, N. Taft, C. Diot, An approach to alleviate link overload as observed on an IP backbone, in:...
  • M. Menth, Efficient Admission Control and Routing in Resilient Communication Networks, Ph.D. Thesis, University of...
  • M. Menth, R. Martin, J. Charzinski, Capacity overprovisioning for networks with resilience requirements, in: ACM...
  • IETF Working Group on Congestion and Pre-Congestion Notification (PCN), Description of the Working Group....
  • P. Eardley (Ed.), Pre-Congestion Notification Architecture....
  • M. Menth et al., PCN-based admission control and flow termination, Currently under revision for IEEE Communications...
  • B. Briscoe et al., An Edge-to-Edge Deployment Model for Pre-Congestion Notification: Admission Control over a DiffServ...
  • J. Babiarz, X.-G. Liu, K. Chan, M. Menth, Three State PCN Marking....
  • A. Charny, F.L. Faucheur, V. Liatsos, J. Zhang, Pre-Congestion Notification Using Single Marking for Admission and...
  • L. Westberg, A. Bhargava, A. Bader, G. Karagiannis, H. Mekkes, LC-PCN: The Load Control PCN Solution....
  • S. Floyd et al.

    Random early detection gateways for congestion avoidance

    IEEE ACM Transactions on Networking

    (1993)
  • B. Braden et al., RFC2309: Recommendations on Queue Management and Congestion Avoidance in the Internet, 1998...
  • K. Ramakrishnan, S. Floyd, D. Black, RFC3168: The Addition of Explicit Congestion Notification (ECN) to IP, 2001...
  • N. Spring, D. Wetherall, D. Ely, RFC3540: Robust Explicit Congestion Notification (ECN), 2003...
  • K. Nichols, S. Blake, F. Baker, D.L. Black, RFC2474: Definition of the Differentiated Services Field (DS Field) in the...
  • S. Floyd, RFC4774: Specifying Alternate Semantics for the Explicit Congestion Notification (ECN) Field, 2007...
  • Cited by (17)

    View all citing articles on Scopus

    Michael Menth studied computer science and mathematics at the University of Wuerzburg/Germany and Austin/Texas. He worked at the University of Ulm/Germany and Wuerzburg and obtained his Ph.D. in 2004. Currently, he is assistant professor and heading the research group “Next Generation Networks” at the Institute of Computer Science in Wuerzburg. His special interests are performance analysis, optimization of communication networks, resource management, resilience issues, and Future Internet. He holds numerous patent applications and received various scientific awards for innovative work.

    Matthias Hartmann studied computer science and mathematics at the Universities of Austin/Texas and Wuerzburg/Germany and obtained its Diploma in 2007. After working at Simula research center in Oslo/Norway, he is now Ph.D. student at the University of Wuerzburg, Germany.

    This work is funded by Nortel Networks, Ottawa, and Deutsche Forschungsgemeinschaft (DFG) under grant TR257/18-2. The authors alone are responsible for the content of the paper.

    View full text