Modeling and performance analysis for multimedia data flows scheduling in software defined networks

https://doi.org/10.1016/j.jnca.2017.01.032Get rights and content

Abstract

Supporting diverse Quality of Service (QoS) performance for data flows generated by multimedia applications has been a challenging issue in Software Defined Network (SDN). However, the current available QoS provision mechanisms proposed for SDN have not fully considered the heterogeneity and performance diversity for multimedia data flows. To this end, this work presents a hybrid scheduling model by combining priority queueing with a packet general-processor sharing to provide diverse QoS guarantees for multimedia applications in SDN. Network Calculus is applied to develop modeling and analysis techniques to evaluate the QoS performance of the proposed scheme. Performance bounds guaranteed by the proposed scheme for heterogeneous data flows, including their worst-case end-to-end delay and queueing backlog, are thus determined. Both analytical and simulation results show that the modeling and analysis techniques are general and flexible thus are fully capable of modeling QoS for diverse requirements of multimedia applications in SDN.

Introduction

Recent proliferation in computing and networking technologies has enabled the rapid development of a wide spectrum of network-based multimedia applications (Lu et al., 2015, Yuan et al., 2015). WebTV, video on demand, online gaming, and video conference are becoming exceedingly popular over the Internet. Internet multimedia applications bring in new challenges to networking technologies as such applications usually expect more bandwidth, little or no packet drop, low packet delay, and small delay variation. The diversity in multimedia applications, which have different traffic characteristics and performance requirements, further complicates the network control for supporting both multimedia and best-effort traffic coexisting in the Internet (Che et al., 2015). Therefore, more flexible network architectures and more effective network control mechanisms are required to achieve diverse service provisioning with Quality of Service (QoS) guarantees.

Traditional networking technologies lack sufficient capabilities to address the above challenges. The IP-based Internet architecture with tightly-coupled data forwarding and routing functions suffers from the ossification that constrains flexible on-demand service provisioning. Tremendous amount of efforts have been made by both academic and industrial researchers for enhancing Internet QoS, resulting in, for example, IntServ (Braden et al.), DiffServ (Carlson et al.), and MPLS (Awduche and Jabbari, 2002). However, the currently available mechanisms are quite limited in several aspects, such as the scalability issue of IntServ, coarse-grain traffic classification of DiffServ, and complex configuration, management, and troubleshoot of MPLS. QoS provisioning for heterogeneous data flows to meet multimedia application requirements remains an open issue.

Software-Defined Network (SDN) as an emerging networking paradigm brings new hope to address this challenging issue. SDN architecture decouples network control and data forwarding functions, thus enabling network control to be directly programmable by upper layer applications (Wang et al., 2016, Wang et al., 2015, Chang et al., 2015, Li et al., 2014, ONF,). Its key features include separated control and data planes, a logically centralized network controller, programmability of the control plane, and standard application programming interfaces (APIs). DN provides greater flexibility and finer granularity on traffic control for meeting the diverse performance requirements of multimedia data flows. Flow tables in SDN switches allow network traffic to be segregated into flows using a wide variety of criteria, including MAC addresses, IP addresses, TCP port numbers, etc. The SDN controller may specify various packet processing actions for steering the traffic of different flows. In addition, the centralized control plane in SDN allows a single controller to obtain a global view of the entire network domain and perform end-to-end flow-level traffic control across multiple switches, which may significantly enhance the capability of QoS provisioning.

While numerous works have been done for QoS provisioning in SDN from the control and management perspective, QoS performance guarantee in SDN requires effective scheduling models for flow forwarding in data plane switches. Flow scheduling in SDN witches for supporting multimedia applications must handle a wide variety of data flows with highly diverse traffic load features to meet different levels of QoS requirements. In addition, the scheduling model should also be able to achieve high network resource utilization for both multimedia data flows and best-effort traffic in SDN. However, none of the currently available single packet scheduling scheme can meet all these requirements. Therefore, a hybrid scheduling model is necessary. Although hierarchical packet scheduling algorithms have been developed for Internet QoS provisioning (Bennett and Zhang, 1997, Stoica et al., 2000), hybrid packet scheduling employed in SDN switches for supporting multimedia data flows is still an open issue that has not been sufficiently studied yet. In this work, we propose a hybrid scheduling model for multimedia flows in SDN to address this issue.

On the other hand, in order to evaluate the QoS capability of the proposed scheduling scheme we must face some new challenges brought in by SDN to network performance analysis. Traditional queueing theory-based analysis methods rely on some strong assumptions about the probabilistic attributes of the service process and arrival load for the studied network system. Nevertheless, decoupling of data and control planes in SDN introduces an abstraction of network infrastructure, which may consist of switches with heterogeneous implementations that are transparent to controllers. Besides, multimedia applications generate data flows with highly diverse traffic loads, and thus it is difficult to have an appropriate assumption about the distributions of packet service time and inter-arrival time in an SDN environment.

To attack this problem, we employ network calculus (Boudec and Thiran, 2012) in this paper as the mathematical tool for modeling and analyzing the proposed scheduling system. The network calculus-based analysis method may obtain the upper bounds of queueing delay and backlog guaranteed to each data flow in an SDN network without assuming any specific switch implementation or arrival traffic distribution. Therefore, application of network calculus makes our analysis agnostic to SDN switch implementations and applicable to various multimedia data flows with diverse traffic characteristics.

In summary, the main challenges we attempt to tackle in this paper lie in the diversity of QoS requirements for the data flows that must be supported simultaneously by SDN and the heterogeneity of switch implementations enabled by SDN data plane abstraction. The network calculus-based modeling and performance evaluation methods developed in this paper are general and flexible, thus are applicable to diverse multimedia data flows processed by heterogeneous switches in an SDN environment. Specifically, we make the following contributions in this paper.

  • We propose to apply a hybrid scheduling mechanism in SDN switches for providing QoS guarantees to multimedia data flows.

  • We develop a network calculus-based model for evaluating QoS capability of the proposed hybrid scheduling scheme in SDN switches.

  • We determine the performance bounds, including the maximum queuing delay and backlog, guaranteed by the proposed scheduling scheme to each traffic flow, for both single hop and multi-hop cases.

  • We conduct extensive simulation experiments to evaluate effectiveness of the proposed scheduling scheme and analyze the impacts of various queueing and scheduling parameter settings on its performance.

The rest of this paper is organized as follows. We review related works in Section 2 and present the hybrid scheduling that we propose to apply in SDN switches for supporting multimedia data flows in Section 3. Then in Section 4 we develop the network calculus-based performance analysis technique to evaluate the QoS capability of the proposed scheduling model. Numerical and simulation results are provided and discussed in Section 5. We draw conclusions and identify the possible future work in Section 6.

Section snippets

Related work

QoS provisioning in SDN has been studied from various aspects (Zinner et al., 2014). The centralized control plane enabled by SDN allows network programmability through an API between network controllers and upper layer applications. This API, often referred to as a northbound interface in SDN, allows applications to specify their QoS requirements, thereby enabling application-based QoS control. Gorlatch et al. (2014) described the specification of a northbound API that allows real-time

Hybrid scheduling in SDN switches

The logically centralized control plane in SDN provides an effective mechanism for flow-level traffic control across the entire network domain. In SDN networks, flows of traffic belong to various media streams are treated differently by data plane devices in order to meet their diverse performance requirements. The scheduling mechanisms implemented in SDN switches realize the advantage of flow-level traffic control. OpenFlow specification defines the concept of OpenFlow port, which may

Network calculus preliminaries

The theory of network calculus was initially developed by Chang (1994) and Cruz (1991) and then extended by others (Le Boudec and Thiran, 2001, Agrawal et al., 1999, Jiang and Liu, 2008). Network calculus gives a theoretical framework that can overcome some limitations of the traditional queueing theory-based approaches for analyzing network performance guarantees. Traditional network modeling and analysis methods typically need to make some assumptions about the technologies employed in packet

Experimental results

In the previous section, we have obtained the service curve for the single and multi-hop cases, and proved their end-to-end delay and backlog upper bounds, respectively. In this section, we evaluate these performance bounds and discuss various aspects that may impact them through numerical experiments. This allows us to have an in-depth understanding of the multimedia flow delivery performance through our model. In these experiments, we assume that there are three variable bit-rate video flows,

Conclusion

To deal with the issue of real-time transmissions of data flows and attain the fairness in scheduling data flows of the same priority in SDN, we for the first time propose a hybrid scheduling model that combines PGPS and preemptive scheduling algorithms by using the theory of network calculus, we have successfully established an end-to-end deterministic performance analysis of the propose hybrid scheduling model for both single and multi-hop cases, which includes service curve, delay, and

Acknowledgements

This work is supported by the NSFC under grant number 61309031 and Chongqing Municipal program under grant number cstc2015jcyjBX0009 and cstcKJCXLJRC20.

References (38)

  • D.O. Awduche et al.

    Internet traffic engineering using multi-protocol label switching (MPLS)

    Comput. Netw.

    (2002)
  • R. Agrawal et al.

    Performance bounds for flow control protocols

    IEEE/ACM Trans. Netw.

    (1999)
  • Azodolmolky, S., Nejabati, R., Pazouki, M., Wieder, P., Yahyapour, R., Simeonidou, D., 2013. An analytical model for...
  • J.C.R. Bennett et al.

    Hierarchical packet fair queueing algorithms

    IEEE/ACM Trans. Netw.

    (1997)
  • Bianco, A., Birke, R., Giraudo, L., Palacina, M., 2011. OpenFlow switching: data plane performance. In: 2010...
  • J.-Y.L. Boudec et al.

    Network calculusa theory of deterministic queueing systems for the Internet

    Springe. Lect. Notes Comput. Sci. LNCS

    (2012)
  • Braden, R., Clark, D., Shenker, S., Integrated services in the Internet architecture: an overview, IETF...
  • Bueno, I., Aznar, J., Escalona, E., Ferrer, J., Antoni Garcia-Espin, J., 2013. An opennaas based sdn framework for...
  • Carlson, M., Davies, E., Nortel, U., Wang, Z., Weiss, W., An architecture for differentiated services, IETF...
  • C.-S. Chang

    Stability, queue length, and delay of deterministic and stochastic queueing networks

    IEEE Trans. Autom. Control

    (1994)
  • Chang, X., Li, J., Wang, G., Zhang, Z., Li, L., Niu, Y., 2015. Software defined backpressure mechanism for edge...
  • X. Che et al.

    A survey of current youtube video characteristics

    IEEE Multimed.

    (2015)
  • R.L. Cruz

    A calculus for network delay. i. network elements in isolation

    IEEE Trans. Inf. Theory

    (1991)
  • Duan, Q., 2014. Network-as-a-service in software-defined networks for end-to-end QoS provisioning. In: Wireless and...
  • Durner, R., Blenk, A., Kellerer, W., 2015. Performance study of dynamic QoS management for OpenFlow-enabled SDN...
  • Egilmez, H.E., Dane, S.T., Bagci, K.T., Tekalp, A.M., 2012. Openqos: An OpenFlow controller design for multimedia...
  • Gorlatch, S., Humernbrum, T., Glinka, F., 2014. Improving QoS in real-time internet applications: from best-effort to...
  • N. Gude et al.

    NOXtowards an operating system for networks

    ACM SIGCOMM Comput. Commun. Rev.

    (2008)
  • Hua, Y., Liu, X., 2011. Scheduling design and analysis for end-to-end heterogeneous flows in an avionics network. In:...
  • Cited by (17)

    • Analytical Modelling of Software and Hardware Switches with Internal Buffer in Software-Defined Networks

      2019, Journal of Network and Computer Applications
      Citation Excerpt :

      The controller processes a packet-in message and generates a packet-out message to the switch updating flow information (ONF, 2013). Most existing research in the literature analyses the performance of SDN switches with no internal buffering (Jarschel et al., 2011; Mahmood et al., 2014; Mahmood et al., 2015; Miao et al., 2016; Goto et al., 2016; Ogasawara and Takahashi, 2016; Shang and Wolter; Sood et al., 2016; Javed et al., 2017; Singh et al., 2017; Singh et al., 2018a; Lai et al., 2017; Fahmin et al., 2018; Azodolmolky et al., 2013; Koohanestani et al., 2017; Huang et al., 2017). This is perhaps attributed to the evolving nature of the OpenFlow specifications which in their current incarnation leaves the buffering of a data packet as an optional feature.

    • Enabling hard service guarantees in Software-Defined Smart Grid infrastructures

      2018, Computer Networks
      Citation Excerpt :

      Yet, computations are performed offline as the approach is not coupled with an actual SDN set-up. Similarly, Huang et al. validate their proposed hybrid scheduling approach for SDN switches by applying offline NC analysis [27]. In [28] NC is employed for the analysis of SDN scalability.

    • Modelling Software-Defined Networking: Software and hardware switches

      2018, Journal of Network and Computer Applications
      Citation Excerpt :

      To identify the potential bottlenecks that could hinder the performance of the SDN, the trade offs between choosing a hardware vs. software need to be studied and investigated to improve the performance of SDN. While other approaches like Network Calculus has been also used to model SDN (Azodolmolky et al., 2013; Koohanestani et al., 2017; Huang et al., 2017), we use queueing theory to model both hardware and software switches in this paper for finer grain analysis. Most analytical models (queueing and network calculus) for SDN characterize the software switch architecture (i.e. a switch equipped with CPU only), with none of the queuing models to the best of our survey have modelled the hardware switch architecture (i.e. switch with CPU and specialized hardware).

    • Fundamental relationship between node dynamic and content cooperative transmission in mobile multimedia communications

      2018, Computer Communications
      Citation Excerpt :

      As a key problem affecting the transmission performance of multimedia service, there are many works on the transmission mechanism of multimedia service and the relationship between the carriers. In order to guarantee the transmission performance and QoS of multimedia service, Huang et al. [32] proposed a hybrid scheduling model based on calculus. By dividing the service priority and using the queue processing technology, the end-to-end delay and the queue length of the service were limited to a certain extent, but this mechanism did not take into account the complex social relations between nodes and the dynamic evolution of the social attributes of nodes.

    View all citing articles on Scopus
    View full text