Elsevier

Computer Networks

Volume 51, Issue 3, 21 February 2007, Pages 621-631
Computer Networks

Upper-level scheduling supporting multimedia traffic in cellular data networks

https://doi.org/10.1016/j.comnet.2006.05.007Get rights and content

Abstract

Wireless data networks such as cdma2000 1x EV-DO and UMTS HSDPA use downlink scheduling that exploits channel fading to increase the system throughput. As future wireless networks will eventually support multimedia and data traffic together, we need a proper criterion for scheduling that can count various service requirements such as delay and packet loss. Although some previous approaches proposed opportunistic schedulers at the lower layer, it has not been investigated well whether they are able to meet explicit QoS defined at the upper layer. Hence, in this paper, we develop a hierarchical scheduling model that considers QoS provisioning and the time-varying channel feature separately. We focus on the upper-level QoS scheduling that supports various traffic classes in a unified manner. Supposing that a user gets some satisfaction or utility when served, we introduce a novel concept of opportunity cost, which is defined as the maximum utility loss among users incurred by serving a particular user at the current turn. We obtain each user’s net profit by subtracting the opportunity cost from its expected utility, and then select a user with the maximum profit for service. Simulation results reveal that our scheme supports various QoS classes well that are represented by delay and packet loss under various traffic loadings.

Introduction

Recently high data rate systems have been deployed for cellular networks of which examples are cdma2000 1x EV-DO (or HDR) in 3GPP2 [1] and High Speed Downlink Packet Access (HSDPA) in 3GPP [2]. They have some features in common with the next generation cellular system and can be good candidates for it. To support a variety of services, it is important to meet each user’s desired QoS in these networks. Especially a cellular data network needs to manage network resources deliberately because it has some different characteristics from a voice network in many aspects. First, in the data network, the traffic volume for downlink is much higher than that for uplink. Second, there are many kinds of services such as HTTP, WAP, VoIP, real time video traffic, and so on, which have their own requirements of delay and loss rate. Last, data traffic is bursty on the whole.

In this paper, we consider a cellular data system that uses transmission rate control instead of power control in the downlink. The base station (BS) transmits at full power [6] and uses a time division multiplexing to maximize the cell throughput like in the EV-DO standard [1]. A lot of works have dealt with downlink scheduling, where all users share a single channel [8], [9], [10], [11], [12], [13], [14], [15], [16], [17]. They exploit channel fading instead of overcoming it. By allocating a time slot to a user with the maximum signal-to-interference-noise-ratio (SINR), the system achieves the maximum throughput. It is a simple concept of opportunistic scheduling that has lots of variants [8], [9], [10], [11], [12], [13]. The opportunistic scheduling is beneficial for best-effort traffic, but it needs further consideration to be applied for multimedia applications because its effect on packet delay has not been investigated enough.

Some previous works have dealt with the problem of scheduling delay-sensitive traffic [14], [15], [16], [17]. They apply opportunistic scheduling for real-time traffic, but cannot guarantee network-level QoS explicitly. Therefore we construct a two-level scheduling framework which comprises of the QoS-aware scheduler at the upper layer and the opportunistic scheduler at the lower layer. This architecture aims at maximizing the system efficiency, while satisfying each user’s QoS requirements. Since there have been a lot of works for lower-level scheduling, we work on the upper-level scheduling that supports diverse classes of traffic in an integrated fashion.

We use the concept of utility as the scheduling metric, which was first introduced in the area of networks by [19]. The utility represents the level of user’s satisfaction for the given service, and many opportunistic schedulers have already taken it as the objective function, i.e., maximizing the utility depending on time-varying SINR [9], [10] or average transmission rate [20]. The QoS of multimedia traffic is usually characterized by delay bound and loss rate, so we reflect them in designing the utility function. Our scheduler chooses a user who can achieve the maximum profit from using the opportunity cost that is determined by considering other users’ utilities.

We organize the remainder of this paper as follows. Section 2 shows a hierarchical scheduling model and Section 3 describes the design principles of a utility function that concerns the QoS of each traffic class. Section 4 proposes our scheduling scheme that considers the utility and the opportunity cost together. Then we investigate the performance of our scheme through simulation experiments in Section 5. Finally, Section 6 concludes this paper.

Section snippets

Hierarchical scheduler model

The transmission rate control enables a scheduler to use opportunistic scheduling that exploits channel fluctuation to achieve high throughput rather than overcomes it. If the scheduler simply chooses a user with the maximum SINR, it unfairly prefers some users with good channel. Thus, various algorithms have been developed in [8], [9], [10], [11], [12], [13]. In [8], the current data rate is normalized by the average data rate and the concept of proportional fairness is applied for scheduling.

QoS and utility functions

3GPP and 3GPP2 have classified traffic into conversational, streaming, interactive and background classes [3], [4]. Their characteristics are shown in Table 1. We represent QoS objective by a utility function that quantifies the satisfaction level of each user, depending on throughput, delay, and loss rate [19]. As shown in Table 1, multimedia traffic is generally sensitive to delay and loss. Hence, we incorporate both delay and data loss into the utility function.

Algorithm

Packet scheduling can select the job with the maximum utility. In this case, some other jobs with low utility may starve. Therefore we introduce the concept of opportunity cost. It is incurred by making a decision and defined as the benefit that the system gives up while carrying out that decision. Here we simply define it as the maximum utility that should be given up among the other users because of their losing the current scheduling turn. For example, consider a case of three jobs; jobs a,

Simulation environments

For simulations, we consider a cluster of seven hexagonal cells as shown in Fig. 6. The six neighboring cells generate signals interfering with the center cell of which performance we are interested in. The cell radius is 500 m and users are grouped according to the channel condition. The channel experiences path loss as dα, where d is the distance of the user from the base-station and α is the exponent that we set equal to 4 approximately for our simulation. The channel also experiences

Conclusions

In this paper, we investigated a new downlink scheduling scheme for wireless data networks. We decoupled the scheduling function into two levels: upper-level and lower-level. Each level of scheduling plays the role of QoS guarantee and opportunistic transmission, respectively, in the network and physical layers. Our upper-level scheduler supports multimedia traffic by an integrated approach that uses the concept of opportunity cost. It used utility functions that consider the delay bound and

Young-June Choi is currently a post-doctoral researcher in the School of Electrical Engineering and Computer Science, Seoul National University. He received his B.S., M.S., and Ph.D. degrees from Seoul National University, in 2000, 2002, and 2006, respectively. His research interests include fourth generation wireless networks, wireless resource management, and cross layer system design.

References (23)

  • X. Liu et al.

    A framework for opportunistic scheduling in wireless networks

    Computer Networks

    (2003)
  • 3GPP2 C.S0024 Ver3.0, cdma2000 High rate packet data air interface specification, December 5,...
  • 3GPP TR 25.858 V5.0.0, High speed downlink packet access: physical layer aspects(Release5), March...
  • 3GPP TS 23.107 V5.4.0, QoS concept and architecture (Release5), March...
  • 3GPP2 SR0035 Ver1.0, Quality of service, October 29,...
  • 3GPP2 WG5 Evaluation adhoc, 1xEV-DV Evaluation methodology-addendum V6, July 25,...
  • A. Bedeker, Downlink Scheduling in CDMA Data Networks, in: Proc. IEEE GLOBECOM 99, Rio de Janeiro, Dec....
  • P. Bender

    CDMA/HDR: a bandwidth-efficient high-speed wireless data service for nomadic users

    IEEE Communications Magazine

    (2000)
  • A. Jalali, R. Padovani, R. Pankaj, Data throughput of CDMA-HDR a high efficient high data rate personal communication...
  • X. Liu et al.

    Opportunistic transmission scheduling with resource-sharing constraints in wireless networks

    IEEE Journal on Selected Areas in Communications

    (2001)
  • S. Borst, P. Whiting, Dynamic rate control algorithms for HDR throughput optimization, in: Proceedings of the IEEE...
  • Cited by (0)

    Young-June Choi is currently a post-doctoral researcher in the School of Electrical Engineering and Computer Science, Seoul National University. He received his B.S., M.S., and Ph.D. degrees from Seoul National University, in 2000, 2002, and 2006, respectively. His research interests include fourth generation wireless networks, wireless resource management, and cross layer system design.

    Jin-Ghoo Choi received the B.S., M.S., and Ph.D. degree in the School of Electrical Engineering & Computer Science, Seoul National University, in 1998, 2000, and 2005, respectively. He is a senior engineer in Samsung Electronics. His research interests include resource management and packet scheduling in wireless networks.

    Saewoong Bahk received B.S. and M.B. degrees in Electrical Engineering from Seoul National University in 1984 and 1986, respectively, and the Ph.D. degree from the University of Pennsylvania in 1991. From 1991 through 1994 he was with the Department of Network Operations Systems at AT&T Bell Laboratories as an MTS where he worked for AT&T network management. In 1994, he joined the school of electrical engineering at Seoul National University and currently serves as a professor. His areas of interests include performance analysis of communication networks and network security.

    This research was supported partially by the University IT Research Center Project and the Ubiquitous Autonomic Computing and Network Project, Ministry of Information and Communication, in Korea.

    View full text