Scheduling Both GBR and Non-GBR Uplink Bearers Based on Moving Average of Data Rate Fang-Chang Kuo, Kuo-Chang Ting * , Chih-Cheng Tseng, Hwang-Cheng Wang, and Ming-Wei Chen National Ilan University, Taiwan kfc@niu.edu.tw * Minghsin University of Science and Technology, Taiwan AbstractMany resource allocation schemes have been proposed based on different criteria such as system throughput, fairness, transmission power, user priority and others. According to the 3GPP specifications, however, as long as a GBR (guaranteed bit rate) bearer is admitted, eNB has to allocate resource blocks (RBs) for this bearer to ensure its transmission right. At the same time, eNB also must distribute the remaining RBs to non-GBR bearers so as to achieve high radio resource utilization. However, 3GPP has not specified how to measure GBR. In this paper, we use ex- ponentially weighted moving average (EWMA) to define the measurement of data rate and show that it is beneficial to the scheduling on both UE and eNB sides. We also present a schedul- ing scheme, AAG-2, which aims at ensuring QoS for all in- progress GBR sessions, while at the same time efficiently allocat- ing RBs to non-GBR sessions. Keywords-uplink scheduling, resouce allocation, EWMA, non- GBR, AAG-2 I. INTRODUCTION Dynamic resource allocation (DRA) for LTE uplink bearers is an important task of radio resource management (RRM). The schemes for allocating resource blocks (RBs) to user equip- ment (UE) have been extensively studied. The schemes pro- posed in the early articles usually emphasize maximizing the system throughput of an eNB. For example, the Recursive Maximum Expansion (RME) and relevant schemes allocate free RBs to a UE which has best channel quality [1, 2]. Many researchers felt that fairness is also a very important criterion. As a result, some paradigms have emerged. A re- source chunk (RC) paradigm algorithm treat RC, which is a fixed number of RBs, as a basic unit of allocation to UEs. Any UE that is not scheduled in a transmission time interval (TTI) will have higher priority in subsequent TTIs [3-5]. One simple scheduling scheme of this paradigm is frequency domain round robin (RR), which allows RCs to be assigned to multiple UEs within one TTI in a cyclic manner [6]. The proportional fair (PF) paradigm allocates resources to users based on the ratio of their channel condition over their lifelong service rate [7, 8]. As a result, users with poor channel conditions will get some re- sources but fewer than those with better channel conditions. However, both of these paradigms do not take UE QoS re- quirements into account when forming the allocation matrix and still suffer problems related to fairness in terms of through- put. Some resource allocation schemes are concerned with the transmission power of UEs [9, 10]. Authors in [11] proposed a QoS-aware resource allocation paradigm for LTE uplink scheduling that gives greater advantage to UEs with high- priority data, while not starving other users. However, bearers in LTE are classified into GBR and non-GBR ones, rather than priorities among UEs. The above-mentioned dynamic resource allocation (DRA) schemes are designed based on different criteria. We would like to briefly review some concepts pertaining to DRA in 3GPP specifications [12, 13] before presenting our scheme. The task of radio admission control (RAC) is to admit or reject the establishment requests for new radio bear- ers. The goal of RAC is to ensure high radio resource utilization and at the same time maintain proper QoS for in-progress sessions. DRA typically takes into account the QoS require- ments associated with radio bearers, the channel quali- ty information for UEs, interference situation, buffer status, etc. Each GBR (guaranteed bit rate) bearer is associated with throughput requirements expressed in terms of (a) GBR, which is the bit rate that can be expected to be provided, and (b) maximum bit rate (MBR), which is the maximum bit rate that can be expected to be pro- vided. The throughput of non-GBR bearers in a UE is denoted in group by aggregate maximum bit rate (AMBR). Based on these key concepts, we come up the following ob- servations, which are often neglected in previous research: As long as a GBR bearer is admitted, eNB has to allo- cate sufficient RBs to meet the guaranteed data rate ir- respective of the channel quality. However, 3GPP has not specified how to measure GBR, MBR and AMBR. Apart from allocating RBs for GBR bearers to ensure their QoS, eNB also must distribute the remaining RBs to non-GBR bearers in order to achieve high radio re- source utilization. Buffer status report (BSR) also provides important in- formation for DRA. In this paper, we present an algorithm that uses exponen- tially weighted moving average (EWMA) to define the meas- urement of GBR and facilitate the operation of DRA. We also present a scheduling scheme, AAG-2 (Allocate As Granted, version 2), which employs BSR to prevent the waste of RBs. AAG-2 aims at ensuring QoS for all in-progress GBR sessions, while at the same time efficiently distributing RBs to non-GBR sessions so as to improve resource utilization. In the following, we describe the use of EWMA for data rate measuring in Section II. Section III briefly describes the operation of AAG scheme, which is used as a module in our This paper is supported by MOST of Taiwan under grant no. 100-2221-E-197-019 QSHINE 2015, August 19-20, Taipei, Taiwan Copyright © 2015 ICST DOI 10.4108/eai.19-8-2015.2260881