Scheduler design for traffic classification in CDMA2000 1xEVDO network
Abstract
In CDMA2000 Evolution Data Optimized (EVDO) network, initial applications have been mainly focused on Best Effort (BE) traffic. Network operators observed that even though BE traffic users were the lowest priority users paying small subscription fees, they were heavily consuming the network resources. Therefore, operators demanded to charge BE users depending on their different levels of data usage. In this paper, we have designed a novel forward link (FL) scheduler called Throughput Based Adaptive Scheduler (TBAS)to differentiate BE traffic users, and at the same time to optimize the sector throughput. A set of live network experiments has been conducted to understand the behavior of current scheduler in CDMA 1xEVDO network, and then better throughput efficiency and traffic differentiation is achieved by introducing new TBAS method. TBAS algorithm utilizes BE traffic classes to exploit the multi-user diversity by combining it with multi-class diversity. TBAS method uses adaptive technique to calculate the scheduler's QoS parameter which is used for BE traffic differentiation. The problem with existing scheduling algorithm is that existing technique uses pre-determined constant QoS parameter to classify the traffic. Used QoS parameter does not reflect the changes in channel or user conditions. Therefore, scheduler can not efficiently exploit the existing channel information for more accurate scheduling decisions and can not provide high sector throughput. TBAS solves this problem by using average values of each BE traffic class throughput to calculate QoS parameter of each class. As average class throughput is a function of channel capacity, TBAS implicitly uses the channel information to calculate accurate QoS parameter for each traffic class. The performances of proposed and existing scheduler are evaluated in terms of average sector throughput, cumulative distribution function (CDF) of BE class throughput values, and throughput fairness. The performance characterization of existing scheduler is investigated through real network experiments conducted in live CDMA2000 1xEVDO network. Throughput measurements, received mobile terminal power, and signal-to-noise-plus-interference ratio (SNIR) were evaluated in the experiments. Since TBAS is out novel scheduling algorithm, only computer simulations were performed in order to assess its performance.