Service-Class Based Scheduling for Long Term Evolution (LTE) Networks

Abstract

LTE (Long Term Evolution) is a wireless communication standard providing support for IP-based traffic as well as end-to-end Quality of Service (QoS) with description of strict margins as to how QoS for different service classes should be maintained. These service quality related requirements including delay budgets and packet loss rates are listed in a standard service class table for LTE. Due to the stringent QoS requirements, simplest schedulers adopted from literature like Proportional Fair, Maximum Throughput and Exponential Rule cannot be applied directly to LTE Networks. Moreover if these schemes are used as such, the non guaranteed traffic is bound to suffer from resource scarcity whilst a number of best effort flows are always present in the network for major scenarios. To achieve a minimum resource level guarantee for service classes, specifically the best effort traffic while satisfying the delay budget requirements, we propose a service delay-budget and channel quality aware LTE network scheduler. The scheduler is implemented for the downlink with a bound on the number of resources that can be assigned to each class. These bounds are dynamically tested for delay budgets and packet loss rates to determine the number of accommodated and sustainable users of different service types. For example in a 10MHz spectrum usage, the number of accommodated users in a 7:3 resource ratio for Guaranteed (Video, VoIP and CBR tested individually) and Non-Guaranteed (Best Effort) traffic came out to be ~ 14, 280, 33 and 480 using theoretical model with delay budget of 100ms. These were validated using simulations with close approximations of ~ 12, 244, 27, and more than 400 users for each category while running Video at 242kbps, VoIP at 12kbps, CBR at 100kbps and BE at 3kbps. Moreover the data rate of best effort is sustained within specific levels corresponding to the assured resources even when the system is loaded with guaranteed users. The throughput of best effort using normal schedulers goes down to zero after the resources get mostly occupied by guaranteed service. Using delay budget parameters, the delay for services is also maintained within margins. Admission Controller for LTE Network forms one of the direct applications of this work. Capacity based Admission Controller design is also discussed in terms of the resource calculations needed to accommodate and sustain specific number of QoS subscribed users.