This chapter introduced and discussed network virtualization as one of the new en- abling tools for the future Internet/networks. The main focus of the work is wire- less virtualization, specifically in mobile networks. The advantages introduced by
the LTE virtualization based on spectrum sharing between different operators is studied and highlighted. An innovative wireless virtualization framework is pro- posed for the virtualization of LTE. However, the framework is not restricted to only LTE and is open to other similar infrastructure based wireless communication systems, e.g., WiMax. The framework was evaluated in several proof of concept scenarios aiming at showing the benefits of virtualization in the wireless domain: mainly in having better multiplexing gain due to spectrum sharing, as well as bet- ter overall system utilization due to the full exploitation of the multi-user diversity gain. In addition to the above achievements, further advantages of LTE virtualiza- tion can also be seen: for example, sharing the infrastructure of the LTE system is a considerably large gain because it leads to having less equipment and thus reduces the power consumption. It also reduces the capital investment since the operator does not need to buy the equipment but only lease virtual resources, which opens the market for small operators. Another advantage of LTE virtualization is the flexibility the virtualization brings into the operator’s networks, the operator can shrink/expand and change his virtual network on the fly very easily which is one of the major advantages of network virtualization.
Scheduling in LTE is one of the main functions of the eNodeB. The eNodeB aims at achieving several goals regarding its scheduling functionality. First of all, the eNodeB targets cell throughput maximization, as well as being able to serve as many active users as possible. This is one of the important features that mobile network operators try to achieve, as it is related to their revenue. The second im- portant objective is to satisfy the user by guaranteeing the QoS of their individual services. Looking at the scheduling problem within LTE, it is very easy to find that it is a complex multi-dimensional problem with conflicting targets and there- fore a proper optimization and trade-off is required which can be achieved by the scheduler design.
As explained earlier, LTE uses OFDMA and SC-FDMA as the multiple ac- cess schemes, for both the downlink and uplink respectively. This means that the resources that the scheduler needs to distribute have both a time and a frequency component. The LTE radio scheduler is responsible for distributing a fixed number of PRBs (obtained from dividing the total system bandwidth) over a varying num- ber of users/bearers in a fixed time interval, so-called for LTE, the Transmission Time Interval (TTI) which is set to 1ms. This scheduling has to be done for both the downlink and uplink direction, where the downlink is mainly constrained by the total transmission power of the eNodeB; while in contrast, the main constraint on the uplink arises from the multi-cell view of inter-cell interference [SBT09]. In LTE, a user can have in general several bearers, one for each service. For example, a user can have a phone call while browsing the web and downloading a file at the same time, which means for this example the user will have three different bearers. However, within this thesis work, the focus is only on a single bearer users, i.e., all the investigations done within this thesis are performed with only one bearer per user. That is why the terms user and bearer are used interchangingly within the thesis.
In this chapter, the LTE radio scheduler is described in detail. First of all, the general scheduling framework is introduced, followed by a state of the art survey on LTE schedulers. Then the LTE radio scheduler proposed and developed in this thesis is presented and explained in detail. The proposed scheduler is called “Op-
timized Service Aware Scheduler” (OSA) which is a proportional fair downlink scheduler with the motivation of providing service differentiation and guarantee- ing quality of service (QoS). Finally, a detailed simulation analysis is presented, highlighting the achievements and performance gain of the proposed scheduler.
The proposed LTE downlink radio scheduler design, achievements and results are published in [ZZW+11] [ZWGTG11b] [ZZW+12].
6.1 LTE Dynamic Packet Scheduling
Two of the main functions of the LTE radio scheduling are the Dynamic Packet Scheduling and Link Adaptation. In the dynamic packet scheduling, the time- frequency resources (i.e., Physical Resource Block (PRB) see section 3.4.3) are distributed between the different active users and their corresponding packets are scheduled at the MAC Layer, the number of packets being scheduled depends on the Modulation and Coding Scheme (MCS) that the scheduler determines. The general packet scheduling framework is shown in Figure 6.1.
Time Domain Scheduling (TD)
Frequency Domain Scheduling (FD) Packet Scheduling
Outer Loop Link Adaptation Link Adaptation
Inner Loop Link Adaptation MIMO Link Adaptation HARQ QoS attributes Buffer information
Rank CQI (N)ACK
Figure 6.1: General packet scheduling framework [HT09]
As can be seen, there are a number of interactions and input parameters be- tween the different entities. Both the HARQ manager and the link adaptation are connected to the packet scheduler, since the scheduler needs to know whether to schedule a new transmission for a certain bearer/user within this TTI, or a pending
retransmission is required (as a user cannot do both at the same time). It can be seen that a number of input parameters are also required, such as: QoS attributes, buffer information and some link parameters that are fed back from the users in the uplink in the form of Channel Quality Indicator (CQI).
The packet scheduling is divided into two different stages: the first stage is the Time Domain Scheduler (TDS) and the second stage is the Frequency Domain Scheduler (FDS). The driver behind such a split is mainly simplicity. The designed scheduler has to work in real life equipment at the end of the day, where resources are limited and decisions have to be made within the one millisecond time interval. Thus, the scheduler design and implementation have to be as simple and efficient as possible. In the LTE literature such a split is recommended and is normally called decoupled time and frequency domain scheduler [KPK+08a].