Design, Implementation and Performance Evaluation of a Scheduling Algorithm for HSDPA System

The first mobile phone systems, introduced in the early 1980s, were based on analog interface technology with voice-only capabilities. These systems are limited in bandwidth and are low in quality. The increased need for more features was addressed by the introduction of the second generation (2G) network. 2G mobile phone systems, such as Global System for Mobile Communications (GSM), were characterized by digital circuit switched transmission and the introduction of advanced and fast phone to network signaling. However data services that use packet data technology were limited. The data packet service enhancement was introduced with 2.5G systems such as General Packet Radio Service (GPRS). GPRS introduced the packet switched domain to the Core Network (CN) of the GSM. The main features of 2.5G are the possibility to be permanently on line, only paying for the information received and a radio interface that typically provides a maximum data rate of 30-40 kbps with latency of 600 ms. The growing number of applications with quality of service (QoS) requirements, like web browsing, music and game download, mobile TV, Real Time (RT) video sharing, push e-mail, on-line gaming, mobile weblog, Voice over IP (VoIP) and so on, has led to the development of the third generation (3G) mobile telecommunications systems: Universal Mobile Telecommunication System (UMTS). UMTS is standardized by the 3rd Generation Partnership Project (3GPP). The most common form of 3G mobile systems adopts the Wideband Code Division Multiple Access (WCDMA) air interface. WCDMA Release 99 provides data rate of 384 kbps for wide area coverage and up to 2 Mbps, variable data rate on demand, a 5 Mhz bandwidth and a reduction of the network round trip time. The strong demand for multimedia services will require higher data rate than 2 Mbps, particularly for the downlink side, where mobile users data traffic will increase. In order to meet the increasing demand for high data-rate multimedia services, High Speed Downlink Packet Access (HSDPA) has been introduced by 3GPP in Release 5 of UMTS. HSDPA, labeled as 3.5G wireless system, will permit to increase user peak data rate up to 14 Mbps. This enhancement is possible for the introduction of a new downlink time shared channel among access users, that support a transmission time interval (TTI) of 2 ms, an hybrid automatic repeat request (H-ARQ), an adaptive modulation and coding scheme (AMC) and multicode transmissions. A key design feature in HSDPA concepts is packet scheduling, which is implemented in the Node B. A packet scheduler is responsible for sharing the available resources among the users eligible to receive data. Due to its function, the packet scheduler has a direct impact on the HSDPA system performance. It determines the perceived satisfaction among the user in the cell. As a result, it determines the overall behavior of the system. The main focus of this thesis is the design of an HSDPA scheduler aimed at servicing both RT traffic (i.e., VoIP) and BE traffic (i.e., TCP flows). The existing schedulers focus only on a limited subset of the overall HSDPA aspects. In fact, the design of the scheduling algorithm should take into account not only the channel conditions of the users, but also the way in which the HSDPA resources are assigned, the QoS requirements of different traffic types, the priority functions used for user selection and the backlog of each UEs. In order to devise an effective scheduling algorithm for HSDPA, we have first derived a CQI based power control algorithm that set dynamically the resources to allocate, in response to the CQI reports, in order to guaranteeing a BLER no larger than 10%. Then, we have designed and implemented an HSDPA scheduler that takes different decisions based on the traffic type of each UEs, using different priority functions. It considers BE traffic when no more RT traffic can be serviced (either because no RT PDU is waiting for transmission, or because all active RT flows are unable to access the medium, e.g. due to ongoing retransmissions). On the other hand, RT traffic is given as few resources as possible. More specifically the amount of resources involved in transmitting RT PDU is limited to the amount which is strictly necessary in order to guaranteeing a good reception at UE side. Moreover, each newly arrived RT packet in the Node B is associated with a discard timer and it is discarded when its deadline is missed. All the resources not used by RT flows are then allocated to BE traffic. The performance evaluation has been carried out through simulation using an event-driven ad hoc simulator of the HSDPA Release 6 system written for the Network Simulator 2. Comparison results are provided for VoIP services and for different scheduling algorithm, such as Round Robin (RR), MAX-C/I and a scheduler that we implement based on a work of Nokia Networks