Performance of Streaming Services on HSDPA

  • Gutiérrez, Pablo Jose Ameigeiras (Project Participant)
  • E. Mogensen, Preben (Project Participant)

Project Details


Multimedia streaming services are becoming increasingly popular among Internet users and are expected to play a decisive role in the introduction of packet data services in Third Generation (3G) mobile networks. These services impose stringent constraints on the conveying networks in order to satisfy their real time QoS requirements. On the other hand, the increased spectral efficiency of High Speed Downlink Packet Access (HSDPA) motivates the mapping of streaming services on to this technology. However, the time-shared nature of HSDPA is very well suited for Non Real Time (NRT) bursty traffic such as Interactive and Background, but the transport of Real Time (RT) traffic over HSDPA faces important challenges in order guarantee its exigent delay requirements. The Modified Largest Weighted Delay First (M-LWDF) has been selected as the Packet Scheduling algorithm for delay-sensitive traffic because it includes both the instantaneous channel quality and the queuing delay in the user?s priority computation. Focusing on Constant Bit Rate (CBR) encoded video streams, from a system level performance point of view, it has been shown that the M-LWDF algorithm is a rather unfair scheduling principle where the users with poor average radio propagation conditions, which suffer from higher delays than the remaining users in the cell, are not able to fulfil the QoS criteria during high load situations. More fair Scheduling algorithms have been proven to degrade the system performance. The conveyance of streaming connections with a source data rate of 128 kbps only represent a minor cell capacity reduction compared to NRT traffic using the Proportional Fair algorithm and guaranteeing a minimum data rate of 128 kbps. Regarding the system capacity sensitivity to the delay jitter introduced by HSDPA, it has been shown that its impact on the system level performance is not as large as might be initially expected. A delay jitter increase from 1 to 4 seconds in a Pedestrian A and 3 km/h environment produces a cell capacity gain of around 35% gain, while the same delay jitter increase in a Vehicular A and 3 km/h environment yields a 20% gain. It is concluded that a jitter of 3 seconds represents a good trade-off between HSDPA cell capacity and Node B & UE buffering requirements. In terms of the power delay profile, the time dispersion of the Vehicular A model yields a cell throughput reduction of 55% versus the Pedestrian A model (for a delay jitter of 3 seconds). (Pablo Ameigeiras, Preben E. Mogensen; Jeroen Wigard, Nokia)
Effective start/end date31/12/200331/12/2003