HSDPA networks are currently being deployed; however, there is little knowledge about how these networks perform and behave, and which will the Quality of Service and Quality of Experience that users will achieve due to the fact that UEs share the downlink channel. Furthermore, HSDPA planning and dimensioning is being done through the traditional mechanisms to plan and dimension UMTS networks. These mechanisms do not provide, though, accurate results for HSDPA. This thesis will focus on doing progress in these two areas.
A HSDPA simulator was built to find some answers. This simulator used a simplistic model to simulate the radio environment and HSDPA features at Node B. Besides, the simulator dynamically created web browsing traffic according to the traffic patterns specified by the 3GPP. Three main simulations were performed. First, the maximum number of HSDPA users that a HSDPA network can support was obtained for different mean cell throughputs. Results also showed that the relationship between the mean cell throughput and the maximum number of users is linear. Second, the effect of the amount of UEs in a HSDPA network was studied. Results showed how the network and end-user performance changed when the number of UEs differed from the maximum number of UEs. Simulations demonstrated that network and end-user performance decreases rapidly and significantly when the maximum number of UEs was exceeded. Finally, the mean session inter-arrival time was modified to observe how this traffic parameter affected the network and the end-user performance. Furthermore, different sets of number of UEs were used to find out any correlation between the number of UEs and the mean session inter-arrival time. Results showed how the mean session inter-arrival time was much more relevant for the network and end-user performance when the maximum number of UEs had been exceeded.
Results will give a glimpse of how HSDPA can perform in real networks. Besides, this simulator can help operators and providers to plan and dimension HSDPA networks more accurately.