Seamless Application Handover Across Radio Access Networks (SAHARA Net)

The explosion in mobile and data traffic in the last decade has led to a rapid proliferation in wireless networks. A plethora of wireless access technologies are available today each with a different offering. Some offer high data rates within a restricted coverage area such as 802.11 hotspots. Others, offer lower data rates but with a much wider coverage such as UMTS. This diversity can be harnessed in a way that creates a ubiquitous communications platform for the user. This is the premise of the heterogeneous networks vision/architecture: an environment where disparate technologies cooperate together and complement each other. However, there are various technical challenges in the way of such convergence. The first obstacle is enabling communication between disparate mobility protocols. Once this is achieved, the diversity of networks in itself poses a challenge for the user as to which network he connects to. This thesis answers the first question by reviewing the low-latency handover literature to identify the most credible solutions. The general consensus amongst researchers in the field has been to bridge the gap between the network and link layers so that IP protocols can react quickly to link changes. To answer the second question, this thesis defines a framework to assess handover decision algorithms based on application performance. The merit of the handover algorithm’s decision is measured by how well the application performs after handover. In order to facilitate this process, a simulation module was created within the NS2 network simulator that allows mobile devices to collect network measurements and feed that information into a decision algorithm to decide whether or not handover should be triggered. Through this evaluation process, a number of issues emerged as possible stumbling blocks. The first such issue is the inconsistency between local network conditions measured at the Access Point or Base Station, and the end to end conditions experienced by the user’s application. Another issue is the algorithm’s adaptability to user and application preferences. Personal users might be cost aware opting to trade off quality for a lower cost in certain circumstances. The Abstract IV handover algorithm must be able to accommodate such scenarios. Furthermore, algorithms must be able to adapt their decisions according to the application’s requirements. Using application profiles with thresholds or utility functions can result in better decisions than using absolute values. If an application is satisfied with the current network conditions, it might not be in its benefit to move to a different network even if it offered better conditions. In fact, it might suffer as a result of possible handover disruptions