Adaptive display virtualization and dataflow model selection (ADVADAMS) for reducing interaction latency in thin clients

Abstract

The advent of cloud computing has driven away the notion of having sophisticated hardware devices for performing computing intensive tasks. This feature is very essential for resource-constrained devices. In mobile cloud computing, it is sufficient that the device be a thin client i.e. which concentrates solely on providing a graphical user interface to the end-user and the processing is done in the cloud. We focus on adaptive display virtualization where the display updates are computed in advance using synchronization techniques and classifying the job as computationally intensive or not based on the complexity of the program and the interaction pattern. Based on application, the next possible key-press is identified and those particular frames are pre-fetched into the local buffer. Based on these two factors, a decision is then made whether to execute the job locally or in the cloud or whether we must take the next frame from the local buffer or pull it from server. Jobs requiring greater interaction are executed locally in the mobile to reduce interaction delay. If a job is to be executed in the cloud, then the results of the processing alone are sent via the network to the device. The parameters are varied in runtime based on network conditions and application parameters to minimise the interaction delay.