QoE Tuning for remote access of interactive volume visualization applications

Remote access of interactive volume visualizations such as e.g., remote MRI (magnetic resonance image) viewing and inspection with three-dimensional images is important in smart health care applications. However, due to the large scale of data sets involved in the computation and various network/system factors (i.e., network bandwidth, CPU/GPU), delivering satisfactory user Quality of Experience (QoE) for remote access is quite challenging. In this paper, we investigate tuning of user QoE based on controllable parameters such as data transmission rate i.e., the client-side encoding scheme selection, and the computation resource scale i.e., the GPU server hardware size/number. The novelty of our studies is in the joint use of a 'network-aware encoding scheme ' on the client-side along with an 'encoding-aware server scaling' on the server-side to guide efficient tuning decisions within a remote access system. We also describe a 'Remote Interactive Volume Visualization System' (RIVVS) case study and analyze utility functions (e.g., bandwidth consumption, GPU utilization) that guide the design of a tournament scheme for subjective testing in an application-specific context. Our RIVVS testbed results with human subjects show that our approach can help in efficient tuning of remote MRI access configurations with satisfactory user QoE for: (a) good-to-poor network health conditions, (b) low-to-high remote access user workloads involving a diverse set of thin clients such as personal computers, smart phones and tablets.