Clean Slate Research Projects
Virtual worlds are networked, multi-user, simulated environments. They simulate physical interaction and decouple this interaction from geographic constraints. Users participate through avatars, manipulating objects and communicating with others. Virtual worlds have a wide variety of uses. Probably the most popular use is in gaming, but virtual worlds have applications ranging from simulation for military training, social interaction, teleconferencing and remote collaboration, distance learning, social science research, and even epidemiology. We are investigating ways to make virtual worlds scalable to global levels while addressing the significant limitations of current systems: efficient, scalable architectures, security, privacy, high quality content creation, extensibility, and integration with other systems.
Virtual world systems are very demanding on the network and push the Internet as currently deployed to its limits. We are using a programmable virtual infrastructure to investigate how we can build a better network to support such demanding applications. Programmable virtual infrastructures provide programmable components at many network layers. Using this infrastructure we're investigating ways to make virtual worlds more efficient. For example, multicast is not currently widely deployed, but we expect virtual worlds could make great use of this capability since the majority of data sent from servers to clients are simple status updates which are identical for each client. By targeting a PVI for deployment we will easily be able to implement and deploy custom routing algorithms, including multicast. The PVI makes it possible to easily gather large scale usage statistics in a realistic setting so we can evaluate the effectiveness of our new approaches. We're also considering ways to improve efficiency and enable new capabilities in other parts of the system. We are investigating routing algorithms based on higher level identifiers such as unique object IDs or location instead of server addresses, meaning routing destinations could change quickly if objects cross server boundaries. Our new approaches must be responsive to these changes, even if they occur as a packet is being routed, in order to maintain the illusion of a consistent world. We are also investigating new content distribution approaches with capabilities targeted to virtual worlds. For instance, in virtual worlds there is the need to be quickly informed when content changes (e.g. "user X changed their outfit to texture Y"). We are looking into how to improve existing content distribution systems to enable this sort of notification. We expect that there are a number of fundamental changes to networking that could enable virtual worlds to be a more efficient and effective new mode of communication.
(Faculty) Assistant Professor of Computer Science