The recent and continuing advent of the World-Wide-Web as the means of choice for communication, advancements in distributed computing systems, new media techonolgy, and software platform-independence, have given rise to the idea of establishing a computing environment whereby the user's computer "is the network." In this project, we take our research group's approach to the network computer, which we call the "hollow client", and charcterize some of the issues involved in feasibly making this computer a powerful window to the computing resources throughout the net. We see this accomplished by placing most of the computing tasks on the net, while keeping enough compute performance on the user's machine to guarantee satisfactory real-time performance. It is our view that even with sufficiently high network bandwidth, and ideal memory system at the user's end, response time for the interactive tasks that interface the user with the network will set a lower bound on the performance requirement on the client machine. With a limit set on average network transfer size, and a limit on the frequency of network access relative to other locally executed tasks, the dependence on network performance can be minimized. It is then under such premises that the local performance lower bound will be the governing factor of the hollow client feasibility.
2.2 define the common operations
4.0 Technical Results and Analysis
4.1.2 Instruction-level Parallelism
4.1.3 Minimum MIPS requirements to achieve real-time performance
4.2.2 Effect of Network Access Frequency on Network Access Performance