Netlets for end-to-end delay minimization in distributed computing over the internet using two-paths

In a number of distributed computing applications, messages must be transmitted on demand between processes running at different locations on the Internet. The end-to-end delays experienced by the messages have a significant "random" component due to the complicated nature of network traffic. We propose a method based on delay-regression estimation to achieve low end-to-end delays for message transmissions in distributed computing applications. Two-paths are realized between various communicating processes in a transparent manner. Our scheme is implemented over the Internet by a network of NetLets, which communicate with one another to maintain an accurate "state" of delay-regressions in the network. NetLets handle all network traffic between the processes and also perform routing at a certain level depending on the underlying network. We present experimental results to illustrate that NetLets provide a viable and practical means for achieving low end-to-end delays for distributed computing applications over the Internet.