Networks of the future will evolve instead of being designed by engineers, according to Yaling Yang, an assistant professor of ECE. She is leading a Virginia Tech effort to replace today's ad hoc, manual process of network design with automatic system design and adaptation.
"Traditionally, network engineers have designed network systems, based on complex collections of objectives, policies, principles, and past experiences, she says. "The design process is top-down and based on a set of over-simplified assumptions about the operation environments and performance objectives." Because of limited human experience and capability, this manual design process makes it impossible to exploit and evaluate the entire design space for network systems, resulting in extreme challenges for identifying the best designs and promptly responding to network environment changes, she explains.
Instead, she says, a network could be automatically assembled from a set of reusable "genes" and tested in its actual environment. In this scenario, a gene is a small piece of computer code that implements a particular design for a small part of a network component, such as an intrusion detection system, a routing protocol, or a transmit power adaptation strategy. The selection of genes in the automatic process would be based on performance analysis and guided by an evolution process based on interoperability and compatibility theory, optimization theory, and game theory.