A New Metric for Network Load and Equation of State

The performance of networks is generally estimated through characteristic indices such as delivery ratio, latency, delay, jitter and loss. Such indices depend on network parameters such as TTL, buffer size, time and load which are at times highly correlated. Factors impacting network measurements include cross traffic, architectures of intermediate nodes, complex interaction of hardware resources and protocols at various levels, as well as implementations that often involve competing and conflicting requirements. Load-balancing at an intermediate node may result in out-of-order arrivals and excessive jitter, and network congestion may manifest as packet losses or large jitter. Out-of-order arrivals, losses, and jitter in turn may lead to unnecessary retransmissions which increases the load. Network load is mainly the result of bandwidth exploited, buffer availability and processing time at intermediate nodes. Indeed, network load is a key parameter with a potentially large impact on network protocols. Besides, Network load is responsible for traffic delay and is sometime confused with congestion. In brief, load is an objective network index, whereas congestion is a subjective measurement from the perspective of a user. With the growth of the Internet in size, speed and traffic volume, and the advent of new network architectures such as MANets/DTN, understanding the impact of underlying network resources and protocols on packet delivery and application performance has assumed a critical importance. Measurements and models explaining the variation and interdependence of delivery characteristics are crucial not only for efficient operation of networks and network diagnosis, but also for developing solutions for future networks. In this paper we give an analytical definition and propose a novel metric to measure network load for mobile ad-hoc and traditional networks by means of an equation of state.