Overview

The drawbacks of centralized scheduling can be resolved by using distributed management schemes. With distributed approaches, each user determines the amount of resources to be used based on local information. Compared to centralized schemes, the main difference of distributed resource management schemes is that there are multiple decision makers in the network, whose decisions may conflict with each other, therefore resulting in performance degradation.

With distributed resource management, the network does not need infrastructure support and therefore can be easily constructed. This is very helpful in some application scenarios. For example, some sensor networks must operate in remote areas and harsh environments, without infrastructure support or maintenance. Sensor nodes must manage themselves autonomously and configure themselves to operate and collaborate with other nodes. In addition, each sensor node must be able to adapt to failure, topology, density, environment, and traffic characteristics automatically without too much overhead. Furthermore sensor networks usually have large scales, which make it not feasible to rely on centralized implementations, because centralized approaches may incur exhaustive overheads for adaptations that may affect all users in the network. Instead, sensor nodes must make local decisions without global knowledge. Therefore, it is essential to design distributed resource management schemes for sensor networks.

A more detailed example is the Request to Send (RTS)/Clear to Send (CTS) option in the IEEE 802.11 standard. In this case, a user that has packets to transmit sends a RTS frame to the destination user.