Decentralized service-level monitoring using network threshold crossing alerts

Service level agreements are at the core of the business relationship between service providers and their customers. Service level monitoring is necessary to validate and ensure that service levels are indeed adhered to. One key tool for this is threshold crossing alerts. TCAs can notify a service provider that a certain parameter has exceeded a certain threshold value, directing attention to those areas where preventive action needs to be taken. This article presents a protocol and an architecture that implements a new category of TCAs for parameters that need to be aggregated across a network, as opposed to parameters that can be observed from a single device; for example, "average utilization across all links in a network" as opposed to "utilization of a particular link." Although highly useful, such parameters are rarely subjected to SLAs today, due to the lack of effective monitoring technology. Our system fills this gap. It does so in a manner that is decentralized inside the network and does not rely on a more traditional centralized management architecture. We focus on business application aspects, as well as the robustness and accuracy properties of our system