What QoS research hasn't understood about risk

Abstract

In an effort to meet application needs, network researchers have designed internet quality of service (QoS) architectures intended to deliver end-to-end performance guarantees to applications. Because such services offer guarantees and must elevate and isolate select traffic from competing traffic, we refer to them as “elevated services ” (this is in contrast to forms of QoS that differentiate two or more best-effort service classes none of which is strictly elevated over another [3]). Despite two decades of vigorous research and standards activity, elevated services have failed to deploy. We argue that a fundamental reason for this failure is a confusion between QoS as an underlying technology and QoS as a service offering. Neither customers nor Internet service providers need or want hard performance guarantees. Instead, each wants tools to understand and manage risk.