Quality of Service Middleware Design Guidelines

As users continue to access the Internet from a growing set of heterogeneous access devices, they demand Web services tailored to their personal preferences and usage type (business or personal, for example). Client Internet access devices range from traditional workstations and PCs to laptops, personal digital assistants, and smart phones, and connectivity can be wired or wireless and continuous or intermittent. User requirements and device heterogeneity call for Web services with differentiated quality of service (QoS). In particular, services with response-time constraints, such as video-on-demand (VoD), require the differentiation, control, and dynamic adaptation of QoS. Service providers and network operators need methodologies and mechanisms for managing runtime QoS. Although several recently proposed ad hoc protocols at the network layer have proven effective in limited networks,1 they are incompatible with the Internet’s best-effort model. Their implementation requires intermediate routers traversed by service packet flows, which will likely incur a long process of acceptance and diffusion. In addition, the protocols work at an abstraction level that complicates the embedding of functions such as application-specific adaptation and secure billing.2 The design, implementation, and deployment of QoS-aware Internet services can significantly benefit from a middleware approach at the application level.3,4 (See the “Related Work on QoS for Internet Video-on-Demand” sidebar.) In VoD services, the middleware should exhibit several enabling properties: