Active middleware for Internet Video on Demand: the QoS-aware routing solution in ubiQoS

Abstract

Several factors are forcing to address the issues of differentiated Quality of Service (QoS) and ubiquitous accessibility in Internet services, from growing user requirements to the increasing heterogeneity of access devices, from the competition of service providers to the severe constraints of resource availability in emerging wireless environments. The paper claims that the provision of services with negotiated and controlled QoS over best-effort networks is achievable via distributed support infrastructures that activate some of the nodes along the network path between clients and servers. The paper proposes mobile agents (MAs) as the activation technology to implement the needed active infrastructure and the MA-based design and implementation of the ubiQoS middleware for Video on Demand. At the starting of the service session, ubiQoS establishes an active path of intermediate nodes capable of tailoring multimedia flow QoS depending on profiles of user preferences and of device characteristics; at provision time, ubiQoS monitors the offered quality and promptly react to changes in resource availability by locally performing multimedia transcoding/downscaling and resource pre-emption. In particular, the paper focuses on how it is possible to determine dynamically a QoS-aware active path between clients and servers, discusses alternative solutions, and evaluates the performance results of the completely decentralized Peer-to-Peer implementation of the active path determination in ubiQoS. q

Mobile Middleware Solutions for the Adaptive Management of Multimedia QoS to Wireless Portable Devices

“…In addition, MAs can simplify the QoS adaptation in response to modifications in the availability of network resources at provision time [9]. For instance, MAs can dynamically migrate where needed to monitor network resources locally; this visibility of local monitoring indicators permits to trigger management operations to correct the actually supplied QoS (re-negotiation, format transcoding, additional communication channels for VoD distribution, …) by exploiting locality to congested resources.…”

“…Even if not specific of the wireless Internet domain, the availability of these solutions is important and can significantly leverage the adoption of MAbased middlewares for pervasive computing [9,13]. Due to the novelty of the MA technology, however, there are not many MA environments already employed to implement middleware solutions for the wireless Internet.…”

“…In server-based tailoring solutions, the service provider is in charge of either selecting the version of the requested VoD flow with the most suitable QoS level among a pool of statically pre-determined ones (offline tailoring) or dynamically operating downscale transformations on the unique high-quality stored version (on-the-fly tailoring). Let us additionally note that the server-based approach makes distributed caching infrastructures less effective because VoD flows are tailored since the beginning to fit the specific client requirements [9]. On the opposite, client-based tailoring decentralizes the different charges: it is the client site that downscales the VoD flow before visualizing it, with a significant engagement of client computational resources.…”

“…Active nodes monitor and control QoS levels during service provisioning and react promptly with service management operations in case of network congestion. In addition, any active node can cache the traversing VoD flows before their possible local QoS downscaling, thus permitting the realization of an effective distributed caching infrastructure [9]. MU extends significantly the applicability of the ubiQoS middleware to portable access devices with wireless connectivity and strict limits on local resources.…”

“…On the other hand, MU provides portable mechanisms to give direct location visibility mechanisms to give direct location visibility to VoD clients when interested in developing novel locationaware service clients. Other details about the MU architecture and implementation can be found in [9].…”