Majority of QoS research has concentrated on analyzing network resource consumption and translations between network service demands and network service parameters, For example, [16] illustrates similarly how to design an architecture to provide end-to-end QoS for applications distributed across a network. In the Omega architecture, QoS parameters are translated between application and network by a QoS broker [16]. [11] presents an ODP view of QoS and proposes an environment contract to specify QoS characteristics of computational objects. The authors also provide a translation of these specifications to ATM related QoS parameters. These network parameters incorporate time related characteristics that record time delay (cell interarrival time, cell delay variation, cell transfer delay),and capacity related characteristics that record communication throughput (for e.g. cell transfer capability, mean cell rate), guarantee levels, channel availability(priority level, cell loss probability) and accuracy (for e.g. cell sequence integrity, cell loss ratio, cell error ratio, cell misinsertion rate, severely-errored cell block ratio, cell insertion rate, bit error probability). The TENET project [1, 8] classifies services into deterministic service, statistical service, predicted service and feedback based schemes to deal with tradeoffs in QoS, network utilization and overload. In [18], a method for modifying the HTML protocol to reduce communication latencies is proposed. In collaboration with SANDIA national labs, the work characterizes video-conferencing applications (trying to understand network support for such applications) and determines QoS parameters for MM networking. [2] describes protocols for connection establishment for supporting real-time multiparty applications like video conferencing with guaranteed QoS. [13] addresses issues of distribution and duration of metrics like call-completion rate in the telecommunication world.
Many approaches for QoS support refer to an existing infrastructure or environment. For example, systems that focus on QoS management and its mapping on to the transport system include the Heidelberg High Speed Transport System [28], and the Lancaster QoS architecture [5]. The approach presented in [6] deals with QoS support in a heterogeneous environment with diverse communication requirements, varying levels of QoS in terms of latency, bandwidth, jitter etc. The need for maintaining a uniform view of the entire system to account for heterogeneity in the network model, request model etc. is emphasized. Mechanisms discussed include the scaling of media streams in terms of picturesize, content,picture rate, cost etc. as an effective technique to deliver acceptable perceptive QoS and avoid overloading a system with unnecessary information.
However, there is less research in the area of determining user-satisfaction. [12] discusses the multilevel specification of QoS factors like synchronization, interactivity, availability that are applicable to a range of applications. The quality function presented in [25], is a comparable effort in providing parameters of user-satisfaction. The approach focuses on preserving fidelity of presentation and hence formalizes presentation features that correspond to parameters in the US function such as jitter and end-to-end delay. An explicitly defined presentation error model serves to assign consistent values to quality parameters that can be used to define the weights represented in the metric we propose in this paper. While the approach proposed by [23, 24] is device and implementation independent, the weighted cost-satisfaction ratio adds (a) a notion of devices and resource consumption and (b)an attributed cost to a service request.