QoS-based service selection

随着web服务的激增，如何识别和选择最符合用户需求的服务已成为一个巨大的挑战。通常在找到功能匹配的服务后，进一步的选择和排序是基于非功能属性，如服务质量（QoS）。基于qos的服务选择已经有了很多研究工作。然而，还有一些独特的挑战尚未完全解决，需要做更多的工作。在本提案中，我们希望解决其中的一些问题，包括模糊和灵活的QoS要求、服务的动态性质以及相对的用户满意度水平。我们的目标是设计并实现一个基于qos的服务选择系统，该系统将浏览和搜索作为集成的选择活动，将时间作为服务选择的重要因素，并通过对用户历史行为的学习来提高体验质量（QoE）。主要包括以下几个步骤：1)设计一种交互性的QoS浏览机制，使用户能够随时了解当前服务的QoS分布情况，并选择自己关注的领域获取更多细节，有望在一定程度上解决需求模糊和服务动态性问题；2)定义一种表达性的查询语言，评估不同的排序模型以选择最合适的模型，并可能进行一些修改以匹配提议的查询能力；3)定义考虑用户个体独特偏好和满意度曲线的QoE模型，理解满意度始终是一个相对主观的概念，并利用该模型提高选择性能。该方案的主要特点是：1)通过提出的选择系统，用户可以通过浏览来明确自己的QoS需求和平衡不同的属性，并且搜索的查询定义更清晰，结果排序更合理，整体满意度提高；2)培养对服务选择、QoS和QoE、使用模式分析等课题有深刻理解的高素质人才，并在系统构建和系统实验方法上有具体经验。

With the proliferation of web services, how to identify and select services best matching users' requirements has become a big challenge. Usually after functionally matching services have been found, the further selection and ranking are based on non-functional properties such as Quality of Services (QoS). There have been many research works on QoS-based service selection. However, there are a few unique challenges which haven't been fully addressed yet, and more works should be done. In this proposal, we would like to address some of them, including the vague and flexible QoS requirements, the dynamic nature of services, and the relative user satisfaction levels. Our objective is to design and implement a QoS-based service selection system, in which browsing and searching are considered as integrated selection activities, time is taken as an important factor for service selection, and Quality of Experience (QoE) is improved through learning from the user history behavior. There are several major steps involved: 1) design an interactive QoS browsing mechanism so that users could always understand the current QoS distribution of services and choose their concerned areas to get more details, which could hopefully solve the vague requirement and service dynamism problems to an extent; 2) define an expressive query language, evaluate different ranking models to choose a most appropriate one and possibly make some modifications to match with the proposed query capabilities; 3) define a QoE model which considers individual user's unique preferences and satisfaction curves and understands that satisfaction is always a relative and subjective concept, and use this model to improve the selection performance. The key features of this proposal would be: 1) with the proposed selection system, users could browse to clarify their QoS requirements and balance among different attributes, and search with more clearly defined queries, properly ranked results and overall improved satisfaction levels; 2) we will train the highly qualified personnel to gain a deep understanding on topics such as service selection, QoS and QoE, and usage pattern analysis, and they will also have concrete experiences on system building and systematic experimental methods.