We address the problem of managing cloud applications, consisting of a set of virtual machines (VMs), characterized by bursty and dynamic workloads, in such a way to provide guarantees on their Quality-of-Services (QoS) and, at the same time, to minimize the energy consumption of the physical infrastructure running them. We propose a fuzzy controller, Fuzzy-Q& E, that is able to allocate to the VMs of each cloud application the minimum amount of physical capacity needed to meet its QoS requirements. In this way, the number of physical resources that must be switched-on at any given time is reduced with respect to the case in which physical machines are statically provisioned and, consequently, less energy is required to run a given cloud workload. We implement a prototype of our controller on a Xen-based testbed, and we perform a set of experiments using an E-Commerce benchmark in which we compare Fuzzy-Q&E against Dyna QoS, a state-of-the-art fuzzy controller for virtualized resources. Experimental results show that Fuzzy-Q&E out performs Dyna QoS both in terms of the ability of meeting the QoS level of the application, and of the amount of physical capacity allocated to each VM.
In this article we propose a novel approach for group-decision making, i.e., proposing and selecting solutions to issues discussed in online settings, based on the structure of the discussion. Our approach consists of three steps: (1) assigning weights to participants based on formal criteria such as degree of engagement in the discussion, (2) assigning scores to comments, considering the weights of authors and raters, (3) assigning scores to proposals, based on the scores of the pro and contra arguments. An important point is that individuals whose behavior is in line with our formal criteria have a higher influence on the decisions. Having built a respective online platform, we have evaluated the proposed model by means of an experiment with more than 100 participants who have discussed several topics relevant to them and a subsequent survey. Looking at both the results of survey and the discussion outcomes, we conclude that our approach yields comprehensive discussions and outcome decisions supported by the community.
Micro logging is emerging as a suitable means for question-answering in working settings. This leads to different efforts to seamlessly integrate microblogging into the daily-used tools. Specifically, microblogging is being regarded as particularly useful during software development, akin to the tradition of Q&A forums. This paper looks at a particular kind of software: the one being developed by domain experts through the use of Domain Specific Languages (DSLs). We believe this setting is specially amenable to benefit from Q&A microblogging due to inherent limitations of the target audience. This brings the twist of domain specific ness into microblogging, i.e. the Q&A process is now framed by the semantics of the DSL constructs. This permits the introduction of editing assistants that embed domain knowledge about the kind of questions that can be posed, and the way answers can be selected. This opens an opportunity for more focused and assisted microblogging. This paper introduces Crowd Call, an in place microblogging mediator for DSL editors. The aim is to make microblogging a natural gesture during the conception of the DSL expressions, making transparent the interplay between the DSL editor and the Social Networking Services. In addition, Crowd Call can be configured to the constructs and resolution strategies of the DSL at hand so that questions and answers are framed by the semantics of the DSL. The approach is illustrated for three DSLs: the Google Spreadsheets formula language, SQL and Sticklet. We show how Crowd Call-mediated microblogging is tuned for the semantics of each DSL.
Current applications tend to use personal sensitive information to achieve better quality with respect to their services. Since the third parties are not trusted the data must be protected such that individual data privacy is not compromised but at the same time operations on it would be compatible. A wide range of data analysis operations entails a similarity detection algorithm between user data. For instance clustering on big data groups together objects based on the heuristic that similar objects are likely to be put under the same cluster. Similarity decisions are important for numerous applications such as: online social networks, recommendations systems and behavioral advertisement. In this paper we propose a mechanism that protects user privacy and preserves data similarity results although encrypted. We analyze the security of the scheme and we further demonstrate its correctness and feasibility through a real life experiment where personality traits by users are collected for a 4square application.
Multi-tenancy is an architectural approach for sharing one application instance among different tenants. It is used to reduce operational costs of SaaS offerings. The major drawback of this approach is the potential mutual influence of one tenant onto the performance experienced by others. In this paper we present a combination of a traditional scheduling mechanism and a feedback control loop based controller to ensure performance isolation while efficiently utilizing the system and faster reaction to workload changes than a control cycle's duration.
