Browsing by department "Tietojenkäsittelytieteen laitos"

In health-care area, computing based solutions are frequently sought for efficient management and operation of hospitals or independent surgeries. The overall surgery scheduling plays an important role in matching of the supply and the demand for the services of a surgery. Internal competition on resources makes a surgery scheduling challenging. A surgery scheduler must consider uncertainty in the duration of each surgery and other critical activities such as the arrival of unexpected urgent patients and cancellations on the day of surgery while efficiently managing a variety of human and technical resources. Surgery schedules must balance the surgery requests and the available resources and at the same time accomplish efficient service to patients. This thesis presents a dynamic surgery scheduling based on the idea of a Virtual Organization creation. Virtual organization (VO) is a collaboration partnership model among small businesses to share resources and to be competitive with larger businesses. Dynamic scheduling of a surgery requires the collaboration of various participant to overcome challenges of coordination. The effectiveness of a surgery schedule can be monitored from the surgeries. Moreover, the inadequacies of the schedule can be detected beforehand and the necessary measures can be taken to avoid waste of resources and cancellation of surgeries due to overbooking. The VO based model is more rigorous than is usually presented in research literature. In this thesis a conceptual model is designed to show the interaction of partners. The model is validated using sample populated data. The roles and dependencies among entities are described in the model that uses fact-oriented approach. Object Role Modeling (ORM) is used to show the interaction of entities and their role. And NORMA is used as a development tool. The validation using fact and sample populated data determine the validity of the conceptual model and demonstrate that the collaboration partners benefit from the coordinated surgery schedule.

Wireless access is expected to play a crucial role in the future of the Internet. The demands of the wireless environment are not always compatible with the assumptions that were made on the era of the wired links. At the same time, new services that take advantage of the advances in many areas of technology are invented. These services include delivery of mass media like television and radio, Internet phone calls, and video conferencing. The network must be able to deliver these services with acceptable performance and quality to the end user. This thesis presents an experimental study to measure the performance of bulk data TCP transfers, streaming audio flows, and HTTP transfers which compete the limited bandwidth of the GPRS/UMTS-like wireless link. The wireless link characteristics are modeled with a wireless network emulator. We analyze how different competing workload types behave with regular TPC and how the active queue management, the Differentiated services (DiffServ), and a combination of TCP enhancements affect the performance and the quality of service. We test on four link types including an error-free link and the links with different Automatic Repeat reQuest (ARQ) persistency. The analysis consists of comparing the resulting performance in different configurations based on defined metrics. We observed that DiffServ and Random Early Detection (RED) with Explicit Congestion Notification (ECN) are useful, and in some conditions necessary, for quality of service and fairness because a long queuing delay and congestion related packet losses cause problems without DiffServ and RED. However, we observed situations, where there is still room for significant improvements if the link-level is aware of the quality of service. Only very error-prone link diminishes the benefits to nil. The combination of TCP enhancements improves performance. These include initial window of four, Control Block Interdependence (CBI) and Forward RTO recovery (F-RTO). The initial window of four helps a later starting TCP flow to start faster but generates congestion under some conditions. CBI prevents slow-start overshoot and balances slow start in the presence of error drops, and F-RTO reduces unnecessary retransmissions successfully.

The Web has introduced a new technology in a more distributed and collaborative form of communication, where the browser and the user replace the web server as the nexus of communications in a way that after the call establishment through web servers, the communication is performed directly between browsers as peer to peer fashion without intervention of the web servers. The goal of Real Time Collaboration on the World Wide Web (RTCWeb) project is to allow browsers to natively support voice, video, and gaming in interactive peer to peer communications and real time data collaboration. Several transport protocols such as TCP, UDP, RTP, SRTP, SCTP, DCCP presently exist for communication of media and non-media data. However, a single protocol alone can not meet all the requirements of RTCWeb. Moreover, the deployment of a new transport protocol experiences problems traversing middle boxes such as Network Address Translation (NAT) box, firewall. Nevertheless, the current implementation for transportation of non-media in the very first versions of RTCWeb data does not include any congestion control on the end-points. With media (i.e., audio, video) the amount of traffic can be determined and limited by the codec and profile used during communication, whereas RTCWeb user could generate as much as non-media data to create congestion on the networks. Therefore, a suitable transport protocol stack is required that will provide congestion control, NAT traversal solution, and authentication, integrity, and privacy of user data. This master's thesis will give emphasis on the analysis of transport protocol stack for data channel in RTCWeb and selects Stream Control Transmission Protocol (SCTP), which is a reliable, message oriented general-purpose transport layer protocol, operating on top of both IPv4 and IPv6, providing congestion control similar to TCP and additionally, some new functionalities regarding security, multihoming, multistreaming, mobility, and partial reliability. However, due to the lack of universal availability of SCTP within the OS(s), it has been decided to use the SCTP userland implementation. WebKit is an open source web browser engine for rendering web pages used by Safari, Dashboard, Mail, and many other OS X applications. In WebKit RTCWeb implementation using GStreamer multimedia framework, RTP/UDP is utilized for the communication of media data and UDP tunnelling for non-media data. Therefore, in order to allow a smooth integration of the implementation within WebKit, we have decided to implement GStreamer plugins using SCTP userland stack. This thesis work also investigates the way Mozilla has integrated those protocols in the browser's network stack and how the Data Channel has been designed and implemented using SCTP userland stack.

Cloud computing offers important resources, performance, and services nowadays when it has became popular to collect, store and analyze large data sets. This thesis builds on Berkeley Data Analysis Stack (BDAS) as a cloud computing environment designed for Big Data handling and analysis. Especially two parts of the BDAS, the cluster resource manager Mesos and the distribution manager Spark will be introduced. They offer important features, such as efficiency, multi-tenancy, and fault tolerance, for cloud computing. The Spark system expands MapReduce, the well-known cloud computing paradigm. Machine learning algorithms can predict trends and anomalies of large data sets. This thesis will present one of them, a distributed decision tree algorithm, implemented on the Spark system. As an example case, the decision tree will be used on the versatile energy consumption data from mobile devices, such as smart phones and tablets, of the Carat project. The data consists of information about the usage of the device, such as which applications have been running, network connections, battery temperatures, and screen brightness, for example. The decision tree aims to find chains of data features that might lead to energy consumption anomalies. Results of the analysis can be used to advise users on how to improve their battery life. This thesis will present selected analysis results together with advantages and disadvantages of the decision tree analysis.

Portable music players have made it possible to listen to a personal collection of music in almost every situation, and they are often used during some activity to provide a stimulating audio environment. Studies have demonstrated the effects of music on the human body and mind, indicating that selecting music according to situation can, besides making the situation more enjoyable, also make humans perform better. For example, music can boost performance during physical exercises, alleviate stress and positively affect learning. We believe that people intuitively select different types of music for different situations. Based on this hypothesis, we propose a portable music player, AndroMedia, designed to provide personalised music recommendations using the user's current context and listening habits together with other user's situational listening patterns. We have developed a prototype that consists of a central server and a PDA client. The client uses Bluetooth sensors to acquire context information and logs user interaction to infer implicit user feedback. The user interface also allows the user to give explicit feedback. Large user interface elements facilitate touch-based usage in busy environments. The prototype provides the necessary framework for using the collected information together with other user's listening history in a context- enhanced collaborative filtering algorithm to generate context-sensitive recommendations. The current implementation is limited to using traditional collaborative filtering algorithms. We outline the techniques required to create context-aware recommendations and present a survey on mobile context-aware music recommenders found in literature. As opposed to the explored systems, AndroMedia utilises other users' listening habits when suggesting tunes, and does not require any laborious set up processes.

Cellular networks are facing a data explosion posed by the increasing bandwidth demand of current mobile applications, and cellular operators are trying to leverage auxiliary networks and offload mobile data for relieving this challenge. However, traffic offloading without comprehensive controlling may result poor network utilization and undesirable user experience. In this thesis, we design and implement an integrated architecture for intelligent traffic offloading over collaborative WiFi-cellular networks. Motivated by our measurement, we formulate a mathematical model to estimate and evaluate potential offloading throughput based on various wireless context information, like AP signal strength and bandwidth. To efficiently manage traffic and collect information, we use a centralized SDN architecture in our design. The proposed system enables mobile devices to choose the most beneficial AP for offloading. The experimental evaluation of our prototype implementation demonstrates that this architecture can achieve optimal traffic offloading by considering different real factors instead of making naive decisions. This effort not only explores the feasibility of context-based traffic offloading, but also provides guidelines for designing and implementing a centralized SDN platform for wireless networks.

Hypervisors and containers are the two main virtualization techniques that enable cloud computing. Both techniques have performance overheads on CPU, memory, networking, and disk performance compared to bare metal. Unikernels have recently been proposed as an optimization for hypervisor-based virtualization to reduce performance overheads. In this thesis, we evaluate network I/O performance overheads for hypervisor-based virtualization using Kernel-based Virtual Machine (KVM) and the OSv unikernel and for container-based virtualization using Docker comparing the different configurations and optimizations. We measure the raw networking latency and throughput and CPU utilization by using the Netperf benchmarking tool and measure network intensive application performance using the Memcached key-value store and the Mutilate benchmarking tool. We show that compared to bare metal Linux, Docker with bridged networking has the least performance overhead with OSv using vhost-net coming a close second.

The way the users interact with Information Retrieval (IR) systems is an interesting topic of interest in the field of Human Computer Interaction (HCI) and IR. With the ever increasing information in the web, users are often lost in the vast information space. Navigating in the complex information space to find the required information, is often an abstruse task by users. One of the reasons is the difficulty in designing systems that would present the user with an optimal set of navigation options to support varying information needs. As a solution to the navigation problem, in this thesis we propose a method referred as interaction portfolio theory, based on Markowitz's 'Modern Portfolio theory', a theory of finance. It provides the users with N optimal interaction options in each iteration, by taking into account user's goal expressed via interaction during the task, but also the risk related to a potentially suboptimal choice made by the user. In each iteration, the proposed method learns the relevant interaction options from user behaviour interactively and optimizes relevance and diversity to allow the user to accomplish the task in a shorter interaction sequence. This theory can be applied to any IR system to help users to retrieve the required information efficiently.

This thesis considers the problem of finding a process that, given a collection of news, can detect significant dates and breaking news related to different themes. The themes are unsupervisedly learned from some training corpora, and they mostly have intuitive meanings, like 'finance', 'disaster', 'wars' and so on. They are constructed only based on textual information provided in the corpora without any human intervention. To conduct this learning, the thesis use various types of component models, specifically Latent Dirichlet Allocation(LDA) and Correlated Topic Model(CTM). On top of that, to enrich the experiment, the Latent Semantic Indexing(LSA) and Multinomial Principal Component Analysis(MPCA) are also adopted for comparison. The learning produces every news coverage a relevance weight for given theme, which can be viewed as a theme distribution from statistical perspective. With the help of news time-stamp information, one can sum up and normalize these distributions from all news in day unit, and then draw the moving of accumulated relevance weights on a theme through time-line. It is natural to treat these curves as describing attention strength paid from media to different themes, and one can assume that behind every peak, there are striking events and associated news can be detected. This thesis is valuable in Media Studies research, and also can be further connected to stock or currency market for creating real value.

Mobile RFID services for the Internet of Things can be created by using RFID as an enabling technology in mobile devices. Humans, devices, and things are the content providers and users of these services. Mobile RFID services can be either provided on mobile devices as stand-alone services or combined with end-to-end systems. When different service solution scenarios are considered, there are more than one possible architectural solution in the network, mobile, and back-end server areas. Combining the solutions wisely by applying the software architecture and engineering principles, a combined solution can be formulated for certain application specific use cases. This thesis illustrates these ideas. It also shows how generally the solutions can be used in real world use case scenarios. A case study is used to add further evidence.