No description available.

In recent years, there has been a growing need for computer systems that are capable of handling unprecedented amounts of data. To this end, Hadoop HDFS and Hadoop YARN have become the de facto standard for meeting demanding storage requirements and for managing applications that can process this data. Although YARN is a major advancement from its predecessor MapReduce in terms of scalability and fault-tolerance, its Resource Manager component that performs resource allocation introduces a potential single point of failure and a performance bottleneck due to its centralized architecture. This thesis presents a novel architecture in which the Resource Manager runs on a distributed network of stateless commodity machines as its state is migrated to MySQL Cluster, a relational write-scalable and highly available in-memory database. By doing so, the Resource Manager becomes more scalable as it can now run on multiple nodes as well as more fault-tolerant as arbitrary node failures do not result in state loss. In this work we implemented the proposed architecture for the Resource Tracker service which performs cluster node management for the Resource Manager. Experimental results validate the correctness of our proposal, demonstrate how it scales well by utilizing stateless Resource Manager machines and evaluate its performance in terms of request throughput, system resource and database utilization.

The amount of data stored in modern data centres is growing rapidly nowadays. Large-scale distributed file systems, that maintain the massive data sets in data centres, are designed to work with commodity hardware. Due to the quality and quantity of the hardware components in such systems, failures are considered normal events and, as such, distributed file systems are designed to be highly fault-tolerant. A common approach to achieve fault tolerance is using redundancy by storing three copies of a file across different storage nodes, thereby increasing the storage requirements by a factor of three and further aggravating the storage problem. A concrete implementation of such a file system is the Hadoop Distributed File System (HDFS). This thesis explores the use of RAID-like mechanisms in order to decrease the storage requirements for big data. We designed and implemented a prototype that extends HDFS with a simple but powerful erasure coding API. Compared to existing approaches, we decided to locate the erasure-coding management logic in the HDFS NameNode, as this allows us to use internal HDFS APIs and state. Because of that, we can repair failures associated with erasurecoded files more quickly and with lower cost. We evaluate our prototype, and we also show that the use of erasure coding instead of replication can greatly decrease the storage requirements of big data without scarifying reliability and availability. Finally, we argue that our API can support a large range of custom encoding strategies, while adding the erasure coding logic to the NameNode can significantly improve the management of the encoded files.

In this work we assess four prototypes of gaze-supported input devices that aim o be more ergonomic alternatives than the mouse and touchpads for day to day omputer usage. These prototypes (nunchuck, ring, smartphone and trackball)have hree properties. First, they enable a distant interaction from the computer, letting he user have a 135-degree body-thigh sitting posture. Second, they allow one-handed nteractions, a convenient setting for people that can only use one hand due to rgonomic complications in the other one. And last, they are based on familiar nput technologies and conventions, so users can easily learn how to use them. An xperiment on pointing and drag & drop tasks was conducted to evaluate quantitative roperties of the devices and get qualitative user feedback. The smartphone ranked first n almost all the user feedback entries. The measures of speed and accuracy presented everal outliers. According to the measured results, the smartphone was the fastest ointing device and the most accurate for every task; the trackball was the fastest evice for drag & drop tasks. The nunchuck ranked last in all the measures. Based on his study, future work should be conducted on improving the design of the companion evices based on multitouch and trackball technologies.

Network latency is one of the most significant causes of poor performance in web services, as it is not uncommon for latency to take up a greater share of the total time from request to response than the actual processing time on the server. Working within the well-established REST architectural style for web services, we examine HTTP pipelining and composite representation using multipart/mixed as potential means of reducing the effects of latency on batched operations, using experimental implementations of both approaches to test their performance in different scenarios. The results indicate that performance improvements of up to 50% are possible using pipelining, and up to 80% when using composite representation, under controlled conditions. This supports the conclusion that significant improvements in performance are achievable in existing RESTful web services given a reasonable development effort.

Today we have many mobile applications that use location services. ith the increasing rate of applications per phone and the act that each phone has a limited amount of power, it becomes more nd more important that we see the battery as the limited resource it ruly is and that we should start to think about how we are spending t. Therefore this thesis will investigate how much power the location ervices actually need to work properly. In addition to the power consumption his thesis will also investigate how accurate each location ervice is so that we can make a statement about what we get for the mount of battery consumed. / Idag så finns det många mobilapplikationer som använder sig av lokaliseringstjänster. Idag när kvoten av applikationer per mobil ökar stadigt och det faktum att mobiler endast har en begränsad batterikappasitet, å blir det mer och mer viktigt att vi verkligen ser batteriet som en begränsad resurs som vi ska vara sparsamma med. därför kommer detta examensarbete att handla om hur mycket batterikapasitet om lokaliseringstjänster verkligen behöver använda för att fungera på ett bra sätt. Detta examensarbete kommer också att undersöka vilken precision varje service har så att vi kan bestämma å ett ungefär vad vi får för vilken batterikonsumtion.

The solar industry is bigger than ever and continues to grow while competition between the solar cell manufacturers is increasing. This is something that has lowered the prices of solar cells and reduces the profitability for manufacturers. This thesis aims to lower the cost of production without losing efficiency by using copper wires as conductors in CIGS thin film solar cells. Silver paste is printed on the surface on the solar cell and represents 1/5 of the cell manufacturing costs. That is why this project focuses on how to lower the amount of silver in the solar cell without losing efficiency, this is a hands-on project. The work is done by improving the solar cell with implementation of own ideas and innovations. This is made by taking advantage of the benefits of copper wires and by changing the solar cell pattern. The results showed that one can reduce the amount of silver paste by up to 70% in each solar cell by producing a new silver print with much less silver. The silver-coated copper wires gave the highest results as conductors to lead out electricity from the surface because they can be made much thinner than today's solar cells conductors and obscure the surface less. This gave efficiencies of 10-11% which is in line with the production solar cells modules on Midsummer. The system has also been proved stable. / Solcellsbranschen är större än någonsin tidigare och fortsätter växa samtidigt som konkurrensen mellan solcellstillverkare ökar, detta är något som pressar ner priserna på solceller och minskar lönsamheten för tillverkarna. Därför fokuserar detta projekt på hur man kan sänka mängden silver i solcellen utan att förlora verkningsgrad. Detta genomförs genom att arbeta som en problemlösare ihop med egna idéer. Genom att dra nytta av fördelarna med koppartrådar och skapa ett nytt solcellsmönster är detta möjligt då silver utgör 1/5 av tillverkningskostnaden. Projektet domineras av praktiskt arbete med fokus på innovation av standardsolcellen. Resultatet visade att man kan sänka silverpasta mängden med upp till 70 % i varje solcell genom att ta fram ett nytt silvertryck med mycket mindre silver ihop med koppartrådar. Det användes silver-belagda koppartrådar som ledare för att leda ut ström från ytan med fördelarna att det är tunna och blockerar därför solljuset mindre. Resultaten gav en verkningsgrad på 10-11% vilket är i nivå med produktionssolcellsmodulerna på Midsummer. Systemet bevisades även stabilt.

Large scale graphs are sometimes too big to store and process on a single machine. Instead, these graphs have to be divided into smaller parts and distributed over several machines, while minimizing the dependency between the different parts. This is known as the graph partitioning problem, which has been shown to be NP-complete. The problem is well studied, however most solutions are either not suitable for a distributed environment or unable to do balanced partitioning of graphs having weighted vertices. This thesis presents an extension to the distributed balanced graph partitioning algorithm JA-BE-JA, that solves the balanced partitioning problem for graph having weighted vertices. The extension, called wJA-BE-JA, is implemented in both the Spark framework and in Scala. The two main contributions of this report are the algorithm and a comprehensive evaluation of its performance, including a comparison with the recognized METIS graph partitioner. The evaluation shows that a random sampling policy in combination with the simulated annealing technique gives good results. It further shows that the algorithm is competitive to METIS, as the extension outperforms METIS in 17 of 20 tests.

In wirelesscommunication systems, a local oscillator (LO) aims at demodulating radio-frequency signals into baseband signals. The performance of these signals determines the quality of communications which is highly affected by the phase accuracy of local oscillators. Therefore, evaluating jitter/phase noise should be an essential part when designing wireless communication systems. Typically, LO is achieved by traditional analog PLL. These prototypes have several drawbacks including low integration, narrow bandwidth and high phase noise. With the development of digital techniques, approaches towards an All-digital Phase-Locked Loop have been forwarded against the traditional analogy type. The thesis mainly deals with the modeling and verification of an All-digital Phase-Locked Loop concerning its architecture, functionality and phase noise modeling and analysis. It starts with a comparison of current frequency synthesizers including direct analog/digital synthesis and indirect synthesis using PLL/ADPLL. The advantage and analogy of ADPLL versus traditional PLL in radio-frequency applications has been discussed. In order to gain overall understanding of ADPLL, a behavioral theory in both time and phase domain has been conducted in detail. Analysis shows that the restrictive factors of proposed ADPLL lie in TDC and DCO phase noise. It is also proved that the bandwidth and settling time of ADPLL is determined by proportional and integrating parameter of loop filter. Upon the completion of ADPLL theory analysis, a model based on simulink has been put forward. The phase noise level of TDC is specified and mode switch is implemented in order to improve the speed of ADPLL. The reason for choosing 2nd-order MASH-1-1 type ߢ modulator is briefly discussed. The phase noise of DCO is generated in time-domain using filtered Gaussian distribution and the free-running DCO achieves -20dB/dec spectrumfrom 10Hz to 500kHz. The results verified the feasibility of proposed ADPLL by achieving -50dBc/Hz in-band noise. Other results including howfractional precision, SDMclock and precision contributed to ADPLL phase noise has been presented. A tradeoff between phase noise shaping quality and settling time was evaluated. Ultimately, global parameters setup for the fulfillment of best performance is demonstrated.

Home automation systems are collections of smart devices that enable various functions within a house or building, such as light and plug control, energy monitoring, temperature metering, air conditioning and heating, etc. Usually, these devices are smart sensors, that are implemented with low power communication protocol like ZigBee and 6LoWPAN and only can be controlled from an Internet gateway. Nowadays, there are lots of home automation products on the market for customers. User can use application on smart phone to control the products they bought. The control command can go through a cloud-based server and be directed to the corresponding gateway, and finally reach to the sensor devices, which is referred to as "cloud-based mode" system. However, this single mode is not effcient under some circumstances where the Internet is not enabled or allowed. In this thesis work, a hybrid system architecture is proposed, implemented and evaluated, which include both stand-only and cloud-based mode. It offers a quick connection when user's smart phone and the sensor gateway are in the same private network. The proposed doublemode system architecture fits user's need and provides high reliability. / Hemautomationssystem är smarta enheter som möjliggör olika funktioner i ett hus eller en byggnad, såsom kontroll av ljus och uttag, energiövervakning, temperaturm ätning, luftkonditionering och uppvärmning, etc. Dessa enheter är vanligtvis smarta sensorer, implementerade med lågeffekt kommunikationsprotokoll som ZigBee och 6LoWPAN som endast kan styras via en Internet-gateway. Numera finns det era hemautomationprodukter på marknaden. Användaren kan med sin smarta telefon styra sina produkter. Kommandot går via en molnbaserad server och vidarebefordras till motsvarande gateway, kommandot når slutligen till sensoranordningarna, även kallat "molnbaserat läge". Detta "molnbaserade läge" är inte effektiv under vissa omständigheter där internetanslutning ej är tillänglig. I detta avhandlingsarbete föreslås, genomförs och utvärderas en hybrid systemarkitektur som inkluderar både stand-only och molnbaserade läget. Detta erbjuder en snabb anslutning när användarens smartphone och sensor gateway är pår samma privata nätverk. Den föreslagna systemarkitekturen passar användarens behov och ger hög tillförlitlighe