BUILDING EFFICIENT AND COST-EFFECTIVE CLOUD-BASED BIG DATA MANAGEMENT SYSTEMS

In today’s big data world, data is being produced in massive volumes, at great velocity and from a variety of different sources such as mobile devices, sensors, a plethora of small devices hooked to the internet (Internet of Things), social networks, communication networks and many others. Interactive querying and large-scale analytics are being increasingly used to derive value out of this big data. A large portion of this data is being stored and processed in the Cloud due the several advantages provided by the Cloud such as scalability, elasticity, availability, low cost of ownership and the overall economies of scale. There is thus, a growing need for large-scale cloud-based data management systems that can support real-time ingest, storage and processing of large volumes of heterogeneous data. However, in the pay-as-you-go Cloud environment, the cost of analytics can grow linearly with the time and resources required. Reducing the cost of data analytics in the Cloud thus remains a primary challenge. In my dissertation research, I have focused on building efficient and cost-effective cloud-based data management systems for different application domains that are predominant in cloud computing environments. In the first part of my dissertation, I address the problem of reducing the cost of transactional workloads on relational databases to support database-as-a-service in the Cloud. The primary challenges in supporting such workloads include choosing how to partition the data across a large number of machines, minimizing the number of distributed transactions, providing high data availability, and tolerating failures gracefully. I have designed, built and evaluated SWORD, an end-to-end scalable online transaction processing system, that utilizes workload-aware data placement and replication to minimize the number of distributed transactions that incorporates a suite of novel techniques to significantly reduce the overheads incurred both during the initial placement of data, and during query execution at runtime. In the second part of my dissertation, I focus on sampling-based progressive analytics as a means to reduce the cost of data analytics in the relational domain. Sampling has been traditionally used by data scientists to get progressive answers to complex analytical tasks over large volumes of data. Typically, this involves manually extracting samples of increasing data size (progressive samples) for exploratory querying. This provides the data scientists with user control, repeatable semantics, and result provenance. However, such solutions result in tedious workflows that preclude the reuse of work across samples. On the other hand, existing approximate query processing systems report early results, but do not offer the above benefits for complex ad-hoc queries. I propose a new progressive data-parallel computation framework, NOW!, that provides support for progressive analytics over big data. In particular, NOW! enables progressive relational (SQL) query support in the Cloud using unique progress semantics that allow efficient and deterministic query processing over samples providing meaningful early results and provenance to data scientists. NOW! enables the provision of early results using significantly fewer resources thereby enabling a substantial reduction in the cost incurred during such analytics. Finally, I propose NSCALE, a system for efficient and cost-effective complex analytics on large-scale graph-structured data in the Cloud. The system is based on the key observation that a wide range of complex analysis tasks over graph data require processing and reasoning about a large number of multi-hop neighborhoods or subgraphs in the graph; examples include ego network analysis, motif counting in biological networks, finding social circles in social networks, personalized recommendations, link prediction, etc. These tasks are not well served by existing vertex-centric graph processing frameworks whose computation and execution models limit the user program to directly access the state of a single vertex, resulting in high execution overheads. Further, the lack of support for extracting the relevant portions of the graph that are of interest to an analysis task and loading it onto distributed memory leads to poor scalability. NSCALE allows users to write programs at the level of neighborhoods or subgraphs rather than at the level of vertices, and to declaratively specify the subgraphs of interest. It enables the efficient distributed execution of these neighborhood-centric complex analysis tasks over largescale graphs, while minimizing resource consumption and communication cost, thereby substantially reducing the overall cost of graph data analytics in the Cloud. The results of our extensive experimental evaluation of these prototypes with several real-world data sets and applications validate the effectiveness of our techniques which provide orders-of-magnitude reductions in the overheads of distributed data querying and analysis in the Cloud.

MICROSCALE DIELECTRIC ANTI-REFLECTION COATINGS FOR PHOTOVOLTAICS

In order to power our planet for the next century, clean energy technologies need to be developed and deployed. Photovoltaic solar cells, which convert sunlight into electricity, are a clear option; however, they currently supply 0.1% of the US electricity due to the relatively high cost per Watt of generation. Thus, our goal is to create more power from a photovoltaic device, while simultaneously reducing its price. To accomplish this goal, we are creating new high efficiency anti-reflection coatings that allow more of the incident sunlight to be converted to electricity, using simple and inexpensive coating techniques that enable reduced manufacturing costs. Traditional anti-reflection coatings (consisting of thin layers of non-absorbing materials) rely on the destructive interference of the reflected light, causing more light to enter the device and subsequently get absorbed. While these coatings are used on nearly all commercial cells, they are wavelength dependent and are deposited using expensive processes that require elevated temperatures, which increase production cost and can be detrimental to some temperature sensitive solar cell materials. We are developing two new classes of anti-reflection coatings (ARCs) based on textured dielectric materials: (i) a transparent, flexible paper technology that relies on optical scattering and reduced refractive index contrast between the air and semiconductor and (ii) silicon dioxide (SiO2) nanosphere arrays that rely on collective optical resonances. Both techniques improve solar cell absorption and ultimately yield high efficiency, low cost devices. For the transparent paper-based ARCs, we have recently shown that they improve solar cell efficiencies for all angles of incident illumination reducing the need for costly tracking of the sun’s position. For a GaAs solar cell, we achieved a 24% improvement in the power conversion efficiency using this simple coating. Because the transparent paper is made from an earth abundant material (wood pulp) using an easy, inexpensive and scalable process, this type of ARC is an excellent candidate for future solar technologies. The coatings based on arrays of dielectric nanospheres also show excellent potential for inexpensive, high efficiency solar cells. The fabrication process is based on a Meyer rod rolling technique, which can be performed at room-temperature and applied to mass production, yielding a scalable and inexpensive manufacturing process. The deposited monolayer of SiO2 nanospheres, having a diameter of 500 nm on a bare Si wafer, leads to a significant increase in light absorption and a higher expected current density based on initial simulations, on the order of 15-20%. With application on a Si solar cell containing a traditional anti-reflection coating (Si3N4 thin-film), an additional increase in the spectral current density is observed, 5% beyond what a typical commercial device would achieve. Due to the coupling between the spheres originated from Whispering Gallery Modes (WGMs) inside each nanosphere, the incident light is strongly coupled into the high-index absorbing material, leading to increased light absorption. Furthermore, the SiO2 nanospheres scatter and diffract light in such a way that both the optical and electrical properties of the device have little dependence on incident angle, eliminating the need for solar tracking. Because the layer can be made with an easy, inexpensive, and scalable process, this anti-reflection coating is also an excellent candidate for replacing conventional technologies relying on complicated and expensive processes.

INVESTIGATION OF MAGNESIUM ION CHARGE STORAGE MECHANISM IN MANGANESE DIOXIDE

Magnesium (Mg) battery is considered as a promising candidate for the next generation battery technology that could potentially replace the current lithium (Li)-ion batteries due to the following factors. Magnesium possesses a higher volumetric capacity than commercialized Li-ion battery anode materials. Additionally, the low cost and high abundance of Mg compared to Li makes Mg batteries even more attractive. Moreover, unlike metallic Li anodes which have a tendency to develop a dendritic structure on the surface upon the cycling of the battery, Mg metal is known to be free from such a hazardous phenomenon. Due to these merits of Mg as an anode, the topic of rechargea¬ble Mg batteries has attracted considerable attention among researchers in the last few decades. However, the aforementioned advantages of Mg batteries have not been fully utilized due to the serious kinetic limitation of Mg2+ diffusion process in many hosting compounds which is believed to be due to a strong electrostatic interaction between divalent Mg2+ ions and hosting matrix. This serious kinetic hindrance is directly related to the lack of cathode materials for Mg battery that provide comparable electrochemical performances to that of Li-based system. Manganese oxide (MnO2) is one of the most well studied electrode materials due to its excellent electrochemical properties, including high Li+ ion capacity and relatively high operating voltage (i.e., ~ 4 V vs. Li/Li+ for LiMn2O4 and ~ 3.2 V vs. Mg/Mg2+). However, unlike the good electrochemical properties of MnO2 realized in Li-based systems, rather poor electrochemical performances have been reported in Mg based systems, particularly with low capacity and poor cycling performances. While the origin of the observed poor performances is believed to be due to the aforementioned strong ionic interaction between the Mg2+ ions and MnO2 lattice resulting in a limited diffusion of Mg2+ ions in MnO2, very little has been explored regarding the charge storage mechanism of MnO2 with divalent Mg2+ ions. This dissertation investigates the charge storage mechanism of MnO2, focusing on the insertion behaviors of divalent Mg2+ ions and exploring the origins of the limited Mg2+ insertion behavior in MnO2. It is found that the limited Mg2+ capacity in MnO2 can be significantly improved by introducing water molecules in the Mg electrolyte system, where the water molecules effectively mitigated the kinetic hindrance of Mg2+ insertion process. The combination of nanostructured MnO2 electrode and water effect provides a synergic effect demonstrating further enhanced Mg2+ insertion capability. Furthermore, it is demonstrated in this study that pre-cycling MnO2 electrodes in water-containing electrolyte activates MnO2 electrode, after which improved Mg2+ capacity is maintained in dry Mg electrolyte. Based on a series of XPS analysis, a conversion mechanism is proposed where magnesiated MnO2 undergoes a conversion reaction to Mg(OH)2 and MnOx and Mn(OH)y species in the presence of water molecules. This conversion process is believed to be the driving force that generates the improved Mg2+ capacity in MnO2 along with the water molecule’s charge screening effect. Finally, it is discussed that upon a consecutive cycling of MnO2 in the water-containing Mg electrolyte, structural water is generated within the MnO2 lattice, which is thought to be the origin of the observed activation phenomenon. The results provided in this dissertation highlight that the divalency of Mg2+ ions result in very different electrochemical behaviors than those of the well-studied monovalent Li+ ions towards MnO2.