The growing performance gap between CPUs and sustainable disk I/O is a major hurdle in supporting modern applications. As the CPUs become faster, this gap is projected to worsen, thus making it a critical problem that should be addressed with high priority. Although efficient algorithms have alleviated this problem, the mechanical nature of the disk places physical limits on the achievable speedup. On the other hand, newer technologies such as flash memory promise significant improvements in access time, power consumption, and storage density. However, the mature disk technology offers the most favorable cost per bit ratio. Since replacing standard hard disks with flash disks is prohibitively expensive, hybrid hard disks augment the standard hard disk with a small amount of flash memory. By exploiting the beneficial aspects of both technologies they aim to provide breakthrough increase in performance. Nevertheless, hybrid hard disks pose several significant design challenges. Effective and efficient algorithms to manage the flash, the disk, and interaction between them are required. To facilitate rapid and easy exploration of the design space for hybrid hard disk algorithms we present the design and implementation of a flexible and extensible simulator that models hybrid hard disks. The simulator is flexible in that it models several configurations in which the flash and the magnetic medium interact. The simulator is extensible in that it provides a simple framework to plug in several algorithms to manage the hybrid hard disk. We validate our simulator and analyze the performance of the hybrid hard disk for real workloads. / Master of Science

Galactic formation models have long sought to reproduce the observed chemical and kinematical properties of the Milky Way's stellar halo and disk. Recently it is the so-called "intermediate population", the stellar thick disk, that is driving advances in our understanding of the formation of spiral galaxies. The thick disk is kinematically more like the thin disk than the halo, for all the thick disk has a velocity dispersion twice that of the thin disk and rotates ~40 km/s more slowly. It is generally accepted that the thick disk's metallicity distribution function peaks at a lower metallicity than the thin disk but at higher metallicity than the halo. The lower bound of the thick disk is still uncertain, as many observational studies have found only a few thick disk candidate stars or clusters that are more metal poor than (Fe/H)=1. Beers et al. (2002) have so far proposed the largest sample of metal poor thick disk candidates, presenting 9 stars at (Fe/H)= -1.2 or lower and 46 more stars at (Fe/H)= -1 or lower, all of which are believed to belong to the thick disk. Beers et al. (2002) present possible thick disk stars as metal poor as (Fe/H)~ -2.5, roughly 1 dex lower than is suggested by current Galactic formation models (Brook et al., 2005). This study is a high-resolution spectroscopic follow-up of 29 of the stars Beers et al. (2002) and Chiba & Beers (2000) identify as potential metal poor members of the thick disk and an additional 40 stars from the cannonical thick disk, halo, and thin disk. None of the very metal-poor stars identified by Beers et al. (2002) can be confirmed as members of the thick disk and many are not metal poor at all. Only two stars more metal poor than (Fe/H)= 1.2 retain their thick disk membership. These two stars exhibit some of the chemical characteristics of the cannonical thick disk: high α-element abundances and a relatively low s--/r-- process element ratio. Also of interest are six stars with thin disk kinematic signatures but thick disk α-element abundances. That only a small number of metal poor thick disk stars could be confirmed in this study indicates that the thick disk is neither as populous nor as metal poor as has been proposed by Beers et al. (2002). / text

Disk drives are the bottleneck in the processing of large amounts of data used in almost all common applications. File systems attempt to reduce this by storing data sequentially on the disk drives, thereby reducing the access latencies. Although this strategy is useful when data is retrieved sequentially, the access patterns in real world workloads is not necessarily sequential and this mismatch results in storage I/O performance degradation. This thesis demonstrates that one way to improve the storage performance is to reorganize data on disk drives in the same way in which it is mostly accessed. We identify two classes of accesses: static, where access patterns do not change over the lifetime of the data and dynamic, where access patterns frequently change over short durations of time, and propose, implement and evaluate layout strategies for each of these. Our strategies are implemented in a way that they can be seamlessly integrated or removed from the system as desired. We evaluate our layout strategies for static policies using tree-structured XML data where accesses to the storage device are mostly of two kinds - parent-tochild or child-to-sibling. Our results show that for a specific class of deep-focused queries, the existing file system layout policy performs better by 5-54X. For the non-deep-focused queries, our native layout mechanism shows an improvement of 3-127X. To improve performance of the dynamic access patterns, we implement a self-optimizing storage system that performs rearranges popular block accesses on a dedicated partition based on the observed workload characteristics. Our evaluation shows an improvement of over 80% in the disk busy times over a range of workloads. These results show that applying the knowledge of data access patterns for allocation decisions can substantially improve the I/O performance.

The thesis aims at verification of using emerging Solid-State drives in disk arrays. The advent of SSD disks caused a small revolution in area of data storage, because the growth performance of hard drives has been slow compared to other PC components. But an entirely different principle of operation could mean compatibility problems between SSD and related technologies, such as RAID. This thesis aims at analyzing all the relevant technologies, mainly HDD, SSD and RAID. To achieve this objective, information from literature, articles and other appropriate sources will be used. Other objectives of this thesis are to determine how much are the SSDs suitable for use in the disk array, because low performance RAID controllers or different principles of operation could limit their efficiency. This question should be answered by submission of selected types of storage arrays to synthetic and practical tests of performance. The final goal is to use financial analysis of the test solutions as a shared file storage. Today, remote access to data is used by a wide range of job positions. Slow storage could mean inefficient use of working time and therefore unnecessary financial costs. The goal of my work is primarily to provide answers to the questions mentioned above. Currently it is very hard to find tests of more complex forms of disk arrays based on solid-state drives. This article can be also very useful for companies where fileservers are used to share user data. Based on the result of cost analysis, the company can then decide what type of storage is best for its purpose.

We present an on-disk sequence cache (ODSC), which improves disk drive performance. An ODSC uses a separate disk partition to store disk data in the order that the operating system requests it. Storing data in this order reduces the amount of seeking that the disk drive must do. As a result, the average disk access time is reduced. Reducing the disk access time improves the performance of the system, especially when booting the operating system, loading applications, and when main memory is limited. Experiments show that our ODSC speeds up application loads by as much as 413%. Our ODSC also reduces the disk access time of the Linux boot by 396%, and speeds up a Linux kernel make by 28%. We also show that an ODSC improves performance when main memory is limited.

A comparing study of planting spots and terrain impact in different types of terrain.

Context. The mechanisms governing planet formation are not fully understood. A new era of high-resolution imaging of protoplanetary disks has recently started, thanks to new instruments such as SPHERE, GPI, and ALMA. The planet formation process can now be directly studied by imaging both planetary companions embedded in disks and their e ff ect on disk morphology. Aims. We image disk features that could be potential signs of planet-disk interaction with unprecedented spatial resolution and sensitivity. Two companion candidates have been claimed in the disk around the young Herbig Ae /Be star HD 100546. Thus, this object serves as an excellent target for our investigation of the natal environment of giant planets. Methods. We exploit the power of extreme adaptive optics operating in conjunction with the new high-contrast imager SPHERE to image HD 100546 in scattered light. We obtained the first polarized light observations of this source in the visible (with resolution as fine as 2 AU) and new H and K band total intensity images that we analyzed with the p y n p o i n t package. Results. The disk shows a complex azimuthal morphology, where multiple scattering of photons most likely plays an important role. High brightness contrasts and arm-like structures are ubiquitous in the disk. A double-wing structure (partly due to angular di ff erential imaging processing) resembles a morphology newly observed in inclined disks. Given the cavity size in the visible (11 AU), the CO emission associated to the planet candidate c might arise from within the circumstellar disk. We find an extended emission in the K band at the expected location of b. The surrounding large-scale region is the brightest in scattered light. There is no sign of any disk gap associated to b.

A Thulium Germanate thin disk laser prototype is developed and its potential applications are discussed. Unfortunately, the thin disk gain material for the CW prototype was unable to lase due to thermal limitations within the disk. However, a CW output power model and a physical pump chamber module have been developed, along with the supporting Zemax models and alignment procedures so other gain materials and future improvements can be tested.

published_or_final_version / abstract / Orthopaedics and Traumatology / Master / Master of Philosophy

Application of conventional theory of transport and deposition to small particles or large colloids, on the order of 1 micron in diameter, has received surprisingly little attention in colloid science. While the favorable deposition of colloidal particles ( < 0.5 micron diameter) has repeatedly been shown to agree with the Smoluchowski-Levich approximation for a convective-diffusion process, larger particles are known to deviate from this solute-like mass transfer behavior. The rotating disk, used in the experiments performed in this work, is a model experimental system that has been employed in the past to de-convolute and quantify the mechanisms of particle transport. Experimental evidence shows that particle transport to the rotating disk deviates from the predictions of the complete three-dimensional convective-diffusion equation, including hydrodynamic and surface-surface interaction forces, in that non-uniform deposition is observed over the surface of the disk. Fluid inertial effects, observed to be significant in capillary flow, have been suggested in the literature as an explanation of non-uniform deposition on the rotating disk. Calculations performed in this work show that while inertial lift forces are significant, they are not the dominant cause of non-uniform deposition. Instead, hydrodynamic blocking of available deposition surface area is shown to accurately describe experimental deposition profiles. The effect of particle size on surface area exclusion and hydrodynamic scattering are separately assessed to demonstrate that the blocking model is not only phenomenologically accurate, but also an important part of the mechanistic description of transport in the rotating disk system.