Browsing by Subject "Fusion"
Now showing 1 - 8 of 8
Results Per Page
Sort Options
Item Active replication vs. fusion as fault tolerance mechanisms(2016-05) Boyd, Jeremy J.; Garg, Vijay K. (Vijay Kumar), 1963-; Aziz, AdnanThis report compares two strategies for crash fault tolerance of nodes in distributed systems: active replication and fusion. To tolerate f crash faults, active replication maintains f backup servers for each primary. Fusion, however, maintains a set of f backup servers that contain the replicated data for all primaries in coded form. If n primaries each contain m data to be backed up, then, active replication requires O(nmf) space, while fusion requires only O(mf) space. These savings come at the cost of additional time during the recovery process due to additional messages and computation. For this report, we have implemented an application in which primary nodes maintain increasingly large data structures and periodically crash. Both active replication and fusion are evaluated as recovery mechanisms for the crashed nodes. The mechanisms are evaluated for performance across three metrics: backup size, time spent during updates to the backup, and recovery time. Our results validate theoretical expectations put forward in the literature – that fusion claims significant space savings at the cost of high recovery time. In the most extreme measured case, fusion costs approximately 83% of the space that replication does, while recovery time is three orders of magnitude more expensive in fusion (3.4s) than in replication (0.0037s). However, we also find that the gap between fusion and replication grows as nodes are introduced to the system. We find furthermore that fusion performs more consistently in update time due to the high variability of multicasting within active replication systems.Item Controlled assembly of biodegradable gold nanoclusters for in vivo imaging(2015-12) Stover, Robert John; Johnston, Keith P., 1955-; Truskett, Thomas M; Fan, Donglei; Korgel, Brian; Sokolov, KonstantinGold nanoparticles are of interest in biomedical imaging applications due to their inert nature and ability to exhibit surface plasmon resonance. These phenomena can result in high near-infrared extinction (NIR) due to asymmetry or close interparticle spacings within gold structures, making these materials ideal for photoacoustic imaging. Using this imaging modality, these materials allow for high contrast compared to the body’s tissues which exhibit a transparent “window” between 700-1100 nm, making them perfect for early cancer detection. However many gold structures designed for this application fail to achieve high NIR-absorbance at the <5 nm sizes which are required for efficient kidney clearance. Therefore, we designed a system which assembles ~4 nm primary gold particles into closely-spaced clusters of controlled size using a biodegradable, weakly adsorbing polymer and balance of colloidal attractive and repulsive forces. Thus, when the polymer degrades in acidic environments – such as within cells – the residual charge on the primary particles leads to dissociation of the clusters back to renal-clearable constituents. Since proteins in the blood and cells can increase the diameter of the primary particles above the 5 nm threshold, nanoparticle surfaces were designed to have a mixture of charged and zwitterionic molecules to limit protein interactions through buried charges and increased particle hydration. Strongly-bound, zwitterionic thiol-containing ligands were also investigated to resist the intracellular exchange of biomolecules which could compromise the clearable nature of the particles. These decorated nanoparticles were then assembled into clusters through one of two methods which varied either gold and polymer concentrations through evaporation, or particle charge via electrolyte addition prior to quenching by dilution in DI water. Once assembled, clusters assembled with polymer showed dissociation behavior after incubation in pH 5 acidic solutions to mimic the cellular pH environment. In other cases, sintering of the gold nanoparticle clusters prevented such dissociation. This thesis demonstrates the ability to not only create biocompatible nanoparticle surfaces, but to establish control size control over nanocluster assemblies which are capable of being used as NIR contrast agents.Item Experimental measurement of energy transport in tokamak plasmas(2010-08) Meyerson, Dmitry; Gentle, Kenneth W.; Horton, WendellA tokamak plasma near equilibrium can be perturbed with modulated power sources, such as modulated electron cyclotron heating, or repeated cold pulse application. Temperature response to cyclical changes in profiles parameters that are induced by modulated power deposition can be used to test theoretical transport models as well as improve experimental phenomenology used to optimize tokamak performance. The goal of this document to discuss some methods of analyzing electron temperature data in the context of energy transport. Specific experiments are considered in order to demonstrate the methods discussed, as well as to examine the electron energy transport properties of these shots. Electron cyclotron emission provides a convenient way to probe electron temperature for plasmas in thermal equilibrium. We can show that in tokamak devices,barring harmonic overlap, we can associate a particular frequency with a particular location in a tokamak, by carefully selecting the detection frequency and line of sight of the responsible antenna. ECE radiometers typically measure temperature at tens of locations at a time with a spatial resolution on the order of a few centimeters. Tracking the evolution of electron energy flux depends on careful analysis of the resulting data. The most straightforward way to analyze temperature perturbations is to simply consider various harmonics of the driving source and consider the corresponding harmonics in the temperature. We can analyze the phase and amplitude of the response to find the effective phase velocity of the perturbation which can in turn be related to parameters in the selected heat flux model. The most common example is to determine , the diffusion coefficient that appears in the linearized energy transport equation. The advantages and limitation of this method will be discussed in detail in Section 3. A more involved approach involves using the perturbed temperature data to compute modulated heat flux at any given point in the perturbation cycle, rather than using the temperature data directly. As before the heat flux can then be related to measured profile parameters and theoretical predictions. The advantages and limitations of this approach will be discussed in more detail. Both of the mentioned analysis methods are used to probe electron energy transport in a quiescent H mode (QH mode) shot conducted at DIIID. The nature of the internal transport barrier that is present in the shot is considered in light of the results.Item Experimental studies of high energy density silicon using ultra-fast lasers(2007) Grigsby, Will Robert, 1978-; Ditmire, Todd R.Understanding material behavior under extreme conditions is an important area of research in physics and material science. One method to study the behavior of materials under these conditions is to drive a strong shock wave through a material and watch its response. In many cases the material response is complicated by phase transitions such as lattice restructuring (Barker 1975; Mabire and Hereil 2000; Swift, Tierney et al. 2005) and melting (Asay 1975; Elias, Chapron et al. 1988; Werdiger, Eliezer et al. 1999; Mabire and Hereil 2000; Swift, Tierney et al. 2005). To study these dynamics we are using lasers in high time resolution pump-probe experiments to develop a real time diagnostic on the phase of a shocked material. This technique enables probing of the entire phase history of a material as it shock compresses and releases. In addition to linear reflectivity and ultra-fast 2D displacement interferometry, we developed a melting diagnostics based on the non-linear optical technique of third harmonic generation (THG) using a circularly polarized laser pulse. This diagnostic resolves the less than 300 fs melting transition of laser excited Si and GaAs, and it also detects a response in shock compressed silicon. Our results show that Si remains crystalline during compression of an elastic 100 kbar shock wave. Results from Si shocked to higher pressures (> 300 kbar) indicate a decrease in THG, suggesting some level of disordering or unexplained phase change.Item Fusion-based Hadoop MapReduce job for fault tolerance in distributed systems(2013-05) Ho, Iat-Kei; Garg, Vijay K. (Vijay Kumar), 1963-Standard recovery solution on a failed task in Hadoop systems is to execute the task again. After retrying for a configured number of times, it is marked as failure. With significant amount of data, complicated Map and Reduce functions, recovering corrupted or unfinished data from a failed job can be more efficient than re-executing the same job. This paper is an extension of [1] by applying fusion-based technique [7][8] in Hadoop MapReduce tasks execution to enhance its fault tolerance. Multiple data sets are executed through Hadoop MapReduce with and without fusion in various pre-defined failure scenarios for comparison. As the complexity of the Map and Reduce function relative to the Recover function increases, it becomes more efficient to utilize fusion and users can tolerate faults by incurring less than ten percent of extra execution time.Item The macro- and micro-instabilities in the pedestal region of the Tokamak(2015-05) Ma, Jingfei; Morrison, Philip J.; Horton, C. W. (Claude Wendell), 1942-; Berk, Herbert; Fitzpatrick, Richard; Hallock, GaryIn this paper, we present the theoretical and numerical studies of the linear characteristics and nonlinear transport features of the instabilities driven by the steep profile gradient and edge current in the pedestal region of the tokamak. Two important instabilities, the peeling-ballooning (P-B) modes (macro-instability) and the drift-Alfven modes (micro-instability), are studied using the fluid analysis and the BOUT++ codes. In particular, the edge-localized modes (ELMs), which appear to be the energy burst in the nonlinear stage of the peeling-ballooning mode, are numerically studied and the results are compared with the experimental measurement. In addition, the features of the impurity transport in the edge region of the tokamak are theoretically analyzed. Firstly, we explore the fundamental characteristics of the P-B modes and the ELM bursts numerically using the three-field reduced MHD model under the BOUT++ framework, in the shifted-circular geometry, i.e. the limiter tokamak geometry. In the linear simulations, the growth rate and real frequency and the mode structure versus the toroidal mode number (n) are shown. The features of the ELM bursts are shown in the nonlinear simulations, including the time evolution of the relative energy loss (ELM size) and the pedestal profile. Secondly, two original research projects related to the P-B modes and the ELM burst are described. One is the study of the scaling law between the relative energy loss of ELMs and the edge collisionality. We generate a sequence of shifted-circular equilibria with different edge collisionality varying over four orders of magnitude using EFIT. The simulation results are in good agreement with the multi-tokamak experimental data. Another is the study of the differences of the linear behaviors of the P-B modes between the standard and snowflake divertor configurations. Using DIII-D H-mode ElMing equilibria, we found that the differences are due to the local magnetic shear change at the outboard midplane, which is the result of the realization of the snowflake configuration. Finally, the micro-instability, the drift-Alfven instability in the pedestal region of the DIII-D tokamak is studied. A modified six-field Landau fluid model under BOUT++ framework is used to study the linear characteristics and transport features of the drift-Alfven modes. Based on the DIII-D H-mode discharge, a sequence of divertor tokamak equilibria with different pedestal height is generated by the ’VARYPED’ tool for our studies. Qualitative agreement is obtained between theoretical analysis and the simulation results in the linear regime. Moreover, the heat transport induced by the drift-Alfven turbulence is explored and the convection level is estimated for both ions and electrons.Item Tearing mode dynamics in tokamak plasmas(2016-05) Vergos, Nikolaos; Fitzpatrick, Richard, 1963-; Hazeltine, Richard; Breizman, Boris; Waelbroeck, Francois; Hallock, GaryOne of the most problematic instabilities in tokamak plasmas is tearing modes; they are driven by current and pressure gradients, and involve a reconfiguration of the magnetic and velocity fields localized into a narrow region located at a resonant magnetic surface. While the equilibrium magnetic field lines are located on concentric nested toroidal flux surfaces, the instability creates magnetic islands in which field lines connect flux tubes together, allowing for a high radial heat transport, and, thus, resulting in a loss of confinement, and, potentially, disruptions. In order for the magnetic field lines to break and reconnect, we need to take into account the resistivity of the plasma and solve the resistive magnetohydrodynamics (MHD) equations. The analytical solution consists of a boundary layer analysis (asymptotic matching) and takes advantage of the small radial width of the region where the perturbations vary significantly. Indeed, ideal magnetohydrodynamics can be used everywhere except in that narrow region where the full resistive problem must be solved. This dissertation addresses two related problems in the study of resistive tearing modes, and their interactions with externally induced resonant magnetic perturbations (error-fields). First, an in-depth investigation of the bifurcated states of a rotating, quasi-cylindrical, tokamak plasma in the presence of a resonant error-field is performed, within the context of constant-ψ resistive MHD theory. The response of the rotating plasma is studied in both the linear, and the nonlinear regime. In general, there is a "forbidden band" of tearing mode rotation frequencies that separates a branch of high-frequency solutions from a branch of low-frequency solutions. When a high-frequency solution crosses the upper boundary of the forbidden band there is a bifurcation to a low-frequency solution, and vice versa. Second, the analysis is extended to include the study of braking and locking of tearing mode rotation by the interaction of the mode with an error-field. It is found that this interaction can brake the plasma rotation, suppress magnetic island evolution and drive locked modes.Item Turbulent particle and thermal transport in magnetized plasmas(2013-05) Fu, Xiangrong; Morrison, Philip J.; Horton, C. W. (Claude Wendell), 1942-The particle and thermal transport by low-frequency drift waves in magnetized plasmas are studied with theories and simulations. Universal in inhomogeneous plasmas, drift waves in Earth's ionosphere, the GAMMA-10 Tandem Mirror machine, the Columbia Linear Machine and C-Mod tokamak are studied in this thesis. The first investigations are E x B particle transport in the given electric and magnetic fields of the GAMMA-10 mirror machine at the University of Tsukuba in Japan. The results show that the formation of E [subscript r]-shear by local heating of electrons can reduced the radial particle loss. The turbulent impurity particle transport driven by various modes in the MIT tokamak Alcator C-Mod is studied by a quasilinear theory and compared to experimental measurement of Boron density profiles. A code is developed for solving eigensystems of drift wave turbulence equations for the multi-component fusion plasmas and calculating quasilinear particle fluxes. The calculations are much faster than nonlinear simulations and may be suitable for real-time analysis and feedback control of tokamak plasmas. The electron temperature gradient (ETG) mode is a candidate mechanism for anomalous electron thermal transport across various magnetic confinement geometries. This mode was produced in the Columbia Linear Machine (CLM) at Columbia University. Large scale simulations of the ETG mode in the CLM by a gyrokinetic code GTC are carried out on supercomputers at TACC and NERSC. The results show good agreement with experiments in the dominant mode number, wave frequencies and the radial structure. Some nonlinear properties are also analyzed using the code.