Search results

(1 - 2 of 2)
IMPROVING FAULT TOLERANCE FOR EXTREME SCALE SYSTEMS
Title
IMPROVING FAULT TOLERANCE FOR EXTREME SCALE SYSTEMS
Creator
Berrocal, Eduardo
Date
2017, 2017-05
Description
Mean Time Between Failures (MTBF), now calculated in days or hours, is expected to drop to minutes on exascale machines. In this thesis, a new...
Show moreMean Time Between Failures (MTBF), now calculated in days or hours, is expected to drop to minutes on exascale machines. In this thesis, a new approach for failure prediction based on the Void Search (VS) algorithm is presented . VS is used primarily in astrophysics for nding areas of space that have a very low den- sity of galaxies. We explore its potential for failure prediction using environmental information and compare it to well known prediction methods. Another important issue for the HPC community is that next-generation supercomputers are expected to have more components and consume several times less energy per operation. Hence, supercomputer designers are pushing the limits of miniaturization and energy-saving strategies. Consequently, the number of soft errors is expected to increase dramati- cally in the coming years. While mechanisms are in place to correct or at least detect soft errors, a percentage of those errors pass unnoticed by the hardware. Techniques that leverage certain properties of iterative HPC applications (such as the smoothness of the evolution of a particular dataset) can be used to detect silent errors at the application level. Results show that it is possible to detect a large number of corruptions (i.e., above 90% in some cases) with less than 100% overhead using these techniques. Nevertheless, these data-analytic solutions are still far from fully pro- tecting applications to a level comparable with more expensive solutions such as full replication. In this thesis, partial replication is explored to overcome this limitation. More speci cally, it has been observed that not all processes of an MPI application experience the same level of data variability at exactly the same time. Thus, one can smartly choose and replicate only those processes for which the lightweight data- analytic detectors would perform poorly. Results indicate that this new approach can protect the MPI applications analyzed with 7{70% less overhead (depending on the application) than that of full duplication with similar detection recall.
Ph.D. in Computer Science, May 2017
Show less
GROUP-LEVEL META-ANALYSES: AN EXAMINATION OF THE EFFECTS OF CHARACTERISTICS OF GROUP-LEVEL STUDIES ON THE ACCURACY OF PARAMETER ESTIMATES
Title
GROUP-LEVEL META-ANALYSES: AN EXAMINATION OF THE EFFECTS OF CHARACTERISTICS OF GROUP-LEVEL STUDIES ON THE ACCURACY OF PARAMETER ESTIMATES
Creator
Burke, Maura Irene
Date
2018, 2018-05
Description
This dissertation was an empirical investigation of how statistical artifacts and characteristics of group-level studies affect meta-analytic...
Show moreThis dissertation was an empirical investigation of how statistical artifacts and characteristics of group-level studies affect meta-analytic parameter estimates in group-level meta-analyses. Simulation procedures were employed to examine how the proportion of available group-level reliability information, the number of studies in a meta-analysis, and the type of group-level reliability estimate affect the accuracy of estimates of the mean and variance of rho when these population values are known. Archival data was used to identify known population parameter values and create group-level meta-analytic conditions commonly seen in the organizational sciences literature. This study has resulted in the following conclusions. When proportions of sample-based reliability are reduced in availability, meta-analyses are relatively accurate in estimating the magnitude of mean rho. As more studies enter meta-analyses, standard errors of mean rho are substantially reduced and confidence bands become increasingly smaller in width and this pattern of results holds regardless of the group-level reliability estimate used to individually correct correlations. Further, when meta-analyses involved the use of completely assumed values, the degree of accuracy in mirroring known population parameters was dependent on the degree to which the group-level reliability value approximates that of the population. Finally, both ICC(2) values and rCG group-based reliability estimates produced relatively accurate meta-analytic findings relative to their respective known population parameter values. Advantages and limitations to the use of each type of reliability estimate are discussed in detail in the manuscript.
Ph.D. in Psychology, May 2018
Show less