As shared memory platforms continue to grow in core counts, the need for context-aware scheduling continues to grow. Context-aware scheduling takes into account characteristics of the system and application performance when making decisions. Traditionally applications run on a static thread count that is set at compile-time or at job start time, without taking into account the dynamic context of the system, other running applications, and potentially the performance of the application itself. However, many shared memory applications can easily be converted to malleable applications, that is, applications that can run with an arbitrary number of threads and can change thread counts during execution. Many new and intelligent scheduling decisions can be made when applications become context-aware, including expanding to ll an empty system or shrinking to accommodate a more parallelizable job. This thesis describes a prototype system called Resizing for Shared Memory (RSM), which supports OpenMP applications on shared memory platforms. RSM includes a main daemon that records performance information and makes resizing decisions as well as a communication library to allow applications to contact the RSM daemon and enact resizing decisions. Experimental results show that RSM can improve turn-around time and system utilization even using very simple heuristics and resizing policies. / Master of Science

Due to the unpredictability in job arrival times in clusters and widely varying resource requirements, dynamic scheduling of parallel computing resources is necessary to increase system throughput. Dynamically resizable applications provide the flexibility needed for dynamic scheduling. These applications can expand to take advantage of additional free processors, or to meet a Quality of Service (QoS) deadline, or can shrink to accommodate a high priority application, without getting suspended. This thesis is part of a larger effort to define a framework for dynamically resizable parallel applications. This framework includes a scheduler that supports resizing applications, an API to enable applications to interact with the scheduler, and libraries that make resizing viable. This thesis focuses on libraries for efficient resizing of parallel applications—efficient in terms of minimizing the cost of data redistribution, choosing and allocating the right set of additional processors, and focusing on the performance of the application after resizing. We explore the tradeoffs between these goals on both homogeneous and heterogeneous clusters. We focus on structured applications that have 2D data arrays distributed across a 2D processor grid. Our library includes algorithms for processor selection and processor mapping. For homogeneous clusters, processor selection involves selecting the number of processors that needs to be added and processor mapping decides the placement of the new processors in the context of the given topology such that it minimizes the amount of data that is to be redistributed. For heterogeneous clusters, since the processing powers of the processors vary, there is also an additional problem of choosing the right set of processors that needs to be added. We also present results that demonstrate the effectiveness of our approach. / Master of Science