Heterogeneous distributed computing systems often must operate in an environment where system parameters are subject to uncertainty. Robustness can be defined as the degree to which a system can function correctly in the presence of parameter values different from those assumed. We present a methodology for quantifying the robustness of resource allocations in a dynamic environment where task execution times are stochastic. The methodology is evaluated through measuring the robustness of three different resource allocation heuristics within the context of a stochastic dynamic environment. A Bayesian regression model is fit to the combined results of the three heuristics to demonstrate the correlation between the stochastic robustness metric and the presented performance metric. The correlation results demonstrated the significant potential of the stochastic robustness metric to predict the relative performance of the three heuristics given a common objective function.

This paper describes the new hardware-based streamingaggregation capability added to Mellanox's Scalable Hierarchical Aggregation and Reduction Protocol in its HDR InfiniBand switches. For large messages, this capability is designed to achieve reduction bandwidths similar to those of point-to-point messages of the same size, and complements the latency-optimized low-latency aggregation reduction capabilities, aimed at small data reductions. MPI Allreduce() bandwidth measured on an HDR InfiniBand based system achieves about 95% of network bandwidth. For medium and large data reduction this also improves the reduction bandwidth by a factor of 2-5 relative to hostbased (e.g., software-based) reduction algorithms. Using this capability also increased DL-Poly and PyTorch application performance by as much as 4% and 18%, respectively. This paper describes SHARP Streaming-Aggregation hardware architecture and a set of synthetic and application benchmarks used to study this new reduction capability, and the range of data sizes for which Streaming-Aggregation performs better than the low-latency aggregation algorithm.

This investigation presents two distinct and novel approaches for the prediction of system failures occurring inOak Ridge National Laboratory's Blue Gene/P supercomputer. Each technique uses raw numeric and textual subsets of large data logs of physical system information such as fan speeds and CPU temperatures. This data is used to develop models of the system capable of sensing anomalies, or deviations from nominal behavior. Each algorithm predicted event log reported anomalies in advance of their occurrence and one algorithm did so without false positives. Both algorithms predicted an anomaly that did not appear in the event log. It was later learned that the fault missing from the log but predicted by both algorithms was confirmed to have occurred by the system administrator.

Heterogeneous parallel and distributed computing systems frequently must operate in environments where there is uncertainty in system parameters. Robustness can be defined as the degree to which a system can function correctly in the presence of parameter values different from those assumed. In such an environment, the execution time of any given task may fluctuate substantially due to factors such as the content of data to be processed. Determining a resource allocation that is robust against this uncertainty is an important area of research. In this study, we define a stochastic robustness measure to facilitate resource allocation decisions in a dynamic environment where tasks are subject to individual hard deadlines and each task requires some input data to start execution. In this environment, the tasks that cannot meet their deadlines are dropped (i.e., discarded). We define methods to determine the stochastic completion times of tasks in the presence of the task dropping. The stochastic task completion time is used in the definition of the stochastic robustness measure. Based on this stochastic robustness measure, we design novel resource allocation techniques that work in immediate and batch modes, with the goal of maximizing the number of tasks that meet their individual deadlines. We compare the performance of our technique against several well-known approaches taken from the literature and adapted to our environment. Simulation results of this study demonstrate the suitability of our new technique in a dynamic heterogeneous computing system.

No abstract