We present VizGenie, a self-improving, agentic framework that advances scientific visualization through large language model (LLM) by orchestrating of a collection of domain-specific and dynamically generated modules. Users initially access core functionalities-such as threshold-based filtering, slice extraction, and statistical analysis-through pre-existing tools. For tasks beyond this baseline, VizGenie autonomously employs LLMs to generate new visualization scripts (e.g., VTK Python code), expanding its capabilities on-demand. Each generated script undergoes automated backend validation and is seamlessly integrated upon successful testing, continuously enhancing the system's adaptability and robustness. A distinctive feature of VizGenie is its intuitive natural language interface, allowing users to issue high-level feature-based queries (e.g., "visualize the skull" or "highlight tissue boundaries"). The system leverages image-based analysis and visual question answering (VQA) via fine-tuned vision models to interpret these queries precisely, bridging domain expertise and technical implementation. Additionally, users can interactively query generated visualizations through VQA, facilitating deeper exploration. Reliability and reproducibility are further strengthened by Retrieval-Augmented Generation (RAG), providing context-driven responses while maintaining comprehensive provenance records. Evaluations on complex volumetric datasets demonstrate significant reductions in cognitive overhead for iterative visualization tasks. By integrating curated domain-specific tools with LLM-driven flexibility, VizGenie not only accelerates insight generation but also establishes a sustainable, continuously evolving visualization practice. The resulting platform dynamically learns from user interactions, consistently enhancing support for feature-centric exploration and reproducible research in scientific visualization.

Large high-performance computing systems are commonly shared among users that submit their workflows to a resource manager and scheduling framework such as SLURM. Most commonly available job schedulers provide built-in algorithms for performing job backfill and placement, where candidate jobs can be run out of order on currently free resources, provided that they do not negatively impact other jobs already waiting in the queue. Backfilling relies on two key requirements: 1) the user's own estimate of the runtime of their job and 2) the ability for the scheduler to create and maintain a future schedule of all jobs in the queue at any one moment. Unfortunately, user-provided estimates are often erroneous, a well-known problem in parallel job scheduling. These estimates cause the scheduler to plan jobs based on inaccurate data, which in turn causes the scheduler-provided estimates of user wait time to be quite inaccurate. As such, in this work, we leverage several machine learning (ML) techniques to provide a more accurate estimate of user waiting time and contrast them across a variety of metrics including wait time and bounded per-processor slowdown using simulated data based on real job workload traces. The presented machine learning models improve overall wait time estimation by a factor of$4.1 \times$over traditional scheduler-provided wait times.

A recent trend in machine learning (ML) is its application to surrogate modeling for computationally intractable simulations. The general approach is to train an ML model using data obtained from the field or extensive results from prior simulations. Although training can be an expensive process, it is typically a one-time requirement and the result can be used to reduce the time otherwise required to calculate the solution. In this work, we train and apply a sophisticated ML model that has both spatial and temporal capabilities to aid in solving a well-known two-dimensional heat transfer finite difference problem which serves as a proxy application for our external funding sponsor, Los Alamos National Laboratory (LANL). Additionally, we train and apply the same model to a 2D boiling water simulation in order to contrast the model's generality in two applications with different levels of mathematical complexity and present the obtained predictive results.

OBJECTIVE: This work investigates how reinforcement learning and deep learning models can facilitate the near-optimal redistribution of medical equipment in order to bolster public health responses to future crises similar to the COVID-19 pandemic. MATERIALS AND METHODS: The system presented is simulated with disease impact statistics from the Institute of Health Metrics, Centers for Disease Control and Prevention, and Census Bureau. We present a robust pipeline for data preprocessing, future demand inference, and a redistribution algorithm that can be adopted across broad scales and applications. RESULTS: The reinforcement learning redistribution algorithm demonstrates performance optimality ranging from 93% to 95%. Performance improves consistently with the number of random states participating in exchange, demonstrating average shortage reductions of 78.74 ± 30.8% in simulations with 5 states to 93.50 ± 0.003% with 50 states. CONCLUSIONS: These findings bolster confidence that reinforcement learning techniques can reliably guide resource allocation for future public health emergencies.

In this paper, we present a resilience analysis of the impact of soft errors on CLAMR, a hydrodynamics miniapp for high performance computing (HPC). Leveraging the conservation of mass law, we design a fault detection mechanism and checkpoint/restart fault tolerance approach to enhance the resilience of CLAMR. Overall, our approach can detect up to 88.3% of faults that propagate into SDC or crashes with minimal (less than 1%) overhead for the optimal configuration. We show that CLAMR's fault-tolerance depends on when a fault is injected into the simulation and we also evaluate the frequency of detection and checkpointing on performance.

This paper describes the adaptation of several common computer science courses to include the use of FPGA-based systems for project and assignment work. One of the principal challenges in modern curriculum design is balancing between breadth and depth of course topics while simultaneously reinforcing the interconnections among topics in the field. To address these challenges, faculty are often forced to approximate real-world systems, sometimes at the expense of any hands-on experience. At Coastal Carolina, we have integrated FPGA systems first used in our hardware-based courses into several higher-level systems and applications courses. This allows us to leverage student familiarity with a hands-on, flexible hardware platform and also strengthens the relationships between different subfields within computer science. We discuss both the changes made to our curriculum and the effectiveness of our approach.

One of the challenges in modern curriculum design is balancing between breadth and depth of topics while simultaneously reinforcing the interconnections among topics in the field. To address these challenges, faculty are often forced to approximate real-world systems, sometimes at the expense of any hands-on experience. We have integrated field-programming gate array (FPGA) systems first used in our hardware-based courses into several higher-level systems and applications courses. This allows us to leverage student familiarity with a hands-on, flexible hardware platform and also strengthen the relationships between different subfields within computer science. In this workshop, we present participants with guided hands-on activities for making use of FPGAs in common computer science courses such as digital logic design, computer organization and programming, and compiler construction. Laptop required.

The field of computer science is continuously growing, requiring curriculum developers to make numerous tradeoffs between depth and breadth with every advance. Faculty must look for new ways to communicate more concepts within a fixed number of contact hours as well as employ techniques that underscore interrelated concepts over multiple courses. With this poster, we present the use of field-programmable gate arrays (FPGAs) as a flexible hardware system that may be modified to suit the needs of a wide variety of classes. While FPGA devices are a natural fit for teaching hardware concepts in organization and architecture classes, we feel that they can also be used in a variety of roles within a number of other courses. We describe our work integrating FPGA curriculum modules into several courses in the CS curriculum: organization and architecture, compiler construction, robotics, operating systems, security, and image processing. We believe that by using these devices in several courses, students gain greater familiarity with a hands-on, reconfigurable hardware platform and can leverage this experience throughout their coursework. This approach is consonant with recent trends that emphasize the importance of students being able to reason about the end-to-end operation of an entire system and helps strengthen the relationships between different subfields within computer science. We describe both the changes made to our curriculum and the effectiveness of our approach.

As computational clusters rapidly grow in both size and complexity, system reliability and, in particular, application resilience have become increasingly important factors to consider in maintaining efficiency and providing improved computational performance over predecessor systems. One commonly used mechanism for providing application fault tolerance in parallel systems is the use of checkpointing.

Abstract Multi‐site parallel job schedulers can improve average job turn‐around time by making use of fragmented node resources available throughout the grid. By mapping jobs across potentially many clusters, jobs that would otherwise wait in the queue for local resources can begin execution much earlier; thereby improving system utilization and reducing average queue waiting time. Recent research in this area of scheduling leverages user‐provided estimates of job communication characteristics to more effectively partition the job across system resources. In this paper, we address the impact of inaccuracies in these estimates on system performance and show that multi‐site scheduling techniques benefit from these estimates, even in the presence of considerable inaccuracy. While these results are encouraging, there are instances where these errors result in poor job scheduling decisions that cause network over‐subscription. This situation can lead to significantly degraded application performance and turnaround time. Consequently, we explore the use of job checkpointing, termination, migration, and restart (CTMR) to selectively stop offending jobs to alleviate network congestion and subsequently restart them when (and where) sufficient network resources are available. We then characterize the conditions and the extent to which the process of CTMR improves overall performance. We demonstrate that this technique is beneficial even when the overhead of doing so is costly. Copyright © 2009 John Wiley & Sons, Ltd.

While measures such as raw compute performance and system capacity continue to be important factors for evaluating cluster performance, such issues as system reliability and application resilience have become increasingly important as cluster sizes rapidly grow. Although efforts to directly improve fault-tolerance are important, it is also essential to accept that application failures will inevitably occur and to ensure that progress is made despite these failures. Application monitoring frameworks are central to providing application resilience. As such, the central theme of this paper is to address the impact that application monitoring detection latency has on the overall system performance. We find that immediate fault detection is not necessary in order to obtain substantial improvement in performance. This conclusion is significant because it implies that less complex, highly portable, and predominately less expensive failure detection schemes would provide adequate application resilience.

The interaction of simultaneously co-allocated jobs can often create contention in the network infrastructure of a dedicated computational grid. This contention can lead to degraded job run-time performance. We present several bandwidth-aware co-allocating meta-schedulers. These schedulers take into account inter-cluster network utilization as a means by which to mitigate this impact. We make use of a bandwidth-centric parallel job communication model that captures the time-varying utilization of shared inter-cluster network resources. By doing so, we are able to evaluate the performance of grid scheduling algorithms that focus not only on node resource allocation, but also on shared inter-cluster network bandwidth.

Summary form only given. We present a bandwidth-centric parallel job communication model that takes into account inter-cluster network utilization as a means by which to capture the interaction and impact of simultaneously co-allocated jobs in a mini-grid. Our model captures the time-varying utilization of shared inter-cluster network resources in the grid. We compare our dynamic model with previous research that utilizes a fixed execution time penalty for co-allocated jobs. We have found that the fixed penalty model is more generous in its prediction of job turnaround time than our dynamic communication model. Additionally, we see that the penalty co-allocated jobs may experience without causing a severe performance degradation decreases as the number of clusters increases.