Multi-agent LLM systems increasingly integrate retrieval, planning, and reasoning, but remain fundamentally text-centric, requiring agents to repeatedly recompute shared context through expensive prefill. Although single-request inference is known to be accelerated by KV-cache management, it is usually restricted to local serving scopes. We introduce AAFLOW+, a stateful extension of agentic workflow operators that makes KV cache a first-class distributed systems object. AAFLOW+ builds processes into communication-aware graphs that concurrently optimize data, prompts, and reusable model state. It also provides operators for KV materialization, transfer, fork, composition, and eviction. Its runtime enables zero-copy, transfer-aware execution, allowing agents to reuse long context without recomputation. AAFLOW+ reduces TTFT by up to 50.2x, achieves up to 7.63x reduced multi-agent compute cost at 16-agent scale, reduces KV memory by 1.72-6.10x, and increases throughput by more than 7.74x, based on an analytical cost model parameterized by empirical hardware microbenchmarks. The results demonstrate that KV transmission outperforms recomputation on networks with moderate to high bandwidth, making sure KV-state sharing greatly increases efficiency in multi-agent LLM systems by replacing text passing.

Agentic workflows in large language model systems integrate retrieval, reasoning, and memory, but existing frameworks suffer from scalability and reproducibility limitations due to fragmented data orchestration, serialization overhead, and non-deterministic execution. Although these frameworks increase flexibility, they don't have a formal execution model that adheres to the principles of high-performance computing. We introduce AAFLOW, a unified distributed runtime that creates communication-efficient execution plans by modeling agentic workflows as an operator abstraction. Using Apache Arrow and Cylon, AAFLOW creates a zero-copy data plane that allows direct interoperability between preprocessing, embedding, and vector retrieval without the need for serialization overhead. To lower coordination costs, it uses resource-deterministic scheduling and asynchronous batching. While retaining comparable LLM generation throughput, experimental results demonstrate up to 4.64 times pipeline speedup and 2.8 times gains in embedding and upsert phases. Rather than LLM inference acceleration, these advantages result from enhanced data flow, batching, and communication efficiency.

Significant obstacles exist in scientific domains including genetics, climate modeling, and astronomy due to the management, preprocess, and training on complicated data for deep learning. Even while several large-scale solutions offer distributed execution environments, open-source alternatives that integrate scalable runtime tools, deep learning and data frameworks on high-performance computing platforms remain crucial for accessibility and flexibility. In this paper, we introduce Deep Radical-Cylon(RC), a heterogeneous runtime system that combines data engineering, deep learning frameworks, and workflow engines across several HPC environments, including cloud and supercomputing infrastructures. Deep RC supports heterogeneous systems with accelerators, allows the usage of communication libraries like MPI, GLOO and NCCL across multi-node setups, and facilitates parallel and distributed deep learning pipelines by utilizing Radical Pilot as a task execution framework. By attaining an end-to-end pipeline including preprocessing, model training, and postprocessing with 11 neural forecasting models (PyTorch) and hydrology models (TensorFlow) under identical resource conditions, the system reduces 3.28 and 75.9 seconds, respectively. The design of Deep RC guarantees the smooth integration of scalable data frameworks, such as Cylon, with deep learning processes, exhibiting strong performance on cloud platforms and scientific HPC systems. By offering a flexible, high-performance solution for resource-intensive applications, this method closes the gap between data preprocessing, model training, and postprocessing.

Data is found everywhere, from health and human infrastructure to the surge of sensors and the proliferation of internet-connected devices. To meet this challenge, the data engineering field has expanded significantly in recent years in both research and industry. Traditionally, data engineering, Machine Learning, and AI workloads have been run on large clusters within data center environments, requiring substantial investment in hardware and maintenance. With the rise of the public cloud, it is now possible to run large applications across nodes without owning or maintaining hardware. Serverless functions such as AWS Lambda provide horizontal scaling and precise billing without the hassle of managing traditional cloud infrastructure. However, when processing large datasets, users often rely on external storage options that are significantly slower than direct communication typical of HPC clusters. We introduce Cylon, a high-performance distributed data frame solution that has shown promising results for data processing using Python. We describe how we took inspiration from the FMI library and designed a serverless communicator to tackle communication and performance issues associated with serverless functions. With our design, we demonstrate that the scaling efficiency of AWS Lambda achieves within 6.5% of serverful AWS (EC2) at 64 nodes, based on implementing direct communication via NAT Traversal TCP Hole Punching.

We study exascale parallel algorithms for the selection of intervention or monitoring strategies in massive realistic socio-technical networks through scalable Influence Maximization (InfMax) algorithms. We employ novel techniques to enable efficient scaling on up to 8k nodes of OLCF Frontier, with 65k AMD GPUs and 458k AMD CPU cores. Current state-of-the-art InfMax tools are limited to networks with only a few million actors (vertices) and a few hundred million interactions (edges). By overcoming these limitations,

Scientific machine learning research spans diverse domains and data modalities, yet existing benchmark efforts remain siloed and lack standardization. This makes novel and transformative applications of machine learning to critical scientific use-cases more fragmented and less clear in pathways to impact. This paper introduces an ontology for scientific benchmarking developed through a unified, community-driven effort that extends the MLCommons ecosystem to cover physics, chemistry, materials science, biology, climate science, and more. Building on prior initiatives such as XAI-BENCH, FastML Science Benchmarks, PDEBench, and the SciMLBench framework, our effort consolidates a large set of disparate benchmarks and frameworks into a single taxonomy of scientific, application, and system-level benchmarks. New benchmarks can be added through an open submission workflow coordinated by the MLCommons Science Working Group and evaluated against a six-category rating rubric that promotes and identifies high-quality benchmarks, enabling stakeholders to select benchmarks that meet their specific needs. The architecture is extensible, supporting future scientific and AI/ML motifs, and we discuss methods for identifying emerging computing patterns for unique scientific workloads. The MLCommons Science Benchmarks Ontology provides a standardized, scalable foundation for reproducible, cross-domain benchmarking in scientific machine learning. A companion webpage for this work has also been developed as the effort evolves: https://mlcommons-science.github.io/benchmark/

A key hurdle is demonstrating compute resource capability with limited benchmarks. We propose workflow templates as a solution, offering adaptable designs for specific scientific applications. Our paper identifies common usage patterns for these templates, drawn from decades of HPC experience, including recent work with the MLCommons Science working group. We found that focusing on simple experiment management tools within the broader computational workflow improves adaptability, especially in education. This concept, which we term benchmark carpentry, is validated by two independent tools: Cloudmesh's Experiment Executor and Hewlett Packard Enterprise's SmartSim. Both frameworks, with significant functional overlap, have been tested across various scientific applications, including conduction cloudmask, earthquake prediction, simulation-AI/ML interactions, and the development of computational fluid dynamics surrogates.

Managing and preparing complex data for deep learning, a prevalent approach in large-scale data science can be challenging. Data transfer for model training also presents difficulties, impacting scientific fields like genomics, climate modeling, and astronomy. A large-scale solution like Google Pathways with a distributed execution environment for deep learning models exists but is proprietary. Integrating existing open-source, scalable runtime tools and data frameworks on high-performance computing (HPC) platforms is crucial to address these challenges. Our objective is to establish a smooth and unified method of combining data engineering and deep learning frameworks with diverse execution capabilities that can be deployed on various high-performance computing platforms, including cloud and supercomputers. We aim to support heterogeneous systems with accelerators, where Cylon and other data engineering and deep learning frameworks can utilize heterogeneous execution. To achieve this, we propose Radical-Cylon, a heterogeneous runtime system with a parallel and distributed data framework to execute Cylon as a task of Radical Pilot. We thoroughly explain Radical-Cylon's design and development and the execution process of Cylon tasks using Radical Pilot. This approach enables the use of heterogeneous MPI-communicators across multiple nodes. Radical-Cylon achieves better performance than Bare-Metal Cylon with minimal and constant overhead. Radical-Cylon achieves (4 15) performing similar join and sort operations with 35 million and 3.5 billion rows with the same resources. The approach aims to excel in both scientific and engineering research HPC systems while demonstrating robust performance on cloud infrastructures. This dual capability fosters collaboration and innovation within the open-source scientific research community.

In this paper, we report on work performed for the MLCommons Science Working Group on the cloud masking benchmark. MLCommons is a consortium that develops and maintains several scientific benchmarks that aim to benefit developments in AI. The benchmarks are conducted on the High Performance Computing (HPC) Clusters of New York University and University of Virginia, as well as a commodity desktop. We provide a description of the cloud masking benchmark, as well as a summary of our submission to MLCommons on the benchmark experiment we conducted. It includes a modification to the reference implementation of the cloud masking benchmark enabling early stopping. This benchmark is executed on the NYU HPC through a custom batch script that runs the various experiments through the batch queuing system while allowing for variation on the number of epochs trained. Our submission includes the modified code, a custom batch script to modify epochs, documentation, and the benchmark results. We report the highest accuracy (scientific metric) and the average time taken (performance metric) for training and inference that was achieved on NYU HPC Greene. We also provide a comparison of the compute capabilities between different systems by running the benchmark for one epoch. Our submission can be found in a Globus repository that is accessible to MLCommons Science Working Group.

MLCommons is an effort to develop and improve the artificial intelligence (AI) ecosystem through benchmarks, public data sets, and research. It consists of members from start-ups, leading companies, academics, and non-profits from around the world. The goal is to make machine learning better for everyone. In order to increase participation by others, educational institutions provide valuable opportunities for engagement. In this article, we identify numerous insights obtained from different viewpoints as part of efforts to utilize high-performance computing (HPC) big data systems in existing education while developing and conducting science benchmarks for earthquake prediction. As this activity was conducted across multiple educational efforts, we project if and how it is possible to make such efforts available on a wider scale. This includes the integration of sophisticated benchmarks into courses and research activities at universities, exposing the students and researchers to topics that are otherwise typically not sufficiently covered in current course curricula as we witnessed from our practical experience across multiple organizations. As such, we have outlined the many lessons we learned throughout these efforts, culminating in the need for benchmark carpentry for scientists using advanced computational resources. The article also presents the analysis of an earthquake prediction code benchmark while focusing on the accuracy of the results and not only on the runtime; notedly, this benchmark was created as a result of our lessons learned. Energy traces were produced throughout these benchmarks, which are vital to analyzing the power expenditure within HPC environments. Additionally, one of the insights is that in the short time of the project with limited student availability, the activity was only possible by utilizing a benchmark runtime pipeline while developing and using software to generate jobs from the permutation of hyperparameters automatically. It integrates a templated job management framework for executing tasks and experiments based on hyperparameters while leveraging hybrid compute resources available at different institutions. The software is part of a collection called cloudmesh with its newly developed components, cloudmesh-ee (experiment executor) and cloudmesh-cc (compute coordinator).

With machine learning (ML) becoming a transformative tool for science, the scientific community needs a clear catalogue of ML techniques, and their relative benefits on various scientific problems, if they were to make significant advances in science using AI. Although this comes under the purview of benchmarking, conventional benchmarking initiatives are focused on performance, and as such, science, often becomes a secondary criteria. In this paper, we describe a community effort from a working group, namely, MLCommons Science Working Group, in developing science-specific AI benchmarking for the international scientific community. Since the inception of the working group in 2020, the group has worked very collaboratively with a number of national laboratories, academic institutions and industries, across the world, and has developed four science-specific AI benchmarks. We will describe the overall process, the resulting benchmarks along with some initial results. We foresee that this initiative is likely to be very transformative for the AI for Science, and for performance-focused communities.

Over the last several years, the computation landscape for conducting data analytics has completely changed. While in the past, a lot of the activities have been undertaken in isolation by companies, and research institutions, today's infrastructure constitutes a wealth of services offered by a variety of providers that offer opportunities for reuse, and interactions while leveraging service collaboration, and service cooperation. This document focuses on expanding analytics services to develop a framework for reusable hybrid multi-service data analytics. It includes (a) a short technology review that explicitly targets the intersection of hybrid multi-provider analytics services, (b) a small motivation based on use cases we looked at, (c) enhancing the concepts of services to showcase how hybrid, as well as multi-provider services can be integrated and reused via the proposed framework, (d) address analytics service composition, and (e) integrate container technologies to achieve state-of-the-art analytics service deployment

The Data Science domain has expanded monumentally in both research and industry communities during the past decade, predominantly owing to the Big Data revolution. Artificial Intelligence (AI) and Machine Learning (ML) are bringing more complexities to data engineering applications, which are now integrated into data processing pipelines to process terabytes of data. Typically, a significant amount of time is spent on data preprocessing in these pipelines, and hence improving its efficiency directly impacts the overall pipeline performance. The community has recently embraced the concept of Dataframes as the de-facto data structure for data representation and manipulation. However, the most widely used serial Dataframes today (R, pandas) experience performance limitations while working on even moderately large data sets. We believe that there is plenty of room for improvement by taking a look at this problem from a high-performance computing point of view. In a prior publication, we presented a set of parallel processing patterns for distributed dataframe operators and the reference runtime implementation, Cylon [1]. In this paper, we are expanding on the initial concept by introducing a cost model for evaluating the said patterns. Furthermore, we evaluate the performance of Cylon on the ORNL Summit supercomputer.

Cloud masking is a crucial task that is well-motivated for meteorology and its applications in environmental and atmospheric sciences. Its goal is, given satellite images, to accurately generate cloud masks that identify each pixel in image to contain either cloud or clear sky. In this paper, we summarize some of the ongoing research activities in cloud masking, with a focus on the research and benchmark currently conducted in MLCommons Science Working Group. This overview is produced with the hope that others will have an easier time getting started and collaborate on the activities related to MLCommons Cloud Mask Benchmark.

Digitization is changing our world, creating innovative finance channels and emerging technology such as cryptocurrencies, which are applications of blockchain technology. However, cryptocurrency price volatility is one of this technology’s main trade-offs. In this paper, we explore a time series analysis using deep learning to study the volatility and to understand this behavior. We apply a long short-term memory model to learn the patterns within cryptocurrency close prices and to predict future prices. The proposed model learns from the close values. The performance of this model is evaluated using the root-mean-squared error and by comparing it to an ARIMA model.

In this paper we apply neural networks and Artificial Intelligence (AI) to historical records of high-risk cryptocurrency coins to train a prediction model that guesses their price. This paper's code contains Jupyter notebooks, one of which outputs a timeseries graph of any cryptocurrency price once a CSV file of the historical data is inputted into the program. Another Jupyter notebook trains an LSTM, or a long short-term memory model, to predict a cryptocurrency's closing price. The LSTM is fed the close price, which is the price that the currency has at the end of the day, so it can learn from those values. The notebook creates two sets: a training set and a test set to assess the accuracy of the results. The data is then normalized using manual min-max scaling so that the model does not experience any bias; this also enhances the performance of the model. Then, the model is trained using three layers -- an LSTM, dropout, and dense layer-minimizing the loss through 50 epochs of training; from this training, a recurrent neural network (RNN) is produced and fitted to the training set. Additionally, a graph of the loss over each epoch is produced, with the loss minimizing over time. Finally, the notebook plots a line graph of the actual currency price in red and the predicted price in blue. The process is then repeated for several more cryptocurrencies to compare prediction models. The parameters for the LSTM, such as number of epochs and batch size, are tweaked to try and minimize the root mean square error.

Data-intensive applications impact many domains, and their steadily increasing size and complexity demands high-performance, highly usable environments. We integrate a set of ideas developed in various data science and data engineering frameworks. They employ a set of operators on specific data abstractions that include vectors, matrices, tensors, graphs, and tables. Our key concepts are inspired from systems like MPI, HPF (High-Performance Fortran), NumPy, Pandas, Spark, Modin, PyTorch, TensorFlow, RAPIDS(NVIDIA), and OneAPI (Intel). Further, it is crucial to support different languages in everyday use in the Big Data arena, including Python, R, C++, and Java. We note the importance of Apache Arrow and Parquet for enabling language agnostic high performance and interoperability. In this paper, we propose High-Performance Tensors, Matrices and Tables (HPTMT), an operator-based architecture for data-intensive applications, and identify the fundamental principles needed for performance and usability success. We illustrate these principles by a discussion of examples using our software environments, Cylon and Twister2 that embody HPTMT.

The Comet petascale system is an XSEDE resource with the goal of serving a large user community. The Comet project has served a large number of users while using traditional supercomputing as well as science gateways. In addition to these offerings, Comet also includes a non traditional virtual machine framework that allows users to access entire Virtual Clusters instead of just focusing on individual virtual machines. The virtual machine framework integrates a custom administration interface, a novel virtual machine image management back-end, industry standard hardware virtualization technology and leverages the Comet resource manager and job scheduler to provide access to Comet compute nodes. However, to access and manage it user input is required. In this paper, we summarize the efforts of human in the loop-for-cloud require as part of the computing activities on comet. This includes a discussion of how to get access, how to use the system, how to obtain support and what lessons we learned form the operation of this facility for users.

We use the bibliometrics approach to evaluate the scientific impact of XSEDE. By utilizing publication data from various sources, e.g., ISI Web of Science and Microsoft Academic Graph, we calculate the impact metrics of XSEDE publications and show how they compare with non-XSEDE publication from the same field of study, or non-XSEDE peers from the same journal issue. We explain the dataset and data soruces involved and how we retrieved, cleaned, and curated millions of related publication entries. We then introduce the metrics we used for evaluation and comparison, and the methods used to calculate them. Detailed analysis results of Field Weighted Citation Impact (FWCI) and the peers comparison will be presented and discussed. We also explain how the same approaches could be used to evaluate publications from a similar organization or institute, to demonstrate the general applicability of the present evaluation approach providing impact even beyond XSEDE.

Project CH-818664, KVM: Using Python, Ansible, Bash Shell, and Cloudmesh Client a fully automated process is created for deploying a configurable MongoDB sharded cluster on Chameleon, FutureSystems, and Jetstream cloud computing environments. A user runs a single Python program which configures and deploys the environment based on parameters specified for numbers of Config Server Replicas, Mongos Instances, Shards, and Shard Replication. The process installs either MongoDB version 3.4 or 3.2 as requested by the user. Additionally, functionality exists to run benchmarking tests for each deployment, capturing statistics in a file as input for python visualization programs, the results of which are displayed in this report. These reports depict the impact of MongoDB version and degrees of sharding and replication on performance. Key performance findings regarding version, sharding, and replication are abstracted from this analysis. As background, technologies and concepts key to the deployment and benchmarking, such as MongoDB, Python, Ansible, Cloudmesh Client, and Openstack are examined while comparing and using them within different clouds.