Efficient RDF, graph-based queries are becoming more pertinent based on the increased interest in data analytics and its intersection with large, unstructured but connected data. Many commercial systems have adopted distributed RDF graph systems in order to handle increasing dataset sizes and complex queries. This paper introduces a distribute graph approach to pre-processing linked data. Instead of traversing the memory graph, our system indexes pre-processed join elements that are organized in a graph structure. We analyze the Dbpedia data-set (derived from the Wikipedia cor-pus) and compare our access method to the graph traversal access approach which we also devise. Results show from our experiments that the distributed, pre-processed graph approach to accessing linked data is faster than the traversal approach over a specific range of linked queries. 
CCS CONCEPTS: Computing methodologies, Distributed algorithms 
KEYWORDS: RDF, High Performance Computing; Distributed Algorithms

(a) Graph-store algorithm Graph construction and neighbor times over 1M, .5M triple dataset.

A Distributed Graph Approach for Pre-processing Linked RDF Data Using Supercomputers

In situ analysis has been proposed as a promising solution to glean faster insights and reduce the amount of data to storage. A critical challenge here is that the reduced dataset is typically located on a subset of the nodes and needs to be written out to storage. Data coupling in multiphysics codes also exhibits a sparse data movement pattern wherein data movement occurs among a subset of nodes. We evaluate the performance of data movement for sparse data patterns on the IBM Blue Gene/Q supercomputing system &ldquo;Mira&rdquo; and identify performance bottlenecks. We propose a multipath data movement algorithm for sparse data patterns based on an adaptation of a maximum flow algorithm together with a breadth-first search that fully exploits all the underlying datapaths and I/O nodes to improve data movement. We demonstrate the efficacy of our solutions through a set of micro benchmarks and application benchmarks on Mira scaling up to 131,072 compute cores. The results show that our approach achieves up to 5X improvement in achievable throughput compared with the default mechanisms. 
Keywords: multiple paths, sparse data movement, topology-aware aggregation, data-intensive, Blue Gene/Q

Efficacy of the approaches to move the data to /dev/null on the I/O nodes.

Improving Sparse Data Movement Performance Using Multiple Paths on the Blue Gene/Q Supercomputer

Achievable networking performance of applications in a supercomputer depends on the exact combination of the communication patterns of the applications and the routing algorithms used by the supercomputer. In order to achieve the highest networking performance for the applications, the routing algorithms need to be designed optimally for those communication patterns. However, while communication patterns usually vary from application to application and even from phase to phase in an application, routing algorithms have limited variation and usually are optimized for typical communication patterns. This results in high networking performance for some communication patterns. In this paper, we present approaches for improving communication performance by using multiple paths and re-balancing load on physical links on the Blue Gene/Q supercomputer. We realize our approaches in a framework called OPTIQ and demonstrate the efficacy of our framework via a set of benchmarks. Our results show that we can achieve 43 -- 67% higher throughput on average from 91 experiments, and can achieve higher throughput than default MPI_Alltoallv used for certain communication patterns.

Improving Communication Throughput by Multipath Load Balancing on Blue Gene/Q

Achievable networking performance of applications in a supercomputer depends on the exact combination of the communication patterns of the applications and the routing algorithms used by the supercomputer. In order to achieve the highest networking performance for the applications the routing algorithms need to be designed optimally for those communication patterns. However, while communication patterns usually have a wide variation from application to application and even from phase to phase in an application, routing algorithms have a limited variation and usually are optimized for typical communication patterns. This results in high networking performance for favored communication patterns but low networking performance for others. In this paper we present approaches for improving networking performance by rebalancing load on physical links on the Blue Gene Q supercomputer. We realize our approaches in a framework called OPTIQ and demonstrate the efficacy of our framework via a set of benchmarks. Our results show that we can achieve 30% higher throughput on experiment with data and patterns from a real application. The improvement can be up to several times higher throughput than default MPI_Alltoallv used in the Blue Gene Q supercomputer for certain communication patterns. 
Keywords: multi-path data movement, BG/Q, optimization, heuristic, interconnection network, communication patterns, network load balancing

Multipath Load Balancing for M x N Communication Patterns on the Blue Gene⁄Q Supercomputer Interconnection Network

SEPTEMBER 22, 2021 - The UIC Electronic Visualization Laboratory&rsquo;s (EVL) long-time collaborator, the University of California, San Diego (UCSD), received a National Science Foundation (NSF) grant, to start October 2021, to Enhance and Sustain their CHASE-CI project. As part of EVL&rsquo;s ongoing commitment to this research effort, EVL will host GPU resources and data servers to enhance and extend UCSD&rsquo;s Pacific Research Platform (PRP) Nautilus Hypercluster through 2021-2024. 
In 2017, UCSD received an NSF award for CHASE-CI, the Cognitive Hardware And Software Ecosystem - Community Infrastructure, which initially provided a community of Computer & Information Science and Engineering researchers doing Machine Learning and Computational Media applications with a distributed cluster containing 256 Graphics Processing Units (GPUs). This cluster - named Nautilus - was originally distributed across 10 California campuses and built on PRP. The NSF-funded PRP project focused on integrating California&rsquo;s high-speed regional research & education network (for fast data transfers among campuses) with DTNs, or Data Transfer Nodes, which are specialized storage systems located at each campus edge, optimized to quickly send and receive large data flows. 
Early on, EVL was among 50 UCSD partners to connect to PRP, thanks to UIC&rsquo;s high-bandwidth connectivity (100 Gigabits per second) to national research and education networks via the StarLight International/National Communications Exchange Facility in downtown Chicago. At that time, UCSD provided EVL with the campus&rsquo;s first DTN, which has since been integrated into the Nautilus cluster, providing 192 Terabytes (TB) of data storage for Big Data applications. More recently, as part of its ongoing PRP award, UCSD is providing EVL with an additional 216 TB of data storage. 
Over the past 5 years, Nautilus has grown. PRP interoperability with national and international research and education networks creates a high-capacity data-centric &ldquo;freeway system&rdquo; now stretching across the nation and across oceans that moves Nautilus data among DTNs located at other research institutions. UCSD&rsquo;s Nautilus cluster contains hundreds of GPUs and many Terabytes of storage. While 25% of the computational hardware and data resources are provided by the UCSD CHASE-CI award, 75% is provided by other researchers and institutions who voluntarily contribute resources, building a shared platform that they jointly own and control. EVL has contributed some of its own GPU nodes to Nautilus, with plans to federate more of its GPU resources over time. However, any Computer & Information Science and Engineering researcher can get an account on Nautilus, whether or not they contribute resources. 
This new CHASE-CI Enhance and Sustain award aims to enhance and sustain Nautilus&rsquo;s user community. EVL, as a UCSD partner institution, will receive an 8 GPU Nautilus Machine Learning node, to be integrated with Nautilus, to provide the community with additional GPUs and data services. While EVL currently has substantial computing facilities to support the needs of UIC College of Engineering faculty who require GPUs and data storage, being part of the CHASE-CI Enhance and Sustain grant and integrating the Lab&rsquo;s resources with Nautilus expands faculty and student access to national (and international) scale community computational and data resources. 
NSF award #2120019 to UCSD, ENS: CHASE-CI, 2021-2024 
NSF award #1730158 to UCSD, CHASE-CI, 2017-2020 
NSF award #1541349 to UCSD, PRP, 2015-2022

EVL Continues UCSD Collaboration to Enhance and Sustain CHASE-CI

EVL at the University of Illinois Chicago has joined Chameleon! Chameleon is a large-scale, deeply reconfigurable experimental platform built to support Computer Sciences systems research. Publicly available since 2015, Chameleon supports projects ranging from systems research developing new operating systems, virtualization methods, performance variability studies, and power management research to projects in software-defined networking, artificial intelligence, and resource management. 
Chameleon supports a bare metal reconfiguration system giving users full control of the software stack including root privileges, kernel customization, and console access. Chameleon adapted OpenStack, a mainstream open source cloud technology, to provide its capabilities. 
The EVL site brings back 8 Haswell nodes retired from <a href="https://www.tacc.utexas.edu/">Texas Advanced Computing Center (TACC)</a> with 13 more nodes coming online next month, for a total of 21. The EVL site can be accessed with an existing Chameleon Account, Projects, and Allocations, either by going directly to <a href="https://chi.evl.uic.edu/">https://chi.evl.uic.edu/</a>, the experiment dropdown on <a href="https://chameleoncloud.org">https://chameleoncloud.org</a>, or the dropdown on horizon for any other sites. 
Each node is equipped with 24 cores (48 threads), 128 GB of RAM, 2x 1G network connections, and a 250GB hard drive. Hardware details and availability can be reviewed on the hardware discovery page. 
<a href="https://github.com/ChameleonCloud/chi-in-a-box">https://github.com/ChameleonCloud/chi-in-a-box</a>

EVL Joins Chameleon

The COMPaaS DLV: Composable Platform as a Service Instrument for Deep Learning & Visualization user community consists of faculty from Physics and four College of Engineering departments (Computer Science, Civil, Materials & Environmental Engineering, Mechanical and Industrial Engineering, and Electrical and Computer Engineering) - primarily run applications that are GPU-centric for compute, with significant variability in storage and networking around their data requirements. 
Computer Science applications primarily focus on security, data science, computer vision and Machine Learning (ML). Security projects explore the complexity of modern web applications and the intricacies of security mechanisms that often result in flaws that expose users to significant security and privacy threats. These projects try to develop methods and tools that enable users to understand and more effectively manage retrospective privacy in the context of modern, long-lived, online archives. Composable resources were used to develop Natural Language Processing (NLP)-based domain-specific classifiers that identified data practices stated in privacy policies. Adherence of corresponding applications were then adjusted based on this ground truth. 
Data-science applications have an intuitive framework that integrates state-of-the-art AI technologies with applications, workflows, smart visualizations and collaboration services to help users access, share, explore and analyze their data, whether local or remote, come to conclusions, and make decisions with greater speed, accuracy, comprehensiveness and confidence. One such project is developing and advancing tools that identify image data in biomedical literature to locate beneficial, targeted publications. This work involves training image classifiers, integrating classifiers into labeling pipelines, designing retrieval user interfaces, and identifying related visual representations. 
Computer vision projects include semantic segregation and 3D human pose estimation. Researchers are developing a novel network architecture, termed DependencyNet (dependency network), for semantic segmentation. They also achieved experimental results that demonstrate an effective approach for 3D human pose estimation. Over the past two years, they found COMPaaS to be consistently stable and efficient, with result output as expected, and the group&rsquo;s models achieved state-of-the-art performance on their respective benchmarks. 
ML applications include frameworks for many complex real-world reinforcement learning problems, such as the coordination of autonomous vehicles, network packet delivery, and distributed logistics. 
Civil, Materials & Environmental Engineering applications focus on simulation and modeling. Researchers run data-driven models on high-performance computers to develop an accurate and general neural network ML model that uses crystallographic data to study patterns of synthesizability. They also perform simulations of mass transport in alloys and ceramics. COMPaaS has been performing 1.5-10 times faster than comparable infrastructures they are familiar with. Additionally, they found our use of Jupyter Notebooks to be a significant asset. 
Mechanical and Industrial Engineering researchers use COMPaaS for three research projects: feature extraction in fluid flow using a Convolutional Neural Network (CNN); column height detection in metallic nanoparticles using a CNN; and, electric vehicle battery state-of-charge estimation using different ML methods. 
Electrical and Computer Engineering researchers recently started running mathematical models of ML algorithms and training language models using long-short term memory (LSTMs). 
In 2020, we began implementing a GPUoE (GPU over Ethernet) prototype. Using a GPU expansion chassis connected to compute nodes over Ethernet, we were able to compose remote GPUs into our existing composable infrastructure. Our APIs developed for composable infrastructure have been extended to support these remote GPUs. 
In 2021, we enhanced COMPaaS with a public-facing JupyterHub server providing secure web-based access to experiments, a modern PCIe fabric and a supporting expansion chassis, thereby making composable co-location available to our users. As familiarity with composable infrastructure grows, the idea of using composable co-location instead of traditional servers co-located in racks is possible. By composable co-location, we mean the integration of emerging hardware needs without changing the existing core infrastructure. Researchers can now co-locate their own accelerators or compute nodes and leverage existing COMPaaS resources natively.

COMPaaS DLV User Community

In today&rsquo;s Big Data era, data scientists require modern workflows to quickly analyze large-scale datasets using complex codes to maintain the rate of scientific progress. These scientists often rely on available campus resources or off-the-shelf computational systems for their applications. Unified infrastructure or over-provisioned servers can quickly become
bottlenecks for specific tasks, wasting time and resources. Composable infrastructure helps solve these problems by providing users with new ways to increase resource utilization. Composable infrastructure disaggregates a computer&rsquo;s components – CPU, GPU (accelerators), storage and networking - into fluid pools of resources, but typically relies upon infrastructure engineers to
architect individual machines. Infrastructure is either managed with specialized command-line utilities, user interfaces or specification files. These management models are cumbersome and difficult to incorporate into data-science workflows. We
developed a high-level software API, Composastructure, which, when integrated into modern workflows, can be used by infrastructure engineers as well as data scientists to reorganize composable resources on demand. Composastructure enables infrastructures to be programmable, secure, persistent and reproducible. Our API composes machines, frees resources, supports multi-rack operations, and includes a Python module for Jupyter Notebooks. 
Keywords - distributed systems, testbed implementation and deployment, composable infrastructure, deep learning, visualization, infrastructure as code

Composastructure System Architecture

Moving from Composable to Programmable

Composable infrastructure holds the promise of accelerating the pace of academic research and discovery by enabling researchers to tailor the resources of a machine (e.g., GPUs, storage, NICs), on-demand, to address application needs. We were first introduced to composable infrastructure in 2018, and at the same time, there was growing demand among our College of Engineering faculty for GPU systems for data science, artificial intelligence / machine learning / deep learning, and visualization. Many purchased their own individual desktop or deskside systems, a few pursued more costly cloud and HPC solutions, and others looked to the College or campus computer center for GPU resources which, at the time, were scarce. After surveying the diverse needs of our faculty and studying product offerings by a few nascent startups in the composable infrastructure sector, we applied for and received a grant from the National Science Foundation in November 2019 to purchase a mid-scale system, configured to our specifications, for use by faculty and students for research and research training. 
This paper describes our composable infrastructure solution and implementation for our academic community. Given how modern workflows are progressively moving to containers and cloud frameworks (using Kubernetes) and to programming notebooks (primarily Jupyter), both for ease of use and for ensuring reproducible experiments, we initially adapted these tools for our system. We have since made it simpler to use our system, and now provide our users with a public facing JupyterHub server. We also added an expansion chassis to our system to enable composable co-location, which is a shared central architecture in which our researchers can insert and integrate specialized resources (GPUs, accelerators, networking cards, etc.) needed for their research. 
In February 2020, installation of our system was finalized and made operational and we began providing access to faculty in the College of Engineering. Now, two years later, it is used by over 40 faculty and students plus some external collaborators for research and research training. Their use cases and experiences are briefly described in this paper. Composable infrastructure has proven to be a useful computational system for workload variability, uneven applications, and modern workflows in academic environments. 
Keywords - composable infrastructure, deep learning, visualization, resource management, workload management, user workflow, composable co-location, infrastructure as code

COMPaaS Hardware Racks

Composable Infrastructures for an Academic Research Environment: Lessons Learned

EVL is a technology Innovator, building industry and cross-institutional connections and partnerships. In 2022, EVL deployed two GPU FIONAs in partnership with the Pacific Research Platform (PRP) and the University of California San Diego. The PRP is a partnership of more than 50 institutions that establishes a science-driven high-capacity data-centric &lsquo;freeway system&rsquo;. PRP has support from the National Science Foundation, and includes the US Department of Energy&rsquo;s Energy Science Network (ESnet), as well as scores of research universities in the nation and around the world. PRP leverages Nautilus, a HyperCluster for running containerized Big Data Applications using Kubernetes, and FIONA, flash input/output network applications. A GPU FIONA is a rack-mounted PCs designed by the PRP team that is optimized for 100Gbps large data storage with Mellanox ConnectX-6 SmartNICs, a 24 core Epic Rome CPU, 256GB of memory and 8 NVIDIA RTX-3090 GPUs. Nautilus provides over 7500 CPU cores, over 500 GPUs and 4 PBs of storage. The addition of two Nautilus GPU FIONAs at EVL builds upon our 2018 DTN (Data Transfer Node) that has provided over 300GB of Ceph Storage to Nautilus and our CAVE2 GPU Nodes that are contributing 5 NVIDIA GPUS. EVL provides Nautilus with 300GB of Ceph Storage, 128 CPU cores, and 21 NVIDIA GPUs. The platform is utilized for a wide variety of applications, including applying advanced astronomy AI to microscopy workflow, simulations for natural disasters, climate modeling, virtual reality and even accelerating NSF&rsquo;s IceCube Photon Propagation. UIC is enabled to participate in these cross-institutional connections using UIC&rsquo;s HPRNet, a network environment tailored to the needs of high-performance science applications, built on a Science DMZ architecture. 
<a href="https://pacificresearchplatform.org/media/">Read more about PRP.</a> 
<a href="https://pacificresearchplatform.org/media/sc21-larry-smarr-on-the-rise-of-supernetwork-data-intensive-computing/">See media coverage from SC21 on PRP and Nautilus.</a>

FIONA

Pacific Research Platform (PRP) - Nautilus

In today&rsquo;s Big Data era, data scientists require modern workflows to quickly analyze large-scale datasets using complex codes to maintain the rate of scientific progress. These scientists often rely on available campus resources or off-the-shelf
computational systems for their applications. Unified infrastructure or over-provisioned servers can quickly become bottlenecks for specific tasks, wasting time and resources. Composable infrastructure helps solve these problems by providing users with new ways to increase resource utilization. Composable infrastructure disaggregates a computer&rsquo;s components - CPU, GPU (accelerators), storage and networking - into fluid pools of resources, but typically relies upon infrastructure engineers to architect individual machines. Infrastructure is either managed with specialized command-line utilities, user interfaces or specification files. These management models are cumbersome and difficult to incorporate into data-science workflows. We developed a high-level software API, Composastructure, which, when integrated into modern workflows, can be used by infrastructure engineers as well as data scientists to reorganize composable resources on demand. Composastructure enables infrastructures to be programmable, secure, persistent and reproducible. Our API composes machines, frees resources, supports multi-rack operations, and includes a Python module for Jupyter Notebooks. 
Keywords: distributed systems, testbed implementation and deployment, composable infrastructure, deep learning, visualization, infrastructure as code

Composable infrastructure holds the promise of accelerating the pace of academic research and discovery by enabling researchers to tailor the resources of a machine (e.g., GPUs, storage, NICs), on-demand, to address application needs. We were first introduced to composable infrastructure in 2018, and at the same time, there was growing demand among our College of Engineering faculty for GPU systems for data science, artificial intelligence / machine learning / deep learning, and visualization. Many purchased their own individual desktop or deskside systems, a few pursued more costly cloud and HPC solutions, and others looked to the College or campus computer center for GPU resources which, at the time, were scarce. After surveying the diverse needs of our faculty and studying product offerings by a few nascent startups in the composable infrastructure sector, we applied for and received a grant from the National Science Foundation in November 2019 to purchase a mid-scale system, configured to our specifications, for use by faculty and students for research and research training. 
This paper describes our composable infrastructure solution and implementation for our academic community. Given how modern workflows are progressively moving to containers and cloud frameworks (using Kubernetes) and to programming notebooks (primarily Jupyter), both for ease of use and for ensuring reproducible experiments, we initially adapted these tools for our system. We have since made it simpler to use our system, and now provide our users with a public facing JupyterHub server. We also added an expansion chassis to our system to enable composable co-location, which is a shared central architecture in which our researchers can insert and integrate specialized resources (GPUs, accelerators, networking cards, etc.) needed for their research. 
In February 2020, installation of our system was finalized and made operational and we began providing access to faculty in the College of Engineering. Now, two years later, it is used by over 40 faculty and students plus some external collaborators for research and research training. Their use cases and experiences are briefly described in this paper. Composable infrastructure has proven to be a useful computational system for workload variability, uneven applications, and modern workflows in academic environments. 
Keywords: composable infrastructure, deep learning, visualization, resource management, workload management, user workflow, composable co-location, infrastructure as code

This academic year 2023 EVL alum Mike Papka has joined UIC&rsquo;s Department of Computer Science and EVL faculty as a professor, bringing years of expertise in high performance computing to campus. For nearly three decades, Papka has held the position of Research Scientist at Argonne National Laboratory contributing to research in high performance computing, scientific visualization of large and complex datasets, information visualization related to high-performance computing, virtual reality for science and the integration of edge services. 
Since 2006, he’s served as Deputy Associate Laboratory Director, Computing, Environment and Life Sciences (CELS) Directorate, providing strategic planning and management of the Laboratory Directorate comprising Biosciences, Computational Science, Data Science and Learning, Environmental Sciences, Mathematics and Computer Science, a joint institute with Northwestern University, and the Argonne Leadership Computing Facility. In his role, Papka aids in executing a research program that aims to couple computing-related activities with several science domains whose future is tied to progress in computing.
 For the last 12 years, Papka has served and continues to serve as Division Director at Argonne Leadership Computing Facility (ALCF). He is responsible for planning and executing facility upgrades and integrated data strategies, risk assessment and management, operations reporting and reviews, staff promotion and retention, and safety. The ALCF is home to some of the world&rsquo;s fastest supercomputers: Theta; a Cray XC40 supercomputer based on Intel’s Xeon Phi KNL processor and NVIDIA DGX-3, Polaris; a HPE Apollo 6500 Gen10+ based on AMD Milan processor and NVIDIA A100s, and Aurora; a Intel CPU/GPU system currently being installed that support scientific and engineering open science research. 
We welcome Mike&rsquo;s return to UIC and EVL! He undoubtedly will be a valued asset to our research and educational enterprise.

Alum Michael E. Papka Joins EVL Faculty

Contrails are condensation trails generated from emitted particles by aircraft engines, which perturb Earth&rsquo;s radiation budget. Simulation modeling is used to interpret the formation and development of contrails. These simulations are computationally intensive and rely on high-performance computing solutions, and the contrail structures are not well defined. We propose a visual computing system to assist in defining contrails and their characteristics, as well as in the analysis of parameters for computer-generated aircraft engine simulations. The back-end of our system leverages a contrail-formation criterion and clustering methods to detect contrails&rsquo; shape and evolution and identify similar simulation runs. The front-end system helps analyze contrails and their parameters across multiple simulation runs. The evaluation with domain experts shows this approach successfully aids in contrail data investigation. 
Index Terms: Scalar Field Data, Physical & Environmental Sciences, Mathematics, Feature Detection, Tracking & Transformation 
 <iframe width="420" height="315"src="https://www.youtube.com/embed/mFllS2wQW8I">
</iframe>

Run analysis.

Visual Analysis and Detection of Contrails in Aircraft Engine Simulations

The Electronic Visualization Laboratory at the University of Illinois Chicago acquired a 24 compute node, 64 GPU composable infrastructure compute cluster, COMPaaS DLV - Composable Platform as a Service: Instrument for Deep Learning & Visualization in 2019.  Since then 40+ users have designed and run computations on COMPaaS.  To track cluster utilization, evaluate hardware
performance, and simply analyze application resource utilization of this experimental composable infrastructure system we set out to
collect metrics on the hardware and software infrastructure.  Users deploy their applications on the cluster using Kubernetes, a container orchestration platform that enables users to deploy applications on the cluster without the need to have complete knowledge of the underlying hardware and with minimal assistance from system administrators.  Following this, we utilized tooling already adapted for Kubernetes-based clusters to collect hardware and software metrics.  The metrics provided by this tooling is useful for system administrators in evaluating cluster performance and utilization, as well as useful for users in determining their application&rsquo;s performance and discovering bottlenecks.

GPU Power Consumption

Monitoring COMPaaS

UIC Department of Computer Science 
Doctoral Admissions Webinar 
November 2nd 2022 @7 am-8:30am Chicago time (CDT) 
Webinar: <a href="https://tinyurl.com/uicwebinar2022">https://tinyurl.com/uicwebinar2022</a> 
See what the Department of Computer Science at the University of Illinois Chicago has to offer! Located in the beautiful city of Chicago, UIC is among the most culturally diverse universities in the United States, and home to one of the most innovative computer science departments in the United States. UIC is Chicago&rsquo;s top public research university, and was recently ranked in the top 20 of the best 100 universities worldwide that are under 50 years of age. Doctoral students in their first five years of enrollment are guaranteed a fully funded appointment that includes a tuition waiver, and a stipend. Students are also eligible for summer support. Our students get jobs in academia, industry, and entertainment upon graduation, including positions at top research labs and companies such as Google, Microsoft, Apple, Disney, and Amazon. 
During this webinar, UIC Computer Science faculty will talk about our research agenda, application fee waivers, the application process, application expectations in terms of GPA/GRE etc, and will take questions. A recording of the webinar will be made available on our Graduate Admissions page. 
Apply: <a href="https://cs.uic.edu/graduate/admissions/">https://cs.uic.edu/graduate/admissions/</a> 
Deadline: December 15th 2022 
See our flyer at: <a href="https://tinyurl.com/uic2022flyer">https://tinyurl.com/uic2022flyer</a> 
 <iframe width="420" height="315"src="https://www.youtube.com/embed/eNoBbCm1skM">
</iframe>

Interested in pursuing a PhD in Computer Science in the United States?

Cluster schedulers are crucial in high-performance computing (HPC). They determine when and which user jobs should be allocated to available system resources. Existing cluster scheduling heuristics are developed by human experts based on their experience with speciﬁc HPC systems and workloads. However, the increasing complexity of computing systems and the highly dynamic nature of application workloads have placed tremendous burden on manually designed and tuned scheduling heuristics. More aggressive optimization and automation are needed for cluster scheduling in HPC. In this work, we present an automated HPC scheduling agent named DRAS (Deep Reinforcement Agent for Scheduling) by leveraging deep reinforcement learning. DRAS is built on a hierarchical neural network incorporating special HPC scheduling features such as resource reservation and backﬁlling. An efﬁcient training strategy is presented to enable DRAS to rapidly learn the target environment. Once being provided a speciﬁc scheduling objective given by the system manager, DRAS automatically learns to improve its policy through interaction with the scheduling environment and dynamically adjusts its policy as workload changes. We implement DRAS into an HPC scheduling platform called CQGym. CQGym provides a common platform allowing users to ﬂexibly evaluate DRAS and other scheduling methods such as heuristic and optimization methods. The experiments using CQGym with different production workloads demonstrate that DRAS outperforms the existing heuristic and optimization approaches by up to 50%. 
Index Terms - High-performance computing, cluster scheduling, deep reinforcement learning, job starvation, backﬁlling, resource reservation, OpenAI Gym 
<a href="https://ieeexplore.ieee.org/abstract/document/9894371">https://ieeexplore.ieee.org/abstract/document/9894371</a>

DRAS overview. DRAS agent (at the bottom) represents the scheduling agent; the environment (at the top) comprises the rest of the system, including job wait queue and HPC cluster.

DRAS: Deep Reinforcement Learning for Cluster Scheduling in High Performance Computing

Powerful detectors at modern experimental facilities routinely collect data at multiple GB/s. Online analysis methods are needed to enable the collection of only interesting subsets of such massive data streams, such as by explicitly discarding some data elements or by directing instruments to relevant areas of experimental space. Thus, methods are required for configuring and running distributed computing pipelines-what we call flows-that link instruments, computers (e.g., for analysis, simulation, artificial intelligence [AI] model
training), edge computing (e.g., for analysis), data stores, metadata catalogs, and high-speed networks. We review common patterns associated with such flows and describe methods for instantiating these patterns. 
We present experiences with the application of these methods to the processing of data from five different scientific instruments, each of which engages powerful computers for data inversion, model training, or other purposes. We also discuss implications of such methods for operators and users of scientific facilities. 
<a href="https://www.cell.com/patterns/fulltext/S2666-3899(22)00231-8">https://www.cell.com/patterns/fulltext/S2666-3899(22)00231-8</a>

Linking scientific instruments and computation: Patterns, technologies, and experiences

SAGE3 beta version software and documentation will be introduced next week at the IEEE/ACM Supercomputing (SC22) conference in Dallas, Texas, USA, November 13-18, 2022. The SAGE User Community is invited to attend the SC22 Birds of a Feather (BOF) meeting and participate in ongoing SAGE3 demonstrations in the SC22 Exhibition. 
Attend the &ldquo;SAGE 12th SC BOF: Smart Amplified Group Environment Enhanced with Artificial Intelligence for Global Collaboration&rdquo;: 
 * Tuesday, November 15, 2022 
 *12:15PM - 01:15PM CST 
 * Room: D168, Kay Bailey Hutchison Convention Center Dallas 
 * <a href="https://sc22.supercomputing.org/presentation/?id=bof141&sess=sess312">https://sc22.supercomputing.org/presentation/?id=bof141&sess=sess312</a> 
Join us for SAGE3 demonstrations in the StarLight research booth # 2847 on the exhibition floor, Tuesday-Thursday. 
SAGE3, the Smart Amplified Group Environment, is the next-generation, &ldquo;human-in-the-loop&rdquo; collaboration platform, providing HPC users with the tools to access, explore, discover, publish, share, integrate, and reuse complex datasets. SAGE3 supports today&rsquo;s hybrid work/home environments (laptops, single monitors and display walls), and interfaces with a variety of computational infrastructures, workflows, notebooks, and analytics software through Artificial Intelligence-enabled services and orchestration services. It lowers the barrier of entry into AI for non-expert users, democratizing access to disruptive technologies for those with varying skills. 
The BOF will not be streamed. SAGE3 software and documentation will be available for download on the SAGE3 website. The SAGE3 team plans to hold SAGE3 educational webinars, starting next year. 
REMINDER: SAGE3 / SAGE2 COMMUNITY USER FORUMS are now hosted on Slack. To join, go to the <a href="https://sage3.sagecommons.org/">SAGE3 website</a> and click on the Support menu. 
 * For SAGE3, subscribe to the sage3-support channel. 
 * For SAGE2, subscribe to the sage2-support channel. 
Thank you for your longstanding support of SAGE and we look forward to continued collaborations. 
Maxine Brown 
SAGE team member 
<a href="mailto:maxine@uic.edu">maxine@uic.edu</a> 
SAGE3 is being developed by the Laboratory for Advanced Visualization and Applications (LAVA) and the Hawaii Data Science Institute (HI-DSI) at University of Hawaiʻi at Mānoa (UHM), the Electronic Visualization Laboratory (EVL) at University of Illinois Chicago (UIC), and the InfoVis Lab at Virginia Tech (VT). SAGE3 receives major funding from the U.S. National Science Foundation (NSF), awards #2004014 (UHM), #2003800 (UIC) and #2003387 (VT). 
SAGE2 received major NSF support, award #ACI-1441963, to UHM and UIC. 
SAGE, SAGE2 and SAGE3 are trademarks of the University of Illinois Board of Trustees.

INTRODUCING SAGE3 at the SC22 - Tuesday, November 15, 12:15pm - 1:15pm

The U.S. Department of Energy (DOE) has awarded a $2.6 million grant to lead institution Northern Illinois University and sub-recipient the University of Illinois Chicago for classroom training and research opportunities in computational high energy physics. 
EVL faculty and Computer Science professor Mike Papka, and EVL director of research, professor Andy Johnson will lead the computational and high-performance computing aspect of this project. They will collaborate with UIC Physics professor Richard Cavanaugh to mentor students in computational high-energy physics education. 
The 5-year award will provide two dozen master&rsquo;s-level students in physics and computer science stipends, tuition reimbursement, and distinctive research experiences in the HEP domain. 
Dr. Cavanaugh specializes in Particle Physics and holds a joint appointment with Fermi National Accelerator Laboratory (FermiLab). As the Deputy Associate Laboratory Director, Computing, Environment, and Life Sciences at Argonne National Laboratory (ANL), Dr. Papka has unprecedented experience in high-performance computing. They will provide students a unique opportunity to collaborate with scientists at the DOE&rsquo;s ANL and FermiLab. 
See the full announcement in the <a href="https://newsroom.niu.edu/2-6m-doe-award-will-support-training-in-computational-high-energy-physics/">Northern Illinois University Newsroom</a>.

DOE Awards NIU and UIC Funds to Support Training in Computational High Energy Physics

MS Candidate

Hal Brynteson

Vamshi Chakrala

MS Student Undergraduate Researcher 2023

Sreten Kljaic

Undergraduate Researcher

Fredric Massa

Aaqel Shaik

Research Experience Undergraduate

Aakash Lilaramani

Standing in front of Argonne National Laboratory&rsquo;s Aurora supercomputer, EVL and computer science department professor
Michael E. Papka is the Featured Photo of the Spring 2023 RISE publication. Papka holds a joint appointment with UIC and is also Deputy Associate Laboratory Director, Computing, Environment, and Life Sciences, Argonne National Laboratory. Among his charge is the new Aurora supercomputer, which once fully online, will be among the United States&rsquo; most powerful exascale machines used for research. Sponsored by the Department of Energy and designed by Intel and Hewlett Packard Enterprise, Aurora will provide scientists a new and extremely powerful tool for a host of research initiatives. 
<a href="https://engineering.uic.edu/about/coe-news/rise/rise-spring-2023/">See the RISE: Spring 2023 feature</a>.

EVL Professor Papka is the Featured Photo in UIC Engineering&rsquo;s RISE 2023

Nguyen Hoa Pham

Honorable Mention in the EuroVis Best Paper Category 
Exploratory analysis of the chemical space is an important task in the field of cheminformatics. For example, in drug discovery research, chemists investigate sets of thousands of chemical compounds in order to identify novel yet structurally similar synthetic compounds to replace natural products. Manually exploring the chemical space inhabited by all possible molecules and chemical compounds is impractical, and therefore presents a challenge. To fill this gap, we present ChemoGraph, a novel visual analytics technique for interactively exploring related chemicals. In ChemoGraph, we formalize a chemical space as a hypergraph and apply novel machine-learning models to compute related chemical compounds. It uses a database to find related compounds from a known space and a machine learning model to generate new ones, which helps enlarge the known space. Moreover, ChemoGraph highlights interactive features that support users in viewing, comparing, and organizing computationally identified related chemicals. With a drug discovery usage scenario and initial expert feedback from a case study, we demonstrate the usefulness of ChemoGraph. 
Keywords: chemical space exploration, cheminformatics, multipartite graphs, data visualization 
CCS Concepts: Applied computing → Chemistry 
<iframe width="560" height="315" src="https://www.youtube.com/embed/kVpz6TDjQMw" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" allowfullscreen></iframe>

An overview of ChemoGraph.

ChemoGraph: Interactive Visual Exploration of the Chemical Space

Most real-world networks are both dynamic and multivariate in nature, meaning that the network is associated with various attributes and both the network structure and attributes evolve over time. Visualizing dynamic multivariate networks is of great significance to the visualization community because of their wide applications across multiple domains. However, it remains challenging because the techniques should focus on representing the network structure, attributes, and their evolution concurrently. Many real-world network analysis tasks require the concurrent usage of the three aspects of the dynamic multivariate networks. In this paper, we analyze current techniques and present a taxonomy to classify the existing visualization techniques based on three aspects: temporal encoding, topology encoding, and attribute encoding. Finally, we survey application areas and evaluation methods; and discuss challenges for future research. 
CCS Concepts Human-centered computing → Graph drawings

Overview of the taxonomy of DMVN visualization techniques along with illustrations of the categories.

The State of the Art in Visualizing Dynamic Multivariate Networks

EVL professor Mike Papka and his research students have been awarded an Honorable Mention in the EuroVis Best Paper Category for &ldquo;ChemoGraph: Interactive Visual Exploration of the Chemical Space&rdquo;. Co-authored by collaborators Bharat Kale, Austin Clyde, Maoyuan Sun, Arvind Ramanathan, Rick Stevens, and Michael E. Papka, ChemoGraph is a novel visual analytics technique for interactively exploring related chemicals. A chemical space is formalized as a hypergraph with novel machine-learning models to compute related chemical compounds applied. It uses a database to find related compounds from an available space and a machine learning model to generate new ones, which helps enlarge the known space. Moreover, ChemoGraph highlights interactive features that support users in viewing, comparing, and organizing computationally identified related chemicals. 
<a href="https://youtu.be/kVpz6TDjQMw ">See a video teaser on EVLTube.</a>

&ldquo;ChemoGraph: Interactive Visual Exploration of the Chemical Space&rdquo; Awarded with Honorable Mention in the EuroVis Best Paper Category

EVL Faculty, Professor, Computer Science

Zhiling Lan

Phd Student

Landon Richard Dyken

PhD Candidate

Ahmedur Rahman Shovon

PhD Student

Matthew Dearing

Ke Fan

Melanie Cornelius

Starting Academic Year 23/24 visiting professor Zhiling Lan joins as faculty in the Department of Computer Science and EVL. Dr. Lan comes with significant expertise in high-performance computing and parallel/distributed systems. Prior to her tenure at UIC, she was a professor of Computer Science, at the Illinois Institute of Technology between 2002 and 2023. She is also a guest research faculty at Argonne National Laboratory. 
Dr. Lan also heads the <a href="https://spear.lab.uic.edu/">SPEAR (Systems for Performance, Energy, And Resiliency) lab</a>, which conducts research spanning various areas of parallel and distributed systems including cluster management, interconnection networking, performance modeling and simulation, power and energy efficiency, and fault tolerance. SPEAR&rsquo;smission is to unify systems and AI, where we are committed to creating software systems to expedite AI applications and exploring advanced AI technologies to tackle critical research challenges in computer systems.

EVL Welcomes Professor Zhiling Lan to High Performance Computing Faculty Team

Teddy Zaremba

Abdullah Ali

EVL Faculty Assistant Professor, Computer Science

Sidharth Kumar

EVL alumni Tommy Marrinan, Victor Mateevitsi and Mike Papka (also EVL Professor) are recognized at <a href="https://sciviscontest2023.github.io/">2023 IEEE VIS SciVis Contest</a> for their research in the visualization of neuronal network simulations of plasticity changes in the human brain. 
The first award was for Best Workflow; enabling a scientific workflow was identified as one of the 4 main tasks for this year&rsquo;s contest. The goal of this component was to combine the functionality of the other tasks into a combined framework and interface that enables experts to explore the data and investigate several questions interactively. Which tools help the users to mark, select, and analyze the data on different levels of complexity and transfer information among these levels? 
The second award was for 1st Place Overall; judges took into account how the submission addressed all 4 main tasks, as well as any additional features. The judging was based on a short paper description of the visualization system and an 8-minute video demo. 
Congratulations, Tommy, Victor, and Mike! 
See <a href="https://insidehpc.com/2023/11/alcf-team-wins-first-place-and-best-workflow-at-2023-ieee-scivis-contest/">Inside HPC</a>, &ldquo;ALCF Team Wins First Place and Best Workflow at 2023 IEEE SciVis Contest&rdquo;

EVL Alumni Win Best Workflow and 1st Place Overall at 2023 IEEE SciVis Contest

Undergraduate Researcher, Data Science - Business Analytics

Athena Angara

EVL alumnus Khairi Reda presents his research at EVL tech meeting: 
Friday, November 3, 2023, 12PM - 1PM 
EVL Continuum, Room 2068 ERF 
Towards Trustworthy Visual Analytics 
Abstract: Interactive visualization tools, such as Tableau, Jupyter notebooks, and R with ggplot, have become vital instruments in the data science ecosystem. These systems empower users to explore data and search for new patterns that might lead to insights. However, these tools can also expose analysts to noise and random data fluctuations. Analysts can frequently misinterpret these spurious patterns and mistake them for reliable effects. This phenomenon has led to a &ldquo;replication crisis,&rdquo; with many previously published results failing to replicate, eroding trust in the scientific process. 
In this talk, Reda will discuss his research on reducing the risk of spurious visualizations and describe techniques we have developed to operationalize analyst hypotheses and beliefs in the visual analytics process. Empirical results showing that these unobtrusive interactions can promote good analytical practices and reduce the incidence of false discoveries will be discussed. Finally, Reda will show how using visual inference as a benchmark can provide a complementary account of visualization utility, leading to new insights into visualization design beyond what can be learned from traditional &ldquo;graphical perception&rdquo; studies. 
Bio: Khairi Reda is an Associate Professor in the Luddy School of Informatics & Computing at Indiana University Indianapolis. He is an alumnus of EVL, having earned a PhD in CS from the University of Illinois at Chicago. His research employs a variety of methods, weaving together techniques from human-computer interaction, machine learning, and design thinking to investigate new approaches for visualizing and interacting with data.

EVL Alumnus Khairi Reda Presents his Research &ldquo;Towards Trustworthy Visual Analytics&rdquo;

Numerous sophisticated profiling and visualization tools have been developed to enable programmers to expose semantic information from their application components. However, effective and interactive exploration of the profiles of large-scale parallel programs remains a challenge due to the high I/O overheads of profiles and the difficulties in scaling downstream visualization tools. In this poster, we present a full-stack approach to a performance introspection framework that tackles key challenges in profiling and visualizing performance data at scale. Our novelty lies in a scalable and compact data model and a two-phase I/O system, which instill scalability into the profiler making it low overhead-- even at high process counts (< 5%). We then build a web-based, visual-analytic dashboard with linked views. Our profiling and visualization tools are both lightweight and easy-to-use, which strikes a balance between providing sophisticated features and operating quickly and efficiently at high process counts. 
This work was funded in part by NSF RII Track-4 award 2132013, NSF PPoSS planning award 2217036, NSF PPoSS large award 2316157 and, NSF collaborative research award 2221811. We thank the ALCF's Directors Discretionary (DD) program for offering us the compute hours to run our experiments on the Theta Supercomputer.

End-to-End Framework - Viveka

Two-phase IO Enabling Large-scale Performance Introspection (poster)

PhD student Ke Fan&rsquo;s research in profiling and visualizing performance data at scale is recognized with a best poster nomination at SC23. &ldquo;Two-phase IO Enabling Large-scale Performance Introspection&rdquo; focuses on developing an effective and interactive exploration of the profiles of large-scale parallel programs, which remain a challenge due to the high I/O overheads of profiles and the difficulties in scaling downstream visualization tools. 
The poster presents Viveka, a lightweight end-to-end system for profiling and visualizing the performance of MPI-based applications. 
- A simple data format for the generated logs that minimizes metadata, leading to a smaller storage footprint and faster load times. (on average 3x more lightweight than Caliper). 
- Development of a custom two-phase data aggregation system to scale parallel I/O (of performance data) to high core counts, ensuring minimal overhead. 
- A lightweight web-based visualization dashboard that is capable of performing interactive analysis of performance data collected at high process counts. 
- A case study on real parallel applications to demonstrate the efficacy of our profiling and visualization system. 
Funding in part by NSF RII Track-4 award 2132013, NSF PPoSS planning award 2217036, NSF PPoSS large award 2316157 and, NSF collaborative research award 2221811. Compute hours provided on Theta Supercomputer Argonne Leadership Computing Facility&rsquo;s Directors Discretionary (DD) program.

Two-phase IO Enabling Large-scale Performance Introspection

The challenges of profiling and visualizing large-scale parallel programs

EVL Affiliated Faculty Sidharth Kumar&rsquo;s Advisee Recognized with Best Poster Nomination at SC23

In the realm of Computational Fluid Dynamics (CFD), the demand for memory and computation resources is extreme, necessitating the use of leadership-scale computing platforms for practical domain sizes. This intensive requirement renders traditional checkpointing methods ineffective due to the significant slowdown in simulations while saving state data to disk. As we progress towards exascale and GPU-driven High-Performance Computing (HPC) and confront larger problem sizes, the choice becomes increasingly stark: to compromise data fidelity or to reduce resolution. To navigate this challenge, this study advocates for the use of in situ analysis and visualization techniques. These allow more frequent data &ldquo;snapshots&rdquo; to be taken directly from memory, thus avoiding the need for disruptive checkpointing. We detail our approach of instrumenting NekRS, a GPU-focused thermal-fluid simulation code employing the spectral element method (SEM), and describe varied in situ and in transit strategies for data rendering. Additionally, we provide concrete scientific use cases and report on runs performed on Polaris, Argonne Leadership Computing Facility’s (ALCF) 44 Petaflop supercomputer and Jülich Wizard for European Leadership Science (JUWELS) Booster, Jülich Supercomputing Centre’s (JSC) 71 Petaflop High Performance Computing (HPC) system, offering practical insight into the implications of our methodology.

Side view visualization of a RBC case.

Scaling Computational Fluid Dynamics: In Situ Visualization of NekRS using SENSEI

A team led by researchers at the Argonne Leadership Computing Facility (ALCF) including EVL Professor Mike Papka was recognized with the Best Paper Award for an innovative scientific visualization study presented at the 2023 In Situ Infrastructures for Enabling Extreme-scale Analysis and Visualization (ISAV) at the SC23 conference in Denver, Colorado. 
The paper, &ldquo;Scaling Computational Fluid Dynamics: In Situ Visualization of NekRS using SENSEI,&rdquo; (Mateevitsi, V.A., Bode, M., Ferrier, N., Fischer, P., Göbbert, J.H., Insley, J.A., Lan, Y., Min, M., Papka, M.E., Patel, S., Rizzi, S., Windgassen, J.) details the team&rsquo;s efforts to equip the NekRS computational fluid dynamics (CFD) code with in situ analysis and visualization capabilities. The winning paper explores the benefits of instrumenting the NekRS code with SENSEI, an open-source in situ interface developed with support from DOE&rsquo;s Exascale Computing Project. The researchers demonstrated the capabilities and scalability of their approach by running two real scientific use cases on the ALCF&rsquo;s Polaris supercomputer and JSC&rsquo;s JUWELS Booster supercomputer. They found that the ability to &ldquo;see&rdquo; the data in real-time significantly augmented their analytical capacities.

Visualization of the pb146 use case simulation, illustrating flow dynamics within a pebble-bed nuclear reactor.

ALCF Team Including EVL Professor Michael Papka Wins Best Paper at SC23 ISAV 2023 Workshop

Marco Bonafini

Riccardo Strina

Pietro Lodi Rizzini

Undergraduate Researcher - Edge Computing

Fatima Jassim

EVL Professor Michael Papka is among 52 ACM, the Association for Computing Machinery Distinguished Members inducted into the International Computing Society for significant contributions towards pivotal education, engineering, and science. Each Distinguished Member is a longstanding ACM Member selected by their peers for work that has advanced computing, fostered innovation across various fields, and improved computer science education. 
Dr. Papka is recognized for his contributions in virtual reality, collaborative environments, scientific visualization, as well as research and operations in high-performance computing. EVL faculty, staff and students congratulate him on this prestigious honor! 
See the <a href="https://www.acm.org/media-center/2024/january/distinguished-members-2023">ACM Media Center</a> for the full press release and information on the other recognized members for 2023.

ACM Distinguished Members 2023 Press Release

EVL Professor Michael Papka Recognized as ACM Distinguished Member

Alessandro Martinolli

The development and construction of facilities that allow for efficient data sharing while also providing computational resources for data analysis is becoming a crucial activity across all fields of science. In many research areas, the generation of large-scale, sophisticated simulations on HPC platforms is now as important a part of science as are the associated experimental and theoretical studies. The simulations can be computationally expensive, and often only a limited number of codes exist that can take full advantage of large-scale HPC systems. Therefore, sharing simulation results within the community to enable comparisons to experimental data and further data exploration in search of new insights is essential to push science forward. 
EVL graduate research assistants work with Argonne National Laboratory scientists on the following tasks: 
• Active development of the data portal to support releasing cosmological simulation data to the community. 
• Test data portal and develop tests to increase the portal&rsquo;s robustness. 
• Develop of documentation and tutorials on the use of data portal.

Data Sharing and Analysis Across Leadership Computing Facilities

EVL graduate research assistants will work with Argonne National Laboratory scientists to develop enhancements to SciStream, a platform that enables memory-to-memory data streaming between federated scientific instruments residing in different security domains with no direct external network connectivity. We will explore utilizing advanced packet processing technologies such as DPDK, programmable processors such as DPU and programmable networking to improve the performance of SciStream data plane. We will also benchmark SciStream performance for different deployment scenarios including deploying SciStream alongside Globus Connect Servers on existing DTNs and deploying SciStream on dedicated streaming DTNs. We will also benchmark the performance of SciStream between Argonne&rsquo;s Advanced Photon Source and the Argonne Leadership Computing Facility. 
Research activities include: 
1) Research and integrate advanced packet processing and programmable data processing in SciStream and characterize performance in testbed environments. 
2) Evaluate the performance of SciStream in testbed and production environments for various deployment scenarios. 
3) Create reports and contribute to the development of scholarly manuscripts.

Enhancements to SciStream, a platform to enable memory-to-memory data streaming between federated instruments

The George Crabtree Institute for Discovery and Sustainability is an institutional partnership between Argonne National Laboratory and the University of Illinois Chicago (UIC), enabling transformative solutions for complex scientific challenges. This collaboration aims to advance knowledge in science and technology, propel sustainable innovation, promote diversity in perspectives, and be supportive of the next generation of researchers. 
EVL and Computer Science professor Michael E. Papka and Jordi Cabana will lead the institute as its founding co-directors, actively engaging with Argonne and UIC in their current leadership roles. Their co-director roles will bring forward collaboration opportunities, coordinate within and across both institutions and support future scientists and interdisciplinary, inclusive research. 
Some of the institute&rsquo;s initial programming includes faculty engagement via the Argonne Collaborative Research Fellowship Program, which is supported through strategic use of the Joint Appointment Program, recruitment of undergraduate and graduate students, and scientific workshops.

EVL Professor Michael Papka co-Directs the George Crabtree Institute for Discovery and Sustainability

Contrails are condensation trails generated from emitted particles by aircraft engines, which perturb Earth’s radiation budget. Simulation modeling is used to interpret the formation and development of contrails. These simulations are computationally intensive and rely on high-performance computing solutions, and the contrail structures are not well defined. We propose a visual computing system to assist in defining contrails and their characteristics, as well as in the analysis of parameters for computer-generated aircraft engine simulations. The back end of our system leverages a contrail-formation criterion and clustering methods to detect contrails&rsquo; shape and evolution and identify similar simulation runs. The front-end system helps analyze contrails and their parameters across multiple simulation runs. The evaluation with domain experts shows this approach successfully aids in contrail data investigation. 
Index Terms: Scalar Field Data, Physical & Environmental Sciences, Mathematics, Feature Detection, Tracking & Transformation 
<a href="http://dx.doi.org/10.1109/TVCG.2022.3209356">http://dx.doi.org/10.1109/TVCG.2022.3209356</a>

Contrail structure groups and evolution over time.

The July 2024 issue of IEEE Computing Edge magazine&rsquo; section on Data Visualization features the contrail visual computing digital twin work conducted by EVL graduate students Nafiul Nipu, and Carla Floricel, along with their advisor professor G. Elisabeta Marai. This research was done in collaboration with professor Roberto Paoli (MIE). 
Contrails are condensation trails generated from emitted particles by aircraft engines, which perturb Earth's radiation budget. Simulation modeling is used to interpret the formation and development of contrails. These simulations are computationally intensive and rely on high-performance computing solutions, and the contrail structures are not well defined. We propose a visual computing system to assist in defining contrails and their characteristics, as well as in the analysis of parameters for computer-generated aircraft engine simulations. The back end of our system leverages a contrail-formation criterion and clustering methods to detect contrails&rsquo; shape and evolution and identify similar simulation runs. The front-end system helps analyze contrails and their parameters across multiple simulation runs. The evaluation with domain experts shows this approach successfully aids in contrail data investigation.

Visual analysis of contrails using multiple coordinated views.

EVL Researchers&rsquo; Contrail Digital Twin Work Featured in IEEE CS Magazine

Amy Byrnes

EVL PhD candidate Ke Fan has been named a recipient of the 2024 ACM-IEEE CS George Michael Memorial HPC Fellowship. The fellowship honors exceptional PhD students throughout the world whose research focus is on high-performance computing applications, networking, storage, or large-scale data analytics using the most powerful computers that are currently available. The fellowship is named after George Michael one of the founders of the Supercomputing Conference. 
Read the <a href="https://www.acm.org/media-center/2024/august/george-michael-fellowship-recipients-2024">AMC news release</a>. 
Join us in congratulating Ke and her advisor Computer Science Assistant Professor and EVL faculty Sidharth Kumar on this exceptional distinction!

EVL PhD Candidate Ke Fan Named 2024 ACM-IEEE CS George Michael Memorial HPC Fellowship

Shambhawi Sharma

Jyotsna Rajaraman

Charis Christopher Hulu

MS Student

Wesley Kwiecinski

Tong Ba Ta

Piotr Brozek

Yash Kurkure

September 16, 2024 - The University of Illinois Chicago (UIC) is proud to become an Intel AI/oneAPI Center of Excellence, advancing computational science education for AI and next-generation heterogeneous programming. The Center will integrate the open, standards-based oneAPI programming model and the latest AI development techniques into UIC&rsquo;s courses, preparing students to meet the demands of AI, high-performance computing (HPC), and data visualization in today&rsquo;s scientific and engineering landscapes. The Center of Excellence builds on the efforts of Dr. Michael E. Papka, who has supported Chicago-area university oneAPI workshops for the past two years and has been a strong advocate for oneAPI since its inception. 
UIC&rsquo;s AI/oneAPI Center of Excellence will be a hub for developing focused modules to enhance current AI, HPC, and visualization courses. The center will initially focus on developing modules to integrate seamlessly into regular classes and outreach activities, expanding access to advanced computational education for a wider audience. The goal at UIC is to prepare students to interpret and visually represent complex data sets using cutting-edge technologies such as Intel Embree and OSPRay and to teach them how to use SYCL in HPC to meet the challenges of exascale computing. 
&ldquo;The AI/oneAPI Centers of Excellence are aiding the next generation of developers in advancing future software innovation using the power of CPUs, GPUs, and other accelerators,&rdquo; says Dr. Michael E. Papka, Professor of Computer Science at UIC and the Director of the Argonne Leadership Computing Facility, a Department of Energy Office of Science user facility at Argonne National Laboratory. &ldquo;This UIC-based collaboration with Intel offers valuable, real-world experience and opportunities for our students.&rdquo;
The UIC Center&rsquo;s AI and HPC curriculum will present essential tools and libraries, such as the oneAPI infrastructure for AI and Intel rendering tools, empowering students to optimize code for parallel architectures and high-performance multiarchitecture environments. Through hands-on labs and coding assignments, UIC students will tackle real-world optimization challenges, preparing them to leverage the immense power of exascale systems for scientific discovery. The next phase will consider adding AI curriculums addressing HPC and large-scale AI problems harnessing the power of Intel hardware. 
&ldquo;We are excited to see the impact of these educational advancements on our students and the broader scientific community,&rdquo; says Dr. Peter Nelson, Dean of Engineering and Professor of Computer Science at UIC. &ldquo;By integrating cutting-edge tools and technologies, we are fostering a versatile, technologically adept workforce capable of leveraging AI and HPC for scientific discovery.&rdquo;
Students will engage in projects such as visualizing climate change effects, showcasing energy resource distributions with interactive maps, and rendering materials science structures at the atomic level. These projects bridge the gap between theoretical knowledge and practical application, connecting students&rsquo; work with real-world scenarios and emphasizing HPC&rsquo;s role in solving global challenges. Moreover, HPC supports the rise of AI applications in scientific computing by training complex AI models, from natural language processing to autonomous systems. This interaction between HPC and AI is crucial for advancing both fields, fostering innovation, and driving technological progress. 
&ldquo;UIC&rsquo;s Intel AI/oneAPI Center of Excellence will challenge students to drive complex and important computational projects such as simulating climate change impacts, modeling energy resource distributions, and exploring new materials properties. This brings real-world applicability of AI and HPC into the classroom, enabling the next generation of scientists and engineers to push the boundaries in helping to solve global challenges,&rdquo; says Joe Curley, vice president, and general manager of Intel Software Products and Ecosystems. 
This initiative represents a strategic alignment between academic curriculum development and industry needs. By equipping students with the skills and knowledge necessary to leverage AI, HPC, and visualization technologies, UIC is contributing to cultivating a highly skilled workforce capable of advancing the frontiers of science and engineering. 
For more information, please contact: 
Professor Michael E. Papka – <a href=mailto:"papka@uic.edu">papka@uic.edu</a>

University of Illinois Chicago Establishes a Center of Excellence focused on AI

Computer Science (CS) Professor and EVL faculty Micheal E. Papka is being honored among only a few of his colleagues with the Warren S. McCulloch Professorship of Computer Science. This recognition is the first-ever CS Professorship Investiture. Dr. Papka and fellow CS Professor Barbara DiEugenio are the first two faculty to receive this award. 
Among the other honorees are Professor Cornelia Caragea - first holder of the Robert V. Kenyon Professorship of Computer Science and Distinguished Professor Bing Liu - first holder of the Peter L. and Deborah K. Wexler Professorship of Computing. 
At this inaugural event, both Bob Kenyon and his wife Cynthia Vahlkamp and Peter and Deborah Wexler will be reconginized for their very generous donations of professorships. 
The investiture will be held: 
Friday, September 20, 2024 
3:00 PM Reception, 3:30 PM Ceremony 
ERF West Atrium 
842 W Taylor St, Chicago, IL 60607 
For questions, contact <a href="mailto:rszandra@uic.edu">Ms. Renata Szandra</a>

Micheal E. Papka Honored with Warren S. McCulloch Professorships of Computer Science

Seyfal Sultanov

Jiaxin Lu

Yiheng Tao

Michael E. Papka has been announced as the new Director of EVL. Mike has a long history with EVL, starting in 1994 when he graduated with his Masters degree in Computer Science - - through many long-term collaborations between Argonne National Laboratory and EVL. 
After receiving his MS at UIC, Mike completed a second MS and PhD with the University of Chicago. Mike holds a joint appointment as a full UIC CS professor and at Argonne, where he has research and senior leadership positions. 
As the Director of the Argonne Leadership Computing Facility (ACLF), Mike leads one of the U.S. Department of Energy&rsquo;s premier facilities for open science. He has spearheaded numerous efforts to build and deploy world-class computing capabilities for open science, notably advancing AI and large-scale data analysis. Under his leadership, ALCF recently joined the exascale era with the development of the Aurora supercomputer. 
We welcome Mike as EVL&rsquo;s new leader!

Michael E. Papka Named EVL Director

Congratulations to Mike Papka, one of nine speakers at <a href="https://www.tedxchicago.com/">TEDxChicago 2024: Lighting the Way</a> to be held Friday, September 27th at Harris Theater, Millennium Park fro 3pm - 7pm. 
Mike holds a joint appointment as a UIC Computer Science faculty member and serves as the Argonne Leadership Computing Facility director at Argonne National Laboratory. As director of ALCF, Mike oversees the development and deployment of <a href="https://www.anl.gov/aurora">Aurora</a>, the <a href="https://top500.org/">world&rsquo;s second exascale supercomputer</a> and the fastest <a href="https://www.intel.com/content/www/us/en/newsroom/news/intel-powered-aurora-supercomputer-breaks-exascale-barrier.html">AI supercomputer</a>. 
Mike is one of the inaugural recipients of the William S. McCulloch Collegiate Professorship of Computer Science. Warren S. McCulloch was a UIC professor of psychiatry who with Walter Pitts co-invented <a href="https://link.springer.com/article/10.1007/BF02478259">artificial neural networks in 1943</a>. Today, neural nets form the foundation of much AI technology, including deep learning and large language models. 
Mike is a professor of Computer Science, director of EVL, and co-director of the <a href="https://crabtree.uic.edu/">George Crabtree Institute for Discovery and Sustainability</a>.

UIC CS William McCulloch Professor Mike Papka speaking at TEDxChicago 2024

Overview 
Advanced cyberinfrastructure (CI) is undergoing disruptive changes in system architectures and application workloads. Each facility typically hosts a blend of high-performance systems with various configurations and capabilities. These systems are composed of a multitude of components, embodying heterogeneity in various aspects, ranging from computing to memory and storage. Concurrently, the landscape of cyberinfrastructure workloads is rapidly expanding beyond traditional computational simulations to encompass a hybrid mix of applications including deep learning applications, data-centric applications, as well as coupled
workflows involving distributed computing systems. The emerging applications come with diverse resource requirements and exhibit distinct computing characteristics. 
Current resource management approaches are designed for the traditional paradigm, namely numerical simulations on homogeneous systems. They heavily rely on heuristics and manual processing, making them inadequate for addressing the substantial challenges introduced by the evolving workload and system architecture. 
We advocate the necessity of an artificial intelligence (AI)-guided approach to tackle the complex challenges of CI resource management. We envision that state-of-the-art AI technologies will play a pivotal role in delivering predictive insights and optimizing decision-making in the constantly evolving CI landscape, while humans can focus on critical tasks of oversight and validation. 
Keywords: advanced cyberinfrastructure; resource management; artificial intelligence; heterogeneous systems; science applications. 
Intellectual Merit 
In this project, we propose designing and evaluating AIMCI (Artificial Intelligence for Managing Cyber Infrastructure), an AI-guided framework for improving CI resource management in a computing facility where a diverse array of science applications are scheduled and executed across multiple computing systems. AIMCI not only accommodates various resource demands from different workloads but also enables us to design efficient resource management methods for improving he productivity of the facility. 
The project will (1) develop new AI models for predictive analysis of resource usage patterns and user behavior; (2) design intelligent strategies to optimize resource management in a complex and dynamically changing computing environment; and (3) build a discrete event-driven simulator for exploratory simulation of CI resource management with human-in-the-loop interaction. Together, these AI-guided components will enable us to address the pressing demands of diverse workloads, optimize resource allocation for heterogeneous computing, and enhance CI productivity. 
Broader Impacts 
This project will advance resource management research from a traditional, manual, static, and single-cluster approach to an automated, dynamic, and facility-wide one. It will introduce new technologies to effectively utilize CI for various applications in science and engineering, AI, and data analytics. The resulting AIMCI framework will have a significant impact on the field of high-performance computing (HPC) and benefit a broad spectrum of scientific domains that rely on HPC CI. The software developed in this project will be made public through an open source license, along with any datasets. UIC is a federally designated Minority-Serving Institution (MSI). An integrated education plan will be carried out to educate and train future workforce from K-12 to graduate level. The PIs are committed to promoting underrepresented participants in computing through student recruitment and outreach programs.

Overview of AIMCI with three research thrusts (T1-3)

OAC Core: AIMCI: Artificial Intelligence for Managing Cyberinfrastructure

We will evolve the national-scale Sage platform to build a new pathfinding testbed that links real-time AI-enabled sensors with existing NSF HPC and networking resources to provide unique cyber-infrastructure supporting research exploring new AI techniques, algorithms, and AI-driven climate understanding. The future national-scale computing continuum can use the testbed to gain critical insights and advance AI, as well as the coupled complex adaptive systems that makeup Earth&rsquo;s biosphere. 
The UIC team, led by Michael E. Papka, will collaboratively work with Northwestern University and other participating institutions to lead the following: 
i. Low-Cost Sage Node Development: UIC will oversee a team of undergraduate students who will research and develop low-cost Sage nodes for use in classroom settings. These nodes will be designed to provide accessible and affordable access to the Sage platform for educational purposes. Deliverables include designing and developing low-cost Sage nodes, documentation, and evaluating their performance and usability in classroom settings. 
ii. Jupyter Notebook Examples for Education and Outreach: The UIC team will lead the development of educational examples using Jupyter notebooks to support the project&rsquo;s education and outreach objectives. These examples will demonstrate the capabilities of the Sage platform and provide accessible learning resources for students and educators interested in AI techniques, algorithms, and AI-driven climate understanding. Deliverables include the creation of Jupyter notebook examples, user guides, and instructional materials to support their use in educational settings, as well as reports detailing the impact and effectiveness of these resources in achieving the project&rsquo;s education and outreach objectives.

NAIRR Pilot: A National-Scale Testbed Supporting Artificial Intelligence Research Spanning the Computing Continuum

High-Performance Computing research and development supports the Argonne Leadership Computing Facility (ALCF) supercomputer&rsquo;s use and operations, including: 
- Visualization and analysis of abstract datasets based on ALCF operational logs, visualization of workflow execution and data transfers, dashboard development for monitoring of systems, and integration of edge devices into HPC workflow.
- Development of code and algorithms that demonstrate the use of Intel oneAPI for the ALCF supercomputer Aurora.
- Evaluation of Unity and Unreal gaming engine visualizations for virtual reality environments from data generated using Argonne supercomputers.
- Development of HPC material focused on undergraduate education.
- Production of documentation and sample code as needed.
- Exploration of alternative administration environments to better support the deployment of science portals (Kubernetes).

High Performance Computing Research

EVL undergraduate researcher Athena Angara discussed her summer internship project with Argonne Leadership Computing Facility (ALCF), &ldquo;Tele-robotics to Tele-autonomous Robotics for Isotope Production,&rdquo; during NVIDIA&rsquo;s weekly livestream, Community Spotlight: Unlocking Physical AI with Isaac Sim and OpenUSD. 
Angara is a senior studying data science. She spent her summer at the ALCF working on a DOE-funded project that leveraged virtual reality and augmented reality to create digital twins of critical components of robotic arms. Using Unreal Engine 5 and NVIDIA&rsquo;s Unity, Omniverse, and Isaac Sim, Angara developed an immersive display system with real-time, low latency, and high spatial resolution for remote control processes. She also converted data to ensure accurate representations between these digital platforms. 
NVIDIA&rsquo;s Omniverse is a real-time 3D graphics collaboration platform. It has been used for applications in the visual effects and &ldquo;digital twin&rdquo; industrial simulation industries. Omniverse makes extensive use of the Universal Scene Description format. 
To see more about the ALCF summer program see <a href="https://www.alcf.anl.gov/news/summer-students-tackle-supercomputing-and-ai-research-projects">&ldquo;Summer students tackle supercomputing and AI research projects&rdquo;</a> 
See the <a href="https://youtu.be/3jM1mjGqZ0Q?si=JWoxznDdeGTAv88H">YouTube video</a> of Angara&rsquo;s project.

&ldquo;Tele-robotics to Tele-autonomous Robotics for Isotope Production&rdquo; Presented at NVIDIA&rsquo;s Livestream

Om Patel

Michael Cortez

Efficient job scheduling is crucial in high-performance computing (HPC), balancing user demands for quick job turnaround with facility goals for high resource utilization. Traditional scheduling requires users to specify a system at job submission, which can lead to inefficiencies. A unified scheduling approach, viewing the resources within a computing facility as an integrated pool, promises improved resource use and reduced job wait times. This paper presents CQSim+, an open-source, discrete event-driven simulator tailored for symbiotic multi-resource scheduling. CQSim+ supports dynamic simulation by continuously integrating real-time data from job schedulers, enabling adaptive scheduling based on the system&rsquo;s current state. Through extensive experimentation, we demonstrate CQSim+&rsquo;s ability to enhance resource utilization and decrease job wait times in both homogeneous and heterogeneous HPC environments. 
Additionally, we present a case study that coordinates job scheduling between two production systems, illustrating how CQSim+ can effectively optimize job scheduling across distinct systems. 
CCS Concepts: Computing methodologies - Modeling methodologies; Discrete-event simulation; Real-time simulation; Simulation tools; Social and professional topics - System management 
Keywords: Multi-resource scheduling, Simulation tool, Resource management systems, High-performance computing 
<a href="https://doi.org/10.1145/3726301.3728404">https://doi.org/10.1145/3726301.3728404</a>

A time diagram of CQSim+ performing meta-scheduling for multiple systems.

CQSim+: Symbiotic Simulation for Multi-Resource Scheduling in High-Performance Computing

MPI_Alltoallv generalizes the uniform all-to-all communication (MPI_Alltoall) by enabling the exchange of data-blocks of varied sizes among processes. This function plays a crucial role in facilitating many computational tasks, such as FFT calculations and graph mining operations. Popular MPI libraries, such as MPICH and OpenMPI, implement MPI_Alltoall using a combination of linear and logarithmic algorithms. However, MPI_Alltoallv typically relies only on variations of linear algorithms, missing the benefits of logarithmic approaches. Furthermore, current algorithms also overlook the intricacies of modern HPC system architectures, such as the significant performance gap between intra-node (local) and inter-node (global) communication. To address these problems, this paper presents two novel algorithms: Parameterized Logarithmic non-uniform All-to-all (ParLogNa) and Parameterized Linear nonuniform All-to-all (ParLinNa). ParLogNa is a tunable logarithmic time algorithm for non-uniform all-to-all, and ParLinNa is a hierarchical and tunable near-linear-time algorithm for non-uniform all-to-all. These algorithms efficiently address the trade-off between bandwidth maximization and latency minimization that existing implementations struggle to optimize. We show a performance improvement over the state-of-the-art implementations by factors of 42x and 138x on Polaris and Fugaku, respectively. 
<a href="https://doi.org/10.1145/3731545.3731590">https://doi.org/10.1145/3731545.3731590</a>

An example of ParLinNa when P = 15, N = 3, r = 2 and Q = 5.

Parameterized Algorithms for Non-uniform All-to-all

Motivation: 
- The Large Hadron Collider (LHC) has created hundreds of petabytes of data. The High-Luminosity LHC (HL-LHC) will produce orders of magnitude more.
The HL-LHC is a major upgrade to the LHC, aiming to be operational by the end of 2030. It will produce more data in one year than the LHC produced during its first 10 years of operation. 
- Significant R&D efforts are needed to handle the incoming tide of data. Without using new technologies, the data volume of the HL-LHC will rapidly outgrow CERN&rsquo;s compute and storage capacities. This means everything is on the table, even historically disfavored methods like lossy compression.

﻿﻿Evaluating SZ3 Compressor Performance on High Energy Physics Data (poster)

In recent years, numerous hardware accelerators have been developed to meet the rising demand for machine learning (ML) workloads, and Large Language Models inference in particular. GPUs are currently the standard for ML training and inference. However, they require substantial data movements that hurt performance and increase power consumption, making systems extremely energy-intensive. In response, many companies, including Intel, AMD, and Google, as well as numerous startups such as Groq, SambaNova, Cerebras, and Graphcore, have introduced specialized accelerators for ML workloads that leverage a dataflow design, which aims to reduce data movement and thus improve both performance and power consumption. This article presents a comparative analysis of the performance and energy efficiency of various AI accelerators and GPUs for large language model (LLM) inference, using popular open-source models evaluated on both synthetic and real-world datasets.

Energy Efficiency of LLM Inference Across Various AI Accelerators (poster)

While large language models (LLMs) are increasingly used for generating parallel scientific code, most current efforts emphasize functional correctness, often overlooking performance and energy considerations. In this work, we propose LASSI-EE, an automated LLM-based refactoring framework that generates energy-efficient parallel code on a target parallel system for a given parallel code as input. 
Through a multi-stage, iterative pipeline process, LASSI-EE achieved an average energy reduction of 47% across 85% of the 20 HeCBench benchmarks tested on NVIDIA A100 GPUs. Our findings demonstrate the broader potential of LLMs, not only for generating correct code but also for enabling energy-aware programming. We also address key insights and limitations within the framework, offering valuable guidance for future improvements.

LASSI-EE: Leveraging LLMs to Automate Energy-Aware Refactoring of Parallel Scientific Codes (poster)

HEP workflows often consist of complex DAGs with streaming job arrivals that challenge traditional static scheduling approaches. We introduce a dynamic pilot job scheduling model that allocates HPC resources on demand when queued DAG jobs reach a configurable threshold. The simulation framework models interactions between the trace parser, DAG manager, pilot job manager, and HPC resources. The system improves scheduling responsiveness and resource utilization by reserving nodes as needed and scheduling jobs in batches based on DAG dependencies.

On-Demand Pilot Job Scheduling for Adaptive Execution of HEP Workflows on HPC Systems (poster)

Motivation: 
•	Parallel Discrete Event Simulations (PDES) offers accurate HPC simulations but is computationally intensive and slow to scale. 
•	Surrogate models can accelerate simulations, and we explore if causal insights can improve their long-term forecasting stability. 
Research Questions: 
•	Can Causal Signals hidden in HPC simulations unlock better forecasting? 
•	Can different Causal Discovery methods identify key drivers for accurate surrogate forecasting?

Temporal Causal Graph Discovery  in Complex HPC Network Traffic Simulations (poster)

Motivation: 
- Dragonfly networks suffer from inter-job interference due to shared global links. 
- Existing routing algorithms fail to account for system-wide job patterns and link congestion. 
- Modern HPC systems offer flexible job placement, allowing users to specify workload locations for optimized performance.

Preventing Workload Interference with Intelligent Routing and Flexible Job Placement Strategy on Dragonfly system (poster)

High-performance computing (HPC) job scheduling has seen promising advances with Deep Reinforcement Learning (DRL). However, challenges such as low interpretability, instability, and high computational cost hinder DRL&rsquo;s practical adoption. We explore a model-based alternative using Monte Carlo Tree Search (MCTS) to overcome these limitations. By leveraging existing HPC simulators as models for MCTS and focusing on transparent decision-making, we aim to develop a scalable and interpretable scheduling solution fit for real-world deployment.

Overcoming RL Limitations in HPC Scheduling: A Model-Based MCTS Approach for Practical Deployment (poster)

Large language model (LLM) requests vary widely: a short factual query may produce only a few dozen tokens, while multi-step reasoning or proof generation can run to tens of thousands. Because generation is autoregressive, inference time grows proportionally with output length. Current serving stacks such as vLLM and Orca schedule requests using a simple first-come-first-serve (FCFS) policy. While fair, FCFS suffers from a well-known issue in LLM serving called head-of-line blocking, where a single long request delays all shorter ones behind it - leading to inflated tail latency and wasted throughput. A classical remedy is the Shortest-Job-First (SJF) policy, which improves efficiency by serving shorter jobs first. However, SJF requires knowledge of job length in advance - information that is unavailable in LLMs until generation completes, due to their autoregressive nature. 
We introduce a prompt-sensitive LLM task scheduler based on pairwise learning method from learning-to-rank utilizing Margin Ranking Loss and show that our method significantly improves scheduling performance with minimal overhead.

From Prompts to Priorities: Pairwise Learning-to-Rank Scheduling for Low-Latency  LLM Serving (poster)

Prior uncore frequency tuning studies have primarily focused on conventional HPC workloads running on CPU-only systems. As HPC advances toward heterogeneous computing, integrating diverse GPU workloads on heterogeneous CPU-GPU systems, it becomes imperative to revisit and enhance uncore scaling. Our investigation reveals that uncore frequency scales down only when CPU power approaches its TDP (Thermal Design Power) -- a rare scenario in GPU-dominant applications -- resulting in unnecessary power waste in modern computing systems.

MAGUS: Memory-Throughput-Based Uncore Frequency Scaling on Heterogeneous Systems (poster)

AllReduce is a critical collective in both HPC and large-scale AI workloads. However, scaling it to Exascale systems presents key challenges due to inter-node communication bottlenecks and underutilization of intra-node resources like shared memory and NVLink. This work analyzes state-of-the-art AllReduce algorithms to identify inefficiencies and opportunities for hybrid strategies that explicitly separate intra- and inter-node communication. 
We introduce a preliminary algorithmic design that leverages tunable intra-node communication patterns and discuss key performance criteria, including message count and data volume. Our early results provide insight into communication trade-offs and guide the development of adaptive AllReduce implementations optimized for Exascale systems.

EMMA: Efficient Multi-node Memory-aware AllReduce Algorithms (poster)

Non-uniform MPI_Alltoallv communication is critical in many high-performance computing (HPC) applications where data exchange patterns vary significantly between processes. However, existing MPI implementations rely on fixed heuristics that are not well-suited for dynamic, irregular workloads, often leading to suboptimal performance. Furthermore, there are customized non-uniform All-to-all functions having performance advantage compared to official MPI Alltoallv, under certain scenarios. Creating a framework which can accurately select the optimal algorithm will be extremely beneficial, but challenging due to its potential complexity. 
Currently, there are research works regarding the data-driven approach for tuning MPI functions, but most of them they focus on MPI functions such as MPI_Scatter, MPI_Reduce, MPI_gather, uniform MPI_Alltoall Non-uniform Alltoall is still an under-developed area. One challenge of developing a tuning framework for non-uniform Alltoall is that the blocksizes between processes are not fixed and it introduces significant additional complexity to the tuning workflow.

Performance Characterization and Tuning of Non-uniform All-to-all Data Exchanges (poster)

Research Contributions: 
- First ever Datalog engine designed for multinode multi-GPU HPC systems, outperforming state-of-the-art shared-memory, distributed-memory, and GPU-based engines 
- Introduces novel GPU-Aware communication and buffer preparation strategies for scalable recursive query evaluation 
- Supports recursive aggregation for Datalog rules using high-throughput GPU kernels

High Performance Computing

Bibliography

pubs (31)

Related Topics

people (35)

events (22)

research (7)

Contact