Welcome to GOOD 2026!
Discover the Appverse—a growing, community-driven catalog of Open OnDemand app configurations designed to make app deployment across HPC environments more consistent, reusable, and efficient.
Open OnDemand enables the execution of passenger applications, which can be utilized for self-paced learning and andragogy through GenAI-powered tools. The GenAI tutor paradigm can now be accessible on demand, within the same platform, to support a variety of user needs as soon as they arise. Open OnDemand passenger applications could enable guidance on basic coding, instructions for submitting jobs with different schedulers, "chatting" with available documentation, and accessing vital information via microservices such as resource availability, job-specific information, and efficiency reports.
We have streamlined LLM access on Tufts University’s HPC Cluster through Open OnDemand (OOD), enhancing AI use in research, teaching, and learning. We present a standardized toolkit of OOD applications: Research Chatbot (OpenWebUI + Ollama), LLM Notebooks (Jupyter + Ollama) and AI Image Generators (StableDiffusion, ComfyUI). Participants will gain insights into improving LLM accessibility via the OOD interface, comparing local and cloud based approaches for compute, aligning LLM tools with research needs, and lessons learned from deployment. Code, slides, and documentation will be available via GitHub.
We share how we built LLM inference services for researchers at Ohio Supercomputer Center (OSC) using Open OnDemand. Our users need LLMs in many different ways, so we support several service models. We started with an Ollama module on the cluster, added OpenWebUI, and later created dedicated OnDemand apps that start the LLM server and give users a simple interface. We also provide a Jupyter + Ollama app that exposes an OpenAI API interface in a Jupyter kernel. We maintain shared models so users can start quickly, while still allowing local models. We have also deployed the Ollama + OpenWebUI service on our Kubernetes cluster, but scaling it for a large number of users remains a challenge. We plan to add additional LLM engines and front-ends as part of our future improvements as well.
During a two-semester internship with Tufts Technology Services, interns Javier Laveaga and William Goldman developed new Open OnDemand applications for displaying cluster statistics and user job data, and improved both the functionality and appearance of the Tufts High-Performance Cluster dashboard. In this talk, we share our experiences building custom apps and dashboard components, the final results, and the programming languages and structure used. We explain how to create fully customizable user-facing applications in Open OnDemand.
Imagine having access to a centrally hosted environment that combines the power of HPC systems with the convenience of desktops. This is the HPC Desktop, designed to lower the barrier of entry and broaden adoption of HPC systems.
The GOOD25 conference served as a significant platform for advancing knowledge beyond the fundamentals of Open OnDemand. Numerous presenters introduced tools and strategies aimed at extending the capabilities of the OnDemand portal and enhancing the overall user experience within high-performance computing (HPC) environments. This presentation will analyze several of the most influential approaches highlighted during the event and discuss their successful integration into our center’s operations over the past year.
Explore how companion apps featured in the Appverse can extend the power of your Open OnDemand deployment.
Meet with Open OnDemand Dev team members to ask questions about the platform and the docs that don’t easily lend themselves to discourse or email. Community members are welcome to chime in on topics outside the scope of what is deployed at Ohio Supercomputer Center.
This BoF brings together users and resource providers that are interested in running interactive graphical applications in Open OnDemand. This can be web applications like Jupyter Notebooks or desktop applications like RStudio, Matlab or Visual Studio Code.
Grab a snack, meet the community, and connect with Governance Committee members, the Appverse team, Affinity Groups, and GOODLUCK leadership to learn how to get involved with Open OnDemand project!
This Birds-of-a-Feather (BoF) session is for GOOD attendees interested in Open OnDemand (OOD) for instructional use. With rising demand for centralized computing in education, OOD’s web interface helps instructors and students access high-end resources easily. The BoF will feature presentations on OOD customizations for teaching, interactive demos, and Q&A, fostering community discussion and sharing of adaptable solutions. Attendees will connect with peers and learn techniques for classroom and broader OOD use.
Invited Speakers:
- Martin Cuma [University of Utah]
- Sean Anderson [Wake Forest University]
A few informal meetup spots have been suggested for this evening. These are not official conference events, just casual places where GOOD attendees may choose to gather.
A core leadership update on recent Open OnDemand updates, new cross-cutting solutions, update on teh OOD affinity Groups, and more.
Executing multi-task computational research workflows on HPC systems often requires manual scripting, resource manager and scheduler expertise. It creates barriers for researchers and educators, especially at smaller institutions or in multidisciplinary collaborations. Although OOD provides an accessible web-based interface for launching standalone applications as jobs, it has lacked an integrated mechanism to easily compose, manage, and execute multiple tasks with dependencies among them as Workflows. Workflows enabled researchers to visually construct, execute, and monitor simple composition of multiple “launchers” (i.e. independent tasks/jobs) that can perform data pre- or post-processing, simulation, and other computational tasks connected via output to input dependencies.
Since GOOD 2025, more HPC Centers have been adopting Open OnDemand. Many sites have integrated MATLAB. Learn what’s new and which tools sites are offering, including:
- Browser-based access to full MATLAB, including GPU and parallel computing tools
- MATLAB Copilot for GenAI-assisted coding and simulation in cloud and desktop
- MATLAB MCP Core Server for integrating MATLAB with agentic AIs
As HPC expands beyond computer science, many users encounter command-line interfaces for the first time. HPCMosaic addresses this gap with a modular OOD dashboard that simplifies cluster management for non-technical researchers while staying customizable for experienced users. This talk presents a framework that reduces support burden by enabling self-service for common tasks (file quotas, account management, and job control) decreasing repetitive tickets. Targeted at HPC developers and cluster admins, this session covers modern dashboard development focused on customizability and user experience. Attendees will learn how to build accessible HPC interfaces that let researchers focus on their work instead of troubleshooting, reducing user frustration and support workload.
At CESNET we have integrated OpenStack with Open OnDemand, enabling users to create and deploy virtual machines, Kubernetes clusters and applications with minimal effort. In this presentation we will demonstrate the full potential of this integration through two concrete use‑cases: (1) provisioning a simple OpenStack virtual machine and (2) launching an on‑premises Kubernetes deployment. The talk will highlight the underlying adapter architecture, share technical insights from the implementation, and discuss current limitations and challenges. The goal is to provide an overview for any organization that provides OpenStack to its users and wishes to support this functionality.
Researchers rely on Open OnDemand (OOD) to design, launch, and manage scientific workflows on high performance computing (HPC) systems. In addition to job creation and submission, this process also includes monitoring job progress. OOD provides the job-listing app and session cards for monitoring running jobs. However, these tools primarily provide scheduler-level information, forcing researchers to manually retrieve workflow and domain-specific information. Drona Workflow Engine bridges this gap by providing integrated, domain-specific monitoring throughout the workflow lifecycle. In this presentation, we will focus on job-monitoring capabilities and explore various types of job monitoring integrated into the representative, commonly used scientific workflows.
User impersonation is a powerful capability within Open OnDemand that enables designated users—such as administrators, support staff, or service accounts—to act on behalf of other users to troubleshoot issues, validate configurations, and streamline workflows. As HPC centers grow more complex and support demands increase, a flexible and secure impersonation model becomes essential.
Open OnDemand (OOD) provides researchers with a streamlined, browser-based HPC experience. This session covers deploying and operating OOD on Azure CycleCloud Workspace for Slurm, including provisioning the OOD VM, integrating Microsoft Entra ID with OpenID Connect and federated credentials, and auto-registering Slurm clusters for dynamic partition discovery. We'll demo browser-based VS Code for interactive development, explore when to use Cendio ThinLinc for enhanced visualization, and show how AI Agents like GitHub Copilot inside VS Code can rapidly build new OOD apps and templates.
As AI, HPC, and data-intensive research workloads continue to grow, many institutions face a common challenge: how to modernize research computing environments without replacing existing infrastructure, expanding staff, or locking into proprietary platforms. This talk presents a real-world case study of deploying projectEUREKA! with integrated Open OnDemand, an AI and HPC orchestration platform on Kubernetes, using a repurposed on-premises HPC cluster at Texas A&M University–San Antonio (TAMUSA).
We describe how Kubernetes was used as a local orchestration layer, enabling flexible, project-centric workflows across resources, while providing a browser-based experience. Demonstrates how institutions can extend the life of existing hardware, simplify user access, and support modern workflows.
The Multitenant Apps framework was developed for supporting LLMs, databases, and other services on traditional, job-based HPC infrastructure through Open OnDemand (OOD). It allows for controlled and secure sharing of these services between select users, and can greatly reduce hardware overhead since users share the same resources. It is also an effective method for delivering content to users within the OOD interface, which is especially useful within classrooms, research groups, and even departments. Featuring an interactive demo!
The EuroHPC Federation Platform is a project to federate access to EuroHPC systems, to make it
easier for researchers and industry users across Europe to access computing resources. Within
the project, a federated instance of Open OnDemand has been developed and deployed. It has many
unique features and integrations, such as remote cluster file system access, Kubernetes-native
deployment, and integration of the EESSI software stack.
This talk targets system administrators, service owners, and others interested in federation of HPC
resources or responsible for maintaining and developing Open OnDemand instances. It discusses
technical details and challenges of providing a single instance of Open OnDemand which is
connected to multiple supercomputers at external organizations.
HPC faces a dual challenge of supporting AI/ML engineers alongside domain-specific researchers who may lack Linux expertise. This presentation explores leveraging OOD as a unified gateway for cancer research and AI/ML pipelines. We detail our technical strategy of configuring OOD interactive applications to launch and manage Apptainer containers. This strategy delivers portable, reproducible, and version-controlled software stacks - from GPU-accelerated PyTorch environments to GUI-based biomedical visualization tools, effectively democratizing access to HPC by abstracting the complexity of the container runtime and Slurm. We share how this approach provides a low entry barrier for oncologists/biologists, environmental consistency for AI experts, and reduces support burden.
Open OnDemand and Globus are natural partners: Open OnDemand lowers the barrier to using advanced computing resources while Globus adds managed services to simplify data transfer, distributed computation and workflow automation. Using Globus to move and share data, provide high performance remote function execution and at-scale automation, makes your Open OnDemand system even more valuable to researchers. In this talk we will demonstrate how the two systems integrate to help researchers curate data from inception to publication.
Kubernetes has emerged as a de facto standard for deploying distributed systems and complex, multi-service web applications, and is gaining popularity in HPC environments. Given its demonstrated utility for research computing, how can we leverage Kubernetes for orchestrating Open OnDemand (OOD) deployments? In this talk, we present a Kubernetes operator we developed for deploying and managing OOD inside a Kubernetes cluster. We demonstrate how our operator can be used with a pilot implementation of a Kubernetes-native OOD architecture that allows for network available PUNs and advanced scheduling. Finally, we’ll close with a discussion of the pros, cons, and potential use cases for the KROOD approach to OOD management.
Meet with Open OnDemand Dev team members to ask questions about the platform and the docs that don’t easily lend themselves to discourse or email. Community members are welcome to chime in on topics outside the scope of what is deployed at Ohio Supercomputer Center.
This session will introduce ThinLinc and how to use it together with Open OnDemand. ThinLinc is a Linux terminal server solution that is used at HPC centers around the world to support access to interactive graphical Linux applications. The ThinLinc team is working with the OOD team to integrate ThinLinc into OOD and this session will go over the technical work as well as the features of the integration. This session will also include a demo of the current beta release.
LAMMPS's complex, non-uniform GPU feature support creates a dilemma: users struggle to find a compatible build, leading to wasted time, admin overload, and underutilized GPUs (e.g., Intel Max). We present an intelligent, wizard-style utility that performs automatic three-way matchmaking between available LAMMPS software builds, GPU accelerators, and user input requirements. The utility suggests optimal, GPU-accelerated builds and clearly explains why others are incompatible. Built atop the Drona job launcher, the wizard uses dynamically repopulating menus to guide users through the selection process, and finally injects the full configuration directly into a ready-to-run Slurm job script.
Large language models are now a common tool for writing code and exploring ideas, but using them on HPC systems can still be a challenge. To make this simpler, I built an interactive Open OnDemand application that lets users access a local LLM directly inside Jupyter Notebook using the Jupyter AI extension. Behind the scenes, user requests are routed through a load balancer to a pool of vLLM inference servers running on Intel GPUs, which serve open-source models up to 70B parameters.
In this talk, I’ll walk through how the system works, why we built it, and how it gives researchers easy, reliable access to LLM assistance without relying on external cloud services.
Machine learning now plays a very important role in a wide range of scientific fields. However many researchers in these scientific fields don't have machine learning expertise and may find it difficult to train a machine learning model themselves.
Auto machine learning bridges this gap by making it very simple to train a state of the art model. Auto machine learning frameworks like AutoGluon make it such that all you need is a dataset and a few lines of code.
The interactive application that I have developed for Open OnDemand closes the gap further by not requiring the user to write any code at all. Through our application, users can simply choose a dataset and their configuration, and train the model.
We are excited to share insights from the inaugural OnDemand internship hosted by Research Technology at Tufts Technology Services. This presentation highlights the interns’ accomplishments, explores challenges encountered, and identifies opportunities uncovered. We look forward to sharing our journey and lessons learned with fellow HPC centers.
CryoSPARC is a leading cryo‑EM data‑processing platform. While CryoSPARC offers various installation methods, each can pose significant challenges for HPC users and system administrators ranging from version updates to port management.
To streamline deployment, we have created a CryoSPARC interactive app and a script for generating a Singularity image that eliminates the need for end‑users to perform any manual installation. This OOD app utilizes a CryoSPARC Singularity image that is shared among all users while providing each user with their own individual database and log files. This dramatically reduces storage overhead and simplifies version control, while providing each researcher with a fully functional CryoSPARC environment.
High-performance computing allows researchers across the globe to conduct research at scale. At Purdue University, Open OnDemand provides an easy way for users to access cluster resources and launch popular applications with minimal barriers. In order to improve how users interact with clusters, we are building a user-friendly, modular web dashboard on Open OnDemand with detailed, organized cluster information and color-coded visualizations. By combining Slurm-based data sources and modular widgets and components, this dashboard allows researchers to access cluster information about their jobs and allocations easily and provides a modular, extensible framework to install on any Slurm-based cluster.
Join us at Red Butte Garden, just a mile from the conference center, for an evening of networking and relaxation. Shuttles will be running between the conference/hotel and the Red Butte from 5:30 to 9:30. Shuttle pick up will be right in front of the conference center and hotel.
This talk goes through the history of the Open OnDemand snap, a single package for running Open OnDemand on any Linux distribution.
The talk covers four sections:
- The Good; successes such as automatic backup/restore.
- The Bad; challenges like bundling Passenger, HTTPD, and NGINX together.
- The Ugly; implementing ad-hoc workarounds like a custom portal generator and custom Lua packages for administering Open OnDemand inside the snap runtime.
- What’s Next?
In the “What’s Next?” section, I'll present the Ubuntu HPC community's plans to integrate the snap into Charmed HPC to be the default user frontend. I will highlight blockers, struggles, hopeful contributions to Open OnDemand, and areas where we need community assistance.
Meet with Open OnDemand Dev team members to ask questions about the platform and the docs that don’t easily lend themselves to discourse or email. Community members are welcome to chime in on topics outside the scope of what is deployed at Ohio Supercomputer Center.