Highlights from GlobusWorld 2024
The Conference for Reimagining Research IT
June 11, 2024 | HPCwire
By Susan Tussy
The Globus user conference, now in its 22nd year, brought together over 180 researchers, system administrators, developers, and IT leaders from 55 top research computing centers, national labs, federal agencies, and universities from around the globe. The conference is designed to provide insights into new methods of managing research, storage, and compute at scale and build cutting-edge data portals and science gateways to advance scientific research.
After a morning of technical tutorials, Rachana Ananthakrishnan, executive director and head of products at Globus, kicked off the conference with the “State of the Globus World” keynote. She described how Globus, with reliable and secure hosted, persistent, scalable and resilient services, combined with local agents and plug-ins, delivers the services required to overcome many data management hurdles researchers face in today’s globally distributed IT environments. Ananthakrishnan spoke about the continuing evolution of the Globus data, compute, and automation ecosystems, highlighting the growing use of Gobus Platform-as-a-Service. In particular, as researchers increasingly need to share and reuse data, they can deploy the newly released Globus Static Portal Framework, eliminating the need for any infrastructure or programming expertise. Researchers will be able to create a single page application running in a user’s browser with a no-code solution, and stand up a data portal in minutes.
Globus co-founder Ian Foster completed the keynote session by focusing on new research directions and how we can accelerate scientific discovery through an accelerated scientific method. In a data-driven fourth paradigm of science, cycles increase in complexity and speed, but now scientists can automate the entire process. With AI, you can make sense of large amounts of data; with HPC, you can validate ideas through simulations; and with robotics, you can test hypotheses. Globus, as an automation platform for science, is the engine powering the cycle, as evidenced by the “self-driving labs” operating at Argonne National Laboratory. Ian went on to highlight a few of the projects he is working on, including the construction of Open Science foundation models like ChatGPT trained on scientific literature and data suitable for fine-tuning specializations for downstream scientific tasks, the formation of the trillion parameter consortium, and AuroraGPT, a foundation model trained on a variety of data sources.
This year’s invited speakers featured a presentation by Ben Brown, Director of the Facilities Division for Advanced Scientific Computing Research at the Department of Energy (DOE), on the Department of Energy’s Integrated Research Initiative (IRI). Brown made a compelling case for increasing collaboration among national facilities and illustrated the role that the broader research community can play in making the IRI vision a reality. The day concluded with a packed session of Globus Office Hours, where attendees met with Globus developers to address their questions about the service.
Day two kicked off with the second invited speaker, Greg Gunther, the Science Data Management Branch Chief at the US Geological Survey (USGS), who described a new collaboration between the Globus team and USGS to upgrade one of the Department of Interior’s premier research data repositories, ScienceBase. ScienceBase is a trusted digital repository that serves as a critical component of the data management, delivery, and analysis-ready services provided by USGS for publicly released datasets. The multi-year collaborative effort seeks to reimagine ScienceBase and overcome the challenges of managing thousands of yearly tera-scale data releases.
Aditya Tanikanti, a Computer Scientist from Argonne National Laboratory, gave one of the many engaging lightning talks and spoke about the remote triggering of large language models on HPC clusters using Globus Compute. Other lightning talks highlighted use cases on how Globus facilitates rapid data analysis, enables the automation of data management tasks, and simplifies the management of petabyte-scale datasets. In addition, Karl Kornel from Stanford University received the Globus 2024 Community award.
