Globus Announces Multi-User Support for Globus Compute
Chicago, Illinois—May 7, 2024— Globus, the de facto standard platform for research IT, announced multi-user support for Globus Compute, a service that enables reliable, scalable, and high performance remote function execution, and delivers the same “fire-and-forget” capabilities for computation as the Globus core platform does for data management.
Now, a Globus Compute agent can be installed on clusters, supercomputers and other multi-user computing environments, and configured to create a multi-user endpoint. This enables users to access remote computing resources using the function-as-a-service (FaaS) model, and allows administrators to implement suitable policies that control usage of the Globus Compute service on their system. Multi-user configurations abstract away complexities of the underlying system resources, eliminating the many bottlenecks users might encounter if attempting to configure the software on their own. For example, an administrator can provide access to a cluster’s scheduler without needing to provide shell access to the cluster. Administrators can also maintain fine grained control over which users can access the Globus Compute endpoint, and what those users can do.
The challenges for users to access remote compute resources are well known—and are compounded as distributed IT infrastructures and massive data sets become commonplace. With the introduction of this new Globus capability researchers can easily access compute resources across disparate domains in order to run their analyses, without concern for differences in system architecture or configuration.
“We believe Globus Compute changes the game for organizations creating federated, distributed workflows,” said Rachana Ananthakrishnan, Globus executive director. “Anyone can now run large-scale workloads on systems ideally configured for the job, using the FaaS approach that is commonplace on cloud infrastructures. From simulation and real-time analyses to ML model training and inference, researchers can access limitless computing resources in solving some of today’s largest problems.”
Multi-user Globus Compute endpoints are currently installed at several institutions, with one early user commenting: “Using Parsl and Globus Compute we will be able to run distributed scientific workflows, and use containers for consistency to run parallel jobs across diverse system architectures, whether they are in the cloud or on-premises.” He went on to say, “With the increased distribution of resources there is no one place with all the diverse computing capabilities that a workflow needs. We need to be able to schedule workloads on the resources where they run best.”
The Globus Compute service is free to use for educational and non profit organizations, with some limits on usage by non-subscribers. As with all other Globus services, usage by commercial organizations requires a subscription.