Skip to main content

IU expands Jetstream cloud for increased access, deep learning

Jetstream, led by the Indiana University Pervasive Technology Institute, was the first production cloud environment funded by the National Science Foundation (NSF).

David Hancock, Jetstream Principal Investigator
David Hancock, Jetstream Principal Investigator

Over the last few years, it became clear that IU researchers, in particular, wanted more access to Jetstream’s resources, including expanded access to deep learning technology. In mid-2019, IU met these needs by expanding the Jetstream system hosted in Bloomington, IN to allow for additional discretionary capacity and GPU pilot projects. The system’s single rack expansion added 6 storage servers (288 TB), 16 larger memory compute nodes (4.1 TiB of RAM), and 6 nodes equipped with NVLink-enabled NVIDIA V100 GPUs. This expansion has grown the pool of discretionary resources by another 5%, which means, in a practical sense, Jetstream can support even more projects. 

The expansion has already made a number of new projects possible. IU researchers recently used Jetstream to produce the first SPEC ACCEL results using virtual graphics processing units (GPUs). Researchers used the 6 GPU-equipped nodes mentioned above as a testbed for evaluating the performance and operations of virtualized GPUs.

Soon, these expanded resources will also support a massive open online course (MOOC) on data science with over 600 registered students for the Cyberinfrastructure for Network Science Center. The course provides an overview of the state of the art in information visualization, teaching the process of producing effective visualizations that take user needs into account.

With a focus on ease of use and broad accessibility, Jetstream is designed for researchers who haven’t had much, or any, experience with high performance computing and software resources. This mission, in particular, fosters twenty-first century workforce development at small colleges and universities, including historically black colleges and universities, minority serving institutions, tribal colleges, and higher education institutions in EPSCoR states.

The project, now in its fourth year, and has been awarded over $13M to date, with operations funded through November 2020. Jetstream’s system is rife with practical resources, including a library of preconfigured virtual machines, on-demand interactive computing and analysis, persistent services like gateways, and programmable cyberinfrastructure. 

The Jetstream environment continues to push the boundaries of traditional NSF-funded cyberinfrastructure (CI) systems with programmable cyberinfrastructure serving over 2200 students and tens of thousands of users via 27 hosted science gateways and direct access through the Jetstream user portal (leveraging CyVerse’s Atmosphere) and multiple APIs (the Atmosphere CLI and OpenStack).