We’re excited to announce several increased compute resources are now available in GeoLab. These resources include scalable compute quota caps for intensive workflows, access to GPUs and larger server sizes, and access to a Dask gateway. These features can be enabled for individuals or defined groups of users on a case-by-case basis. This makes it possible to tailor the GeoLab experience to meet the needs of users and across a spectrum of use cases, from students just getting started with exploratory classroom exercises in scientific computing, to graduate students and advanced researchers executing large, compute-intensive workflows.

By enabling these resources on a by-request basis, we are able to ensure that we can limit excessive use of resources and continue to make the platform freely available to everyone. This added flexibility allows us to give users access to the resources that will accelerate their research while supporting the long-term sustainability of the platform.

Increased resources that can be enabled on request include:

• Access to larger server sizes: By default, GeoLab users have access to servers with up to 4 CPUs and 29 GB RAM. Larger server sizes, with up to 119 GB RAM and 15 CPUs, are available on a temporary basis for more compute-intensive projects.
• Access to GPU servers: Nvidia Tesla T4 servers with 16 GB RAM and 4 CPUs are available on request for those performing advanced machine-learning tasks or intensive image processing workflows. 
• Access to a Dask gateway: Dask is a powerful orchestration tool for executing parallel compute operations. Running Dask locally is available for all GeoLab users and allows a person to fully leverage all CPUs available in their instance size in a parallelized workflow. The Dask gateway, however, allows a user to connect to an auxiliary cluster (when available) and leverage more CPUs than are available in their core GeoLab server.
• Increased compute quotas: All GeoLab users start with a cap on the number of compute-hours they can use in a 30-day rolling window. This compute cap meets the needs of over 90% of our users; however, if you require more compute time, for example to develop course materials in GeoLab or to execute a more involved analysis you’ve been prototyping, you can request an override.

Access to any of these resources can be requested here or via the links in our docs. Please be prepared to prototype your work in GeoLab first and be able to justify why your workflow requires advanced access. 

These new capabilities are especially valuable for workshops, short courses, and training events. For example, administrators can right-size the compute resources available to groups of users, giving students the resources they need to complete hands-on exercises and computationally intensive workflows. Once the course concludes, participants seamlessly return to the standard resource allocation, with an option of extending their increased resource allocation if actively pursuing research that requires it. 

Whether you’re teaching a class, supporting a collaborative project, or conducting research, these resources provide GeoLab the flexibility to better support your work.

Questions? Check out the GeoLab documentation or reach out to us on our community forum!