Navigating an ocean of computational environments: understanding researcher needs in selecting and building virtual machine and container images
TimeTuesday, July 306:30pm - 8:30pm
LocationCrystal Foyer and Crystal B
DescriptionMany research computing services are seeing rapid growth in the use of virtual machine images (VM images) and container images to support research computation, applications, and workflows (Belmann, 2015). These images are used across a range of compute models. Researcher interest in VMs and containers is driven by a wide range of considerations, including need to provision and run VM images when using cloud resources; portability of environments to multiple computational resources or platforms on which researchers work; efficiency in reproducing and sharing environments that are proven and configured for a researcher’s workflow; and citability in descriptions of research methodology. Researchers can be significantly supported by a maintenance regime that minimizes time and attention they must pay OS and application maintenance tasks, and by mechanisms that ease discovery of images best suited to their need. From a researcher’s perspective, an initial step in their research workflow is the discovery and tailored configuration of reliable, preferably vetted images, from which to select an appropriate fully-functional or substrate image for their computational research (by “substrate image” we mean an image to which project-specific software can be added with reasonable ease).
To begin to understand how researchers undertake the discovery of such images, a small group of UC Berkeley researchers were interviewed to elicit an overview of their computational research needs and practices, as well as past or prospective characteristics or features they do or expect to look for when choosing or augmenting a VM or container in which to run computational research.