Foster’s rule (also known as the island rule or the island effect) is an ecogeographical rule in evolutionary biology stating that members of a species get smaller or bigger depending on the resources available in the environment.

Source: https://en.wikipedia.org/wiki/Foster’s_rule

Cyberinfrastructure at the University of Hawaii is relatively new and small compared to the other organizations in ACI-REF. The evolving Hawaii team consists of four members and has only been in existence for about 2 years.  The team at Hawaii is part of a larger organization IT organization, UH Information Technology Services (ITS), which services all of the University of Hawaii system (3 Universities, and 7 Community Colleges).  Our team is part of a much larger organization that supports academic and enterprise computing, but since our focus is research computing much of our work is independant of the rest of ITS.  As a result, each team member is required to wear many different hats: ACI-REF, System administrator, software developer, procurement specialist, sales rep, user support.  With each team member having a wide breadth of skills, problem resolutions takes on many different forms.  Here are two different examples of how the diversity found within our small team yields varying results.

Full Time Bioinformatics Researcher —

A bioinformatician who was working with a pipeline that utilized the Spades assembler was encountering performance issues.  Upon closer inspection of the problem this researcher was experiencing two issues.  One issue was general performance issues with IO to and from our Lustre file system.  At different points in the researchers pipeline, applications would treat previously generated files as databases, resulting in many small random IO operations.  The other issue was SPAdes causing kernel panics when interacting with the Lustre client on each node.  The makers of SPAdes were able to reproduce the problem, but a fix to SPAdes would still leave the researcher with poor IO performance.  In order to allow this researcher to perform their work in an optimal manner, we worked with them to identify a hardware solution to the problem.  This ultimately lead us to installing and configuring SSD drives  in each of the nodes this researcher owned.  These SSD drives allowed the researcher to have a local fast scratch space. Removing the IO bottleneck he was experiencing.

Master Student in Oceanography —

A graduate student in Oceanography was working on a phytoplankton simulation for the Atlantic Ocean to better understand why there are distinct populations of phytoplankton in the different hemispheres. Her simulations were taking months to complete.  We sat down and did a code review with her and the PI and recommend some methods to parallelize her R simulation code to improve performance. After some back and forth review and suggestions the implemented optimizations resulted in the simulation runtime requiring only a few days.  In addition, she was able to run further simulations using different time scales with higher timestep granularity over the span of 100 simulation years.