Gateways 2019 has ended
Back To Schedule
Wednesday, September 25 • 1:00pm - 1:20pm
Instant On: Caching Simulation Results for Science Gateways

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

Powered by the HUBzero platform, nanoHUB is the science gateway built and operated by the Network for Computational Nanotechnology (NCN). Like many science gateways, nanoHUB offers a variety of content. Among all HUBzero hubs, nanoHUB is unique for its large catalog of simulation tools and its community of tool users. In 2018, nanoHUB saw 16,750 users execute more than 750,000 simulation jobs using some 600 simulation tools. The resources applied to computing these jobs totaled some 145,000 CPU hours.

While the CPU allocation is significant, what is arguably more significant is the “wall” time experienced by the users running the simulation. Our own internal studies have shown a relationship between usage and wall time. Tools that have a low expected wall time typically have the highest utilization. The bulk of nanoHUB Rappture tools typically execute jobs in the range of almost 0.0 seconds to the maximum allowed session time of 2 weeks. Across these jobs, the expected (median) wall time is approximately 17.0 seconds.

Starting in 2011, the combined efforts of the leadership teams of both nanoHUB and HUBzero were awarded an NSF grant for the “Instant On” project. This project invested in several strategies to reduce resource consumption and improve user experience by reducing the turn around time between submitting a simulation job and receiving the computed result. One of the strategies would invest in developing a system to re-use
simulation results when possible. This development ultimately became a part of the HUBzero middleware as a caching system. It is this caching system upon which the remainder of this paper will focus.

In Section 2, we will describe the design goals of the “Instant On” cache and highlight some of the implementation details and features. In Section 3, we will discuss the operation of the cache with respect to utility and economy and also some of the pitfalls both experienced and potential. Section 4 will present some future directions in which the cache is but one of several services built on top of the underlying archive of simulation results. We will conclude in Section 5 with an invitation for other science gateways to use “Instant
On” as part of their tool and workflow pipelines.

Wednesday September 25, 2019 1:00pm - 1:20pm PDT
Kon Tiki Room, Catamaran Resort 3999 Mission Boulevard, San Diego, California 92109