In the last post, we learned about a project that has expanded the Arbutus research computing cloud site. This project began in 2017 and has seen the installation of more storage, compute, management and control nodes to bring the Arbutus ARC system to full specification: 20,000+ virtual cores and 3.5 PB of storage. These are some impressive specifications, but what can a researcher do with this kind of processing power? It turns out the system has unlocked some pretty incredible capabilities for researchers across Canada and the world.
When it comes to research computing, it is important to understand the two types of workload the Arbutus ARC system can perform: persistent and compute. Persistent workload refers to tasks that run 24 hours a day, 7 days a week for months or years. A typical persistent use case is hosting a web server or database server to provide information to users at all times in the form of an online service or web portal. Although persistent tasks do not usually require as many cores as compute and their processing is usually very limited, they offer great value as they continuously provide a service to users. Compute workloads are usually ephemeral processes which are memory- and/or storage-intensive and tend to start and finish quickly. These jobs can make use of more cores, up to 1400, and usually take several hours to several days to complete.
Both of these functions support incredible research across Canada and offer novel insights right at our fingertips. Below are some local projects currently hosted in the Arbutus system and information about their impact on the Canadian research landscape.
SPH Project
The Smooth Particle Hydrodynamics (SPH) project is a persistent workload in the Arbutus system. The computation method for the project is hosted in the system, as is a web server to share the results of the project. The SPH project uses an in-house parallel Smooth Particle Hydrodynamic computation method that can be used to simulate and support the solutions to fluid dynamic challenges. Fluid dynamic challenges are those that rely on understanding the movement behaviour of liquids and gases for a solution. With SPH, researchers can better determine the scope of impact expected from a variety of events, from sedimentation in a river, to a tsunami. By modeling these processes, as well as those of dam-breaks and landslides, more accurate information is made available to support effective emergency preparation and disaster management interventions. Beyond emergency preparedness, the SPH method can be applied to renewable energy research, more specifically by modelling ocean wave interactions with wave energy convertors. To learn more about this project, visit the SPH project site.
Linked Modernisms
Another interesting persistent project on the Arbutus system is the Linked Modernisms project. This project is led by Prof. Stephen Ross and is using machine learning to create more powerful database searches that present large volumes of complex content in a more digestible way. Linked Modernisms presents search results from the modernism encyclopaedia as a network graph, linking vast stores of information about authors, their relationships and influences to build a more intuitive story of these important figures. This project brings the layers of prolific literary figures throughout history to life, even offering information about their degrees of separation from one another. This project uses the Resource Description Framework (RDF) metadata model to link search results with other entries in the encyclopedia in a way that allows users to build a narrative about the lives of these figures, and offers a more relatable framework for the information they seek. The project relies on the virtuoso platform for managing, storing and presenting the network graph.
Z Axis
Another interesting persistent workload on Arbutus is the Z Axis project, led also by Prof. Stephen Ross. This project creates a three-dimensional geospatial map of any story and highlights locations mentioned in the story’s text as bumps on a map. The more frequently a location-based word, such as the “left bank”, appears in a given text, the higher the peak at this area on the map. The project can perform a similar function with emotion-based words and map the location of negative and positive sentiment language that is used in a story. Users of this system have uploaded texts and an accompanying historically accurate map to perform analysis on it in real time (about thirty seconds, give or take). This research is valuable because it allows researchers to develop new perspectives of stories, using the language in the tales to examine the social context in which they were written and understand more about the time period of these stories. This research could be used as the first step in the cinematic process since it offers users guidance on where various elements of a story took place in the real world and improve the accuracy of visual storytelling. More colloquially, imagine building the itinerary of your next vacation based on your favourite story set in that place!
These projects are just the beginning of the incredible insights the Arbutus ARC cloud will be able to generate for researchers in Canada and across the world. As the capacity of the system grows through Systems projects like the most recent expansion, more national and international research projects will be able to keep their data in Canada and contribute to research innovation.