If anything, it seems that disasters such as wildfires, hurricanes, extreme flooding, earthquakes, tsunamis, winter weather conditions, health emergencies, and accidents are becoming more commonplace in the world. At the same time our technological capability, for instance, the ability to capture data and capacity of HPC machines, is growing at an astonishing rate. So it makes a lot of sense to look at combining HPC computational models with real-time data in order to aid in urgent decision making as a part of disaster response. Ultimately the goal here is to save lives and reduce economic loss, but whilst HPC has a long history of simulating disasters, it’s not really designed for the type of workload that is required to support emergency, urgent, decision making, such as fast, real-time acquisition of data and the ability to execute jobs within specific time constraints.
This idea of fusing HPC with real-time data for urgent decision making moves HPC well beyond traditional computational workloads and while the challenges are significant, if HPC can be proven as a tool in responding to these disasters, we think the impact for our community and the wider world is huge. As such we are running a workshop at SC19, called UrgentHPC, around using HPC for urgent decision making and submissions are currently open for research papers (deadline 15th August.) All submitted papers will be peer-reviewed and accepted papers published in IEEE TCHPC.
This workshop will run on Sunday afternoon, 17th of November, and we were motivated to organize this because we think that exploiting HPC in this manner requires expertise in a wide range of areas. So, bringing together people from across the community with a wide range of expertise is hugely valuable. For instance, dealing with real-time data, working with real-time constraints (generating results within specific time frames), and appropriate visualizations techniques enabling front-line responders to make the right choices the first time and every time, are some of the areas that must be considered to meet our end goal. But don’t just think this involves technical challenges either because there are also questions around policy, such as how batch systems can be utilized on our HPC machines in a manner more conducive to running urgent jobs in bounded time, that must be addressed.
As I say, submission for papers is currently open and we are after a wide range of paper topics, from success stories of using HPC and real-time data for urgent decision making, to existing case-studies, to challenges and their potential solutions. There is further information on the workshop website at www.urgenthpc.com and the ultimate goal of this activity is to further build up a community around leveraging HPC as an important tool in urgently responding to disasters and societal challenges.
Source: Nicholas Brown, University of Edinburgh