Harnessing the Computing Continuum for Programming Our World

Abstract

The number of network-connected devices (sensors, actuators, instruments, computers, and data stores) now substantially exceeds the number of humans on this planet. This is a tipping point, and the societal and intellectual effects of this are not yet fully understood. Billions of things that sense, think, and act are now connected to a planet-spanning network of cloud and high-performance computing (HPC) centers that contain more computers than the entire Internet did just a few years ago. We are now critically dependent on this expanding network for our communications and social discourse; our food, health, and safety; our manufacturing, transportation, and logistics; and our creative and intellectual endeavors, including research and technical innovation. Despite our increasing dependence on this massive, interconnected system of systems in nearly every aspect of our social, political, economic, and cultural lives, we lack ways to analyze its emergent properties, specify its operating constraints, or coordinate its behavior. Simply put, today we have the tools to instrument and embed intelligence in everything, and we are doing so at a prodigious pace. Although we are the globally distributed designers, builders, and users of this immense, multi-layered environment, we are not truly its masters. Each of us manages only some of the networks components, and we can neither predict its aggregate behavior nor easily specify our intensional goals in intuitive language. For all of us collectively, and each of us individually, this must change. Today, we program in the relatively small confines of a single node, defining individual device, instrument, and computing element behaviors, and we are regularly confounded by unanticipated outcomes and unexpected behavior that results once this individual node/device is exposed to the network collective. As consumers, we want our Internet-capable environmental devices (e.g., thermostats, lighting, and entertainment preferences) to adapt seamlessly to our changing roles and expectations, regardless of location. And yet, rather than specifying the ends we seek, we must specify detailed behaviors for home, office, car, and transient locale. In environmental health, we build and deploy arrays of wireless environmental sensors and edge devices when our goal may really be to “reprioritize edge resources to search for mosquitoes, given a statistically significant change in seasonal temperature and humidity across the nearby river basin”. In disaster planning, when satellites show hurricane formation, we manually redirect data streams and simulation software stacks, when our goal is really to “retarget advanced computing resources to predict storm surge levels along the eastern seaboard”. In science, when the Laser Interferometer Gravitational-Wave Observatory (LIGO) detects