Project Overview
The Information Technology (IT) industry has recently been estimated to have the same carbon footprint (that is, energy consumption) as the airline industry.
Airlines have invested heavily for decades in more efficient engines, lighter airplanes and optimized scheduling to save energy consumption.
Meanwhile, the energy usage per compute server rack has grown from about 2 KW/rack in 2000 to an estimated 30 KW/rack in 2010, to the point where the cooling and power issues are now a major factor in system design.
In the last several years, the IT industry has begun to develop new strategies for “greening” traditional data centers, yet the physical reality of modern campus CyberInfrastructure (CI) is a complex network of ad hoc and sub-optimal energy environments in departmental facilities.
But because the value of computational and data-intensive approaches to research is increasingly embraced, this number of departmental facilities is swelling fast and creating campus-wide crises of space, power, and cooling.
This GreenLight Instrument Development will enable 5 communities of application scientists, drawn from metagenomics, ocean observing, microscopy, bioinformatics, and the digital media,
to come to understand, through this instrumentation, how to measure and then minimize energy consumption, to make use of novel energy/cooling sources coming online at UCSD,
and employ middleware that automates optimal choice of compute/power strategies.
This will enable domain application researchers to continue to exploit the exponential improvements in silicon technology, and compete globally.
Using OptIPuter and Quartzite methods developed under earlier
NSF awards, computing clusters, truthfully inconvenient and energy-inefficient as now found in users’ labs, can be resituated in pre-fabricated campus “machine rooms” where detailed measurements of energy efficiency can be made,
but with the clusters still operated by their virtual “owners” remotely from their labs. Unfortunately, almost nothing is known about how to make these shared virtual clusters energy efficient,
since there has been no financial motivation to do so. How can better power efficiency configurations and deeper architectures be researched and discovered?
What can/should one teach the next generation of engineers who must scale from an education in Computer Science (CS) to a deep understanding in engineering physics, or vice-versa?
As a step, a full-scale virtualized device, the GreenLight Instrument, will be developed to measure, monitor, and make publicly available, via service-oriented architecture methodology,
real-time sensor outputs, thus allowing researchers anywhere to study the energy cost of at-scale scientific computing.
This Instrument will enable an experienced team of CS researchers to make deep and quantitative explorations in advanced computer architectures, including alternative circuit fabrics such as Field-Programmable Gate Arrays,
direct-graph execution machines, graphics processors, solid-state disks, and photonic networking.
The enabled computing and systems research will yield new quantitative data to support engineering judgments on comparative “computational work per watt” across full-scale applications running on at-scale computing platforms.
This will help re-define fundamentals of systems engineering for green CI, a transformative concept.