In the rapidly changing world of computing technology, creativity is key to staying ahead of the curve. Ever-increasing demand for speed and power, at both large and small scales, constantly puts pressure on technical boundaries and calls old certainties into question.
In an industry where paradigm shifts can happen in a matter of months, educators must be constantly teaching with one foot in the future in order to train students to solve the problems of tomorrow. It is on this forefront of innovation that Dr. Ziliang Zong, associate professor of computer science at Texas State University, works to bridge the gap between education and industry by instilling a mindset of constant growth in the next generation of developers.
In recent years, we have seen an end to laws describing the relationship between size and power for computing cores. Previously, the number of calculations that could be performed by a chip of a certain size was expected to double every 18 months. Due to power constraints, the industry shifted from single-core design to multi-core design, which led to the dual-, quad-, and other multi-core CPUs we’ve become familiar with in the last decade. Very recently, with the emerging power-hungry software applications such as AI and Bitcoin mining, new barriers to computing power on multi-core devices have arisen, and new architecture based on Application Specific Integrated Circuit (ASIC) is emerging to address these challenges. “The trend has been very clear for the last three years,” explains Zong. A design shift is under way, from generalized chips capable of performing any function requested of them to domain-specific chips housed independently from one another and specialized to certain tasks. In both mobile computing devices and massive data centers, complex interconnected cores perform calculations specific to their domain and share the workload across the system, thus doing more work more efficiently.
Because this type of design is so new and demand for computing power is so high, Zong teaches from the most recently published papers in the field rather than from any prescribed program. “It is very exciting; I never stop learning.” He says it is a joy to work with students, knowing just as much about the new material as they do and learning alongside them. He encourages students to not fear failure, because new methods are emerging all the time.
Zong specializes in green computing, a set of practices and design principles that seeks to limit the environmental impact of information technology.
Today, there are over 20 million software developers in the world, according to Evans Data Corporation. Billions of lines of code are written every year and deployed on mobile devices, embedded systems, PCs and servers, which consume a significant amount of energy.
Software developers used to rely on new generations of hardware to improve performance and energy efficiency or treated software energy efficiency as an afterthought. As hardware is quickly approaching its physical limits on further improving energy efficiency, software will play an increasingly important and more proactive role in advancing green IT. However, Zong says that in recent years this has been changing as engineering teams seek to improve power-usage effectiveness, a measure of how much energy the device draws upon and how many calculations it performs. The industry is in the midst of a “best practices transition” as it seeks to get the most out of new and existing systems. According to Zong, the largest opportunity for implementing green practices is in application development, which he feels is where his students can be most effective. Efficiency will be expected to be built-in to emerging technology, so students are experimenting now to develop techniques that cut down on power-hungry apps. To put their theory into practice, students turn to the Marcher System, a unique system developed at Texas State to support green computing research and education.
The Marcher System allows students developing software to test their program’s energy efficiency by measuring power-consumption in real time. The system contains 13 high-performance computing nodes. Each node has four quad-core processors, a K20 GPU and an Intel Xeon Phi many-core coprocessor. Students use an application called Greencode to run their programs on the Marcher System and improve their energy efficiency. By developing and implementing green software practices into their designs, students are cultivating a set of high-demand skills that they can take into the highly competitive IT industry and create their own paradigm shift toward a more environmentally conscious future.
Accurate as of June 2018