University of Arizona Systems Engineering senior Paul Thibodeaux adjusts a robotic car as part of a simulator that researchers in the Systems & Industrial Engineering Department are building to evaluate distributed computing systems. The cars simulate automated vehicles, such as those found in an open pit mine. In this case, however, they are directed to various sites to load up on ones or zeros instead of ore: hence the name of this simulator - the Open Bit Mine.  

Distributed computing systems

From e-groceries to mines, they decide what should be done when

    
Instead of strolling down the grocery aisle, you soon may log onto a web-based e-grocery, click on your favorite frozen dinners and remotely direct a robotic shopper to the frozen food aisle. Later, you'll drive past the e-grocery pickup window and collect the groceries you ordered on-line.

Computers will link you to the store's computer, which, in turn, will be hooked to the robotic shoppers. Other computers handle inventory, delivery and packaging. The whole setup is called a "distributed computing system." These systems are also found at mines with automated trucks, production shops with robotic vehicles, and similar facilities.

University of Arizona Systems and Industrial Engineering researchers, along with those in UA electrical engineering and computer science, are designing more efficient distributed systems under the NSF-funded Speed-CS program (Simulation Platform for Experimentation and Evaluation of Distributed Computing Systems).

"We're building a simulation that spans both the automation and operations parts of these systems," says Suvrajeet Sen, principal investigator on the $990,000, three-year project.

Automation includes hardware - processors, computer architecture and such. Operations centers on software. Electrical engineering is providing simulation software support, computer science is developing communications and networking systems, and Sen and SIE Assistant Professor Frank Ciarallo are providing performance analysis tools and putting it all together within a simulator.

"Many stand-alone modules, are already available," Sen says. "We are building all this into an integrated platform so we can develop and test algorithms and policies."

Past simulations of distributed computer systems have not integrated operations and automation, so results and applications have been limited. But hardware-drives-software-drives-hardware in chicken-and-egg fashion and Sen's group hopes that creation of multi-scale simulations will lead to more efficient and robust systems.

"This is in the realm of integrative engineering," he says. "There are many layers of engineering that we are trying to integrate into this software."

  

Mining ones, zeros at the Open-Bit Mine

To test algorithms and hardware for distributed computing systems, SIE researchers need a facility like an open pit mine.

"We want to mimic a mine, but on a laboratory-sized scale," says Professor Suvrajeet Sen. "So we designed the Open Bit Mine, a series of tiny roadways on a four-by-eight area on which we operate toy robotic cars."

Processors on the cars collect ones and zeros from relay stations. Researchers can ask a car to collect four zeros and four ones in an alternating series, for instance. The car travels to the various relay stations, processes the bits and sends the information to the main computer.

"The physical simulation will help us experiment with hybrid (virtual and physical) models. Such models will make it possible to build more realistic simulations."