Oracle Funds Coskun to Improve Server Efficiency


Temperature and power profiles for a server, as monitored by Oracle's system telemetry
Temperature and power profiles for a server, as monitored by Oracle's system telemetry

As dependency on all-things-Internet grows, so does the amount of energy consumed by servers. Some estimates suggest that servers were responsible for approximately 5% of all electricity consumption in the United States last year when you factor in the amount of energy used by cooling systems required for the servers.

Assistant Professor Ayse Coskun (ECE) and Dr. Kenny Gross of Oracle are teaming up to research ways to decrease those percentages, and their one-year project, “Temperature-Aware Workload Scheduling for Optimal Energy Efficiency and Reliability for Servers,” recently received $70K from Oracle. This interdisciplinary project will combine research in computer engineering, systems, control theory, and energy efficiency. Oracle will also support the research by providing high-end servers to Coskun’s lab.

Assistant Professor Ayse Coskun (ECE)
Assistant Professor Ayse Coskun (ECE)

According to Coskun, servers run inefficiently because energy is used even when systems are idle. This “leakage power” grows exponentially with rising chip temperatures.

By teaming up with Oracle, Coskun will be able to combine her expertise in intelligent temperature-aware workload scheduling with Oracle’s strength in optimal thermal control research. Oracle also possesses pioneering technology in system telemetry that will allow Coskun and Gross to monitor servers in detail during execution.

“Collecting feedback from the servers is very important for achieving efficiency, high reliability, adaptability, and energy-proportional computing,” said Coskun. “Oracle’s expertise will be highly valuable to advance the state-of-the-art.”

Ultimately, Coskun and Gross hope to create new control algorithms to minimize “thermal-shock” dynamics – unexpected hot and cold shocks that happen due to the complex interplay among load, leakage, and fan control decisions.

“The expected outcome of these new control algorithms will be improved energy efficiency, better vibrational stability, better acoustics (owing to lower fan speed), and better performance for servers and database appliances,” said Coskun.

-Rachel Harrington (