Computer Science & Engineering Assistant Professors Feng Yan, Dongfang Zhao, and their co-advised Ph.D. student Xinying Wang received the Best Student Paper Award for their paper "Toward Cost-effective Memory Scaling in Clouds: Symbiosis of Virtual and Physical Memory" at the 11th IEEE International Conference on Cloud Computing (IEEE CLOUD 2018). IEEE CLOUD is the flagship conference in IEEE Computer Society and the most prestigious conference in cloud computing.
The project investigates how to deploy memory-intensive applications to public clouds in the most cost-effective manner by using a symbiosis of virtual and physical memory. This work is motivated by the fact that there are many instance types available on public cloud with discrepant memory capacity and price. Current state-of-the-practice uses trial and error to find the cheapest instance type with sufficient physical memory to execute the application, causing both performance overhead and additional monetary cost.
This research advances the state-of-the-art by using a symbiosis of virtual and physical memory and developing a performance-cost model that can accurately quantify the performance impact of virtual memory, which is much slower than physical memory and can decrease an application's performance. The goal is to help application owners to significantly reduce the cloud usage cost without compromising the application performance, such that a larger user base can benefit from cloud computing with lower cost. This in turn benefits cloud service providers, as potentially more applications are able to move to the cloud.
This research was conducted in partnership with HP Lab's Cong Xu and Microsoft's Ke Wang.