Collaborative Research: CNS Core: Small: Resource-efficient, Strongly Consistent Replication for the Cloud

Data storage within cloud computing systems relies upon replication protocols that store copies of data on multiple servers for reliability. A desirable property of a replication protocol is strong consistency - the ability of multiple servers with copies of data to act as a single, highly performant system with one copy of the data, even when some of the servers fail. Existing strongly consistent protocols improve performance at the cost of sacrificing resource efficiency, which increases the cost of data storage on the cloud. This project aims to explore the inefficiencies in current protocols and design new protocols for cloud computing systems.

This project will study the resource efficiency of existing replication protocols, focusing on cloud deployments in resource-shared settings. Such investigation would be incomplete without including other environmental factors, such as programming language and framework choices. In addition, the project will use the investigation results to design new resource-efficient protocols and optimizations. These will leverage the core algorithmic improvements in addition to new hardware technologies, such as Remote Direct Memory Access (RDMA) and Non-volatile Memory (NVM). The developed protocols will streamline communication, avoid unnecessary message exchange, prioritize lower overhead communication strategies, and reduce work amplification.

Educational and technology transfer aspects play a significant role in this project. This work will facilitate bidirectional technology transfer between academia and industry through meetings and collaborations. To further remove technology-transfer barriers, all protocols and algorithms will be well-documented and open-sourced. This project will bring under the spotlight the importance of building resource-efficient software in cloud computing environments and will develop a new class, projects, and lab modules emphasizing design techniques and programming practices that increase resource efficiency in the cloud software. Through the curriculum and teaching, the project aims to engage undergraduate students and students from underrepresented groups.

This project will release all code artifacts, data, and curriculum materials on the GitHub platform. If applicable, any large datasets or raw data materials will be stored in a public cloud storage system. The project will maintain the GitHub repository, available at https://github.com/resource-efficient-replication. Upon the completion of the project, the GitHub handle will remain active for historical purposes.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.