CAREER: Rethinking Virtualization in Cloud-Native Systems

University Of Texas At Arlington, Arlington TX

Investigators

Abstract

Cloud-native technologies offer advanced and effective means to develop and manage today’s ubiquitous cloud applications. For example, monolithic cloud applications are being replaced with graphs of smaller, simpler “microservices” for reduced development complexity and increased code velocity. Further, such microservices are managed by cloud-native systems automatically, liberating cloud users from onerous operations of their in-cloud applications. Unfortunately, systems software that underpins cloud-native systems, such as virtualization, faces critical challenges to meeting the stringent needs of emerging microservices-based cloud applications, such as strong-yet-lightweight isolation, fast inter-microservice communication, and fine-grained elasticity control. This CAREER project proposes to conduct a holistic study of virtualization techniques to 1) identify critical bottlenecks in resource isolation, I/O communication, and service elasticity; 2) propose new system virtualization solutions to address these bottlenecks; and 3) validate the proposed solutions with real-world prototypes. The knowledge developed in this project will advance the key aspects of systems software in cloud-native systems, thus benefiting all cloud applications that are integral to society running on these systems. The research outcomes will have influences on the design and implementation of production cloud-native systems via university-enterprise collaborations and technology transfer and be integrated into core computer science courses. This project will provide training in the computer systems domain to graduate students, undergraduate students, and high school students, via diverse outreach plans of summer camps, high school student/instructor mentorship, and the inclusion of underrepresented minority and women engineers. The overarching goal of the CAREER project is to investigate and develop virtualization techniques for highly secure, efficient, and elastic cloud-native systems. First, the project will develop a strong-yet-lightweight isolation architecture for efficiently and securely sandboxing distributed microservices. It will move monolithic kernel functions to more isolated and lightweight userspace kernel microservices to provide system services to (and sandbox) all microservices of a cloud-native application. Second, the project will build fast userspace (network and storage) data planes tailored for interactive microservices. It will retrofit kernel I/O stacks to preserve high-quality kernel functionality meanwhile augmenting existing I/O stacks to address critical inter-layer data communication/processing bottlenecks. 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.