Collaborative Research:CNS Core:Small:Towards Efficient Cloud Services

点击翻译按钮获取中文摘要

Cloud computing frameworks enable a wide range of services while sharing computation resources and infrastructure costs. To achieve these benefits, cloud computing frameworks rely on layers of abstractions to reduce the complexity of distributed and heterogeneous computational infrastructure. Abstractions hide resource management complexities and improve programmability. However, abstractions make cloud frameworks less observable, resulting in various forms of inefficiencies. This project will address the challenges of practical cloud monitoring techniques to guide cloud application development and system design.This project will explore the inefficiency patterns in cloud computing infrastructures and applications. More specifically, it will provide novel measurement techniques to enable monitoring these inefficiencies across the cloud layers of abstraction. Additionally, the project will develop tools that will provide actionable insights for high-performance cloud frameworks and application development. This project has three thrusts. First, it will measure language-level abstractions for intra-application inefficiencies. Second, it will explore the inefficient communication patterns among microservices for inter-service optimization. Third, it will develop tools to analyze inefficiencies in the entire stack of cloud software layers of abstraction.This project will bridge the knowledge gap between application developers and system designers to provide more efficient cloud environments. It will advance the state-of-the-art cloud monitoring techniques and address the current and future challenges in the cloud computing community. The tools developed from this project will have broad interest from industry, research institutes, and laboratories for efficient code execution and high system throughput. Furthermore, the project will disseminate the obtained knowledge through hands-on training sessions and tutorials. Finally, the project will facilitate curriculum development with a particular focus on involving minority and under-represented students.The project will maintain a website at https://www.probir.info/cloudprof. The website will host all the project outcomes, including the publications, open-source code, toolkits, datasets, documentation, and tutorials. The website will be accessible to the public throughout the project lifetime and beyond.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.