Graduation Semester and Year

Summer 2024

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Computer Science

Department

Computer Science and Engineering

First Advisor

HUI LU

Abstract

Overlay networks are the de facto network virtualization technique for providing flexible and customized connectivity among distributed containers in the cloud. Despite their widespread adoption, overlay networks incur significant overhead due to their complexity, resulting in notable performance degradation compared to physical networks.

In this dissertation, I present our three-stage solutions aimed at addressing the challenges of efficiency and scalability in cloud-based container overlay networks: Firstly, we conduct a comprehensive empirical performance study of container overlay networks, identifying crucial parallelization bottlenecks within the kernel network stack. Our observations and root cause analysis uncover that these inefficiencies primarily arise from the increased complexity and prolonged packet processing paths introduced by additional network devices. Secondly, we propose Falcon, a fast and balanced container networking approach designed to scale the packet processing pipeline in overlay networks. Falcon pipelines software interrupts associated with different network devices of a single flow across multiple cores, thereby preventing the execution serialization of excessive software interrupts from overloading a single core. Additionally, Falcon supports multiple network flows by effectively multiplexing and balancing software interrupts among available cores. Lastly, we introduce MFLOW, a novel packet steering approach to parallelize the in-kernel data path of network flows. MFLOW exploits fine-grained packet-level parallelism by splitting packets of the same flow into multiple micro-flows, allowing parallel processing on multiple cores. This approach devises new generic mechanisms for flow splitting while preserving in-order packet delivery with minimal overhead.

Keywords

Cloud computing, Networking systems, Kernel Optimization, Datacenter Infrastructure

Disciplines

OS and Networks

License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.