Rethinking Network Quality of Service in the Cloud Era
Haq, Osama.
2021
-
Thesis (Ph.D.)--Tufts
University, 2021.
Submitted to the Dept. of Computer Science.
Advisor: Fahad Dogar.
Committee: Alva Couch, Jeffrey Foster, Usman Khan, and John Byers.
Keyword: Computer science.
To provide quality of service (QoS) to the users, interactive applications (video conferencing, augmented and virtual reality, online gaming, etc.) ... read morerequire strict guarantees from the network. However, the Internet -- due to its best effort nature -- does not provide any packet loss, latency, or bandwidth guarantees to applications, which can lead to degraded application performance and frustrating user experience. In this thesis, we revisit the long-standing problem of network QoS and propose the concept of judicious QoS -- combining the cheaper, best-effort IP service with the cloud, which offers a highly reliable infrastructure and the ability to add in-network services, albeit at a higher cost. We first conduct a measurement study of current cloud infrastructure and compare the network performance of major cloud providers with one another and against the public Internet. We confirm the existence of dedicated paths between the data centers of the cloud providers and observe no significant performance difference among major cloud providers. We also find that cloud inter-data center paths offer high reliability, predictable latency, low jitter, and higher bandwidth than public Internet paths. Using insights from our measurement study, we demonstrate the use of judicious QoS by designing and implementing ReWAN, a packet recovery service that leverages the cloud to recover from lost or delayed packets. ReWAN supplements and protects best-effort connections by sending a small number of coded packets along the highly reliable but expensive cloud paths. When receivers detect packet loss, they recover packets with the help of the nearby data center, not the sender, thus providing quick and reliable packet recovery for latency-sensitive applications. Finally, we propose the judicious QoS framework (J-QoS), which supports diverse use cases of cloud overlay e.g. path switching, packet recovery, and mobility. The J-QoS framework identifies the requirements for supporting these use cases and offers a of variety services, both in the cloud and at the end hosts, to meet application QoS demands. J-QoS cloud services include forwarding, caching, and coding of application streams. Its enhanced endpoint support includes mechanisms such as selective duplication, multipath routing, and loss detection. We demonstrate the feasibility of ReWAN and other J-QoS use cases by: i) using measurements from multiple vantage points, ii) analyzing pricing data in our cost model, and iii) deploying on a real network and cloud. We also consider case studies on how our proposed solutions work with services up and down the network stack, including Skype video conferencing, TCP-based web transfers, and cellular access networks.read less - ID:
- 1g05fr81p
- To Cite:
- TARC Citation Guide EndNote
- Usage:
- Detailed Rights
