当前位置: X-MOL 学术arXiv.cs.OS › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
BumbleBee: Application-aware adaptation for container orchestration
arXiv - CS - Operating Systems Pub Date : 2020-08-27 , DOI: arxiv-2008.11868
HyunJong Lee, Shadi Noghabi, Brian Noble, Matthew Furlong, and Landon P. Cox

Application-aware adaptation is the key to maintaining acceptable quality when resources become scarce. Application-oblivious responses to resource scarcity, such as TCP congestion control, may fairly reallocate a diminishing resource pool, but only the application knows how to adjust its fidelity under resource scarcity. Unfortunately, modern container-orchestration platforms like Kubernetes do not offer good support for application-aware adaptation. Thus, this paper presents a set of orchestration extensions to support application-aware adaption called BumbleBee. BumbleBee provides a clean abstraction for making decisions about network data using application semantics. Experiments with a BumbleBee prototype show that it can help a video-analytics application utilize cloud resources when available and operate without interruption when disconnected, and it can help the Fugu video server eliminate all stalling while sacrificing only 6-24% mean resolution.

中文翻译:

BumbleBee:容器编排的应用感知适配

当资源变得稀缺时,应用感知适应是保持可接受质量的关键。应用程序对资源稀缺的响应,例如 TCP 拥塞控制,可能会公平地重新分配不断减少的资源池,但只有应用程序知道如何在资源稀缺的情况下调整其保真度。不幸的是,像 Kubernetes 这样的现代容器编排平台并没有为应用程序感知适配提供良好的支持。因此,本文提出了一组编排扩展,以支持称为 BumbleBee 的应用程序感知适应。BumbleBee 为使用应用程序语义做出有关网络数据的决策提供了一个清晰的抽象。
更新日期:2020-09-18
down
wechat
bug