当前位置: X-MOL 学术IEEE Internet Things J. › 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.)
Early-Warning-Time-Based Virtual Network Live Evacuation Against Disaster Threats
IEEE Internet of Things Journal ( IF 8.2 ) Pub Date : 12-31-2019 , DOI: 10.1109/jiot.2019.2963319
Ning-Hai Bao , Ming Kuang , Subhadeep Sahoo , Guo-Ping Li , Zhi-Zhong Zhang

Network virtualization enables cloud service providers sharing various substrate resources and serving end users with heterogeneous virtual network (VN) applications. As more and more Internet applications migrate to the cloud, the survivability of VNs, particularly for disaster survivability, has become a crucial issue. In this article, we study the VN survivability problem and propose an early-warning-time-based VN live evacuation (EVLE) scheme to combat against disaster threats. Aiming at an upcoming disaster destruction, EVLE tries to evacuate as many VNs as possible from the disaster risk zone (DRZ) before the given deadline, while sustaining their online services. The evacuation mainly includes two processes, viz., VN reconfiguration and virtual machine (VM) live migration. For a threatened VN, EVLE first remaps it at the outside of the DRZ with minimal resource cost, then exploits the postcopy technique to migrate the impacted VM to the corresponding reconfigured virtual node. During the VM migration process, the operations of basic bandwidth deployment and bandwidth upgradation are implemented, so as to encourage parallel migrations and maximize the resource utilization within the period of early-warning time. In our simulations, the EVLE and a baseline scheme, as well as a counterpart named as best-effort VN live evacuation (BVLE), are tested with different disaster scenarios. Numerical results illustrate that the EVLE, even if with a longer average evacuation time, can outperform the baseline scheme and BVLE in terms of the evacuation completion ratio under different early-warning times, and particularly can achieve distinct dominance under small early-warning times.

中文翻译:


基于预警时间的虚拟网络实时疏散应对灾害威胁



网络虚拟化使云服务提供商能够共享各种底层资源,并为最终用户提供异构虚拟网络(VN)应用程序。随着越来越多的互联网应用迁移到云端,虚拟网络的生存能力,特别是灾难生存能力,已经成为一个至关重要的问题。在本文中,我们研究了 VN 生存性问题,并提出了一种基于预警时间的 VN 实时疏散 (EVLE) 方案来应对灾害威胁。针对即将到来的灾难破坏,EVLE 试图在给定的期限之前从灾害风险区 (DRZ) 疏散尽可能多的 VN,同时维持其在线服务。撤离主要包括两个过程,即VN重配置和虚拟机热迁移。对于受到威胁的虚拟网络,EVLE 首先以最小的资源成本将其重新映射到 DRZ 外部,然后利用后复制技术将受影响的虚拟机迁移到相应的重新配置的虚拟节点。在虚拟机迁移过程中,实现基础带宽部署和带宽升级操作,以鼓励并行迁移,在预警时间内最大化资源利用率。在我们的模拟中,EVLE 和基线方案以及名为尽力而为 VN 实时疏散 (BVLE) 的对应方案在不同的灾难场景中进行了测试。数值结果表明,即使平均疏散时间较长,EVLE在不同预警时间下的疏散完成率方面也能优于基线方案和BVLE,特别是在较小的预警时间下可以取得明显的优势。
更新日期:2024-08-22
down
wechat
bug