[论文笔记] Live Migration of Virtual Machines (NSDI, 2005)
2013-01-17 13:23
459 查看
Timespan: 1.15 – 1.17
Christopher Clark, Keir Fraser, Steven Hand, Jacob Gorm Hansen, Eric Jul, Christian Limpach, Ian Pratt, and Andrew Warfield. 2005. Live migration of virtual machines. In Proceedings of the 2nd conference on Symposium on Networked Systems Design & Implementation - Volume 2(NSDI'05), Vol. 2. USENIX Association, Berkeley, CA, USA, 273-286. (gs: 1194)
在google scholar里搜“live migration”,这篇论文排在第一,引用数将近1200,是这个方向的经典之作了,本文中提到的"pre-copy"、"writable working set(WWS)"被其后的论文引用较多。作者Christopher Clark的单位是剑桥大学,网上搜这个名字,最有名的一人也是在剑桥大学工作,然而却是历史学家,并非同一人;却没有找到作者本人的主页。另外顺便找到介绍Xen历史的一张网页。
本文介绍了在局域网环境中,使用pre-copy方式对虚拟机进行动态迁移的方法,使用了"page-level protection hardware"来确保被传输快照的一致性,使用"rate-adaptive algorithm"来控制迁移对虚拟机上服务正常运作的影响。使用本文方法,可以达到完成迁移只会造成毫秒级宕机的效果。
以下是论文笔记和摘录:
1. .虚拟机迁移时要考虑以下方面的要求和目标(S1)
minimize the downtime
minimize the total migration time
尽量降低对被迁移虚拟机的干扰(如迁移工作会占用本来由虚拟机使用的带宽资源)
(S3.1)中将memeroy transfertransfer分成了"push phase, stop-and-copy phase, pull phase"三个阶段,并对比了几种迁移方式对downtime和migration time的关系:
2. 另外design decisions还要考虑network resources和local storage(S3.2)。
networks resources: 本文考虑的是source/destination machine使用同一个switched LAN的情况,可以使用广播等方式。
storage: 如果使用了NAS device,就很容易解决。本文不考虑使用了local-disk storage的情形。
3. S3中介绍了迁移的概念流程:
4. (S4)中提出了pre-copy方法的一个关键问题:何时应该结束pre-copy阶段而进行stop-and-copy?
作者认为可以分成两种pages:第一种不怎么变动的,因此适合pre-copy;第二种频繁变动,更适合stop-and-copy。文中把第二种称为"writable working set(WWS)"。
作者实现是基于Xen,Xen中的shadow page tables可以用来追踪特定OS产生的dirtying statistics。(S4.1)
5. 其他摘录
(1) 为什么要进行OS instance migration?
a clean separation between hardware and software
facilitates fault management
facilitates load balancing
facilitates low-level system maintenance
(2) 虚拟机迁移的优势(相比于process-level migration)(S1)
避免"residual dependencies(残余依赖)"问题
in-memory state can be transferred in a consistent and efficient fashion
seperation of concerns between the users(使用虚拟机的人) and operator (迁移移虚拟机的人)
Christopher Clark, Keir Fraser, Steven Hand, Jacob Gorm Hansen, Eric Jul, Christian Limpach, Ian Pratt, and Andrew Warfield. 2005. Live migration of virtual machines. In Proceedings of the 2nd conference on Symposium on Networked Systems Design & Implementation - Volume 2(NSDI'05), Vol. 2. USENIX Association, Berkeley, CA, USA, 273-286. (gs: 1194)
在google scholar里搜“live migration”,这篇论文排在第一,引用数将近1200,是这个方向的经典之作了,本文中提到的"pre-copy"、"writable working set(WWS)"被其后的论文引用较多。作者Christopher Clark的单位是剑桥大学,网上搜这个名字,最有名的一人也是在剑桥大学工作,然而却是历史学家,并非同一人;却没有找到作者本人的主页。另外顺便找到介绍Xen历史的一张网页。
本文介绍了在局域网环境中,使用pre-copy方式对虚拟机进行动态迁移的方法,使用了"page-level protection hardware"来确保被传输快照的一致性,使用"rate-adaptive algorithm"来控制迁移对虚拟机上服务正常运作的影响。使用本文方法,可以达到完成迁移只会造成毫秒级宕机的效果。
以下是论文笔记和摘录:
1. .虚拟机迁移时要考虑以下方面的要求和目标(S1)
minimize the downtime
minimize the total migration time
尽量降低对被迁移虚拟机的干扰(如迁移工作会占用本来由虚拟机使用的带宽资源)
(S3.1)中将memeroy transfertransfer分成了"push phase, stop-and-copy phase, pull phase"三个阶段,并对比了几种迁移方式对downtime和migration time的关系:
包含的阶段 | downtime | migration time | performance | |
pure stop-and-copy | stop-and-copy | proportional to the amount of physical memeory allocated to the VM | proportional to the amount of physical memeory allocated to the VM | |
pure demand-migration | short stop-and-copy phase, pull phase | much shorter | much longer | performance after migration is likely to be unaccetably degraded |
pre-copy | push phase, short stop-and-copy phase | very short |
networks resources: 本文考虑的是source/destination machine使用同一个switched LAN的情况,可以使用广播等方式。
storage: 如果使用了NAS device,就很容易解决。本文不考虑使用了local-disk storage的情形。
3. S3中介绍了迁移的概念流程:
4. (S4)中提出了pre-copy方法的一个关键问题:何时应该结束pre-copy阶段而进行stop-and-copy?
作者认为可以分成两种pages:第一种不怎么变动的,因此适合pre-copy;第二种频繁变动,更适合stop-and-copy。文中把第二种称为"writable working set(WWS)"。
作者实现是基于Xen,Xen中的shadow page tables可以用来追踪特定OS产生的dirtying statistics。(S4.1)
5. 其他摘录
(1) 为什么要进行OS instance migration?
a clean separation between hardware and software
facilitates fault management
facilitates load balancing
facilitates low-level system maintenance
(2) 虚拟机迁移的优势(相比于process-level migration)(S1)
避免"residual dependencies(残余依赖)"问题
in-memory state can be transferred in a consistent and efficient fashion
seperation of concerns between the users(使用虚拟机的人) and operator (迁移移虚拟机的人)
相关文章推荐
- [论文笔记] Performance and energy modeling for live migration of virtual machines (Cluster Comput, 2011)
- [论文笔记] Live Gang Migration of Virtual Machines (HPDC, 2011)
- [论文笔记] Live Migration of Multiple Virtual Machines with Resource Reservation in Cloud Computing Environments (CLOUD, 2011)
- [论文笔记] Optimizing the live migration of virtual machine by CPU scheduling (JoNCA, 2011)
- [论文笔记] Costs of virtual machine live migration A survey (SERVICES, 2012)
- [论文笔记] An Optimized Control Strategy for Load Balancing Based on Live Migration of Virtual Machine (ChinaGrid, 2011)
- [论文笔记] Exploiting Data Deduplication to Accelerate Live Virtual Machine Migration (CLUSTER, 2010)
- [论文笔记] Live deduplication storage of virtual machine images in an open-source cloud (Middleware, 2011)
- [论文笔记] Post-copy based live virtual machine migration using adaptive pre-paging and dynamic self-ballooning (VEE, 2009)
- [论文笔记] On Construction of Cloud IaaS for VM Live Migration Using KVM and OpenNebula (ICA3PP, 2012)
- 【云迁移论文笔记】A Comparison of On-premise to Cloud Migration Approaches
- [论文泛读] Dynamic placement of virtual machines for managing sla violations (IM, 2007)
- [论文笔记] A service creation environment based on end to end composition of Web services (WWW, 2005)
- 论文笔记之Label-Free Supervision of Neural Networks with Physics and Domain Knowledge
- 【论文阅读笔记】Makeup-Go: Blind Reversion of Portrait Edit
- TreeFTL:Efficient RAM Management for High Performance of NAND Flash-based Storage Systems-论文注释笔记
- 【论文笔记】One Millisecond Face Alignment with an Ensemble of Regression Trees
- 论文笔记-Augmented Lagrange Multiplier Method for Recovery of Low-Rank Matrices
- 《Higher-order organization of complex networks》-论文学习笔记
- Processor topology inside of Hyper-V virtual machines