Timespan: 1.28-1.29
Michael R. Hines and Kartik Gopalan. 2009.Post-copy based live Virtual Machine migration using Adaptive Pre-paging and dynamic self-ballooning. InProceedings of the 2009 ACM sigplan/sigops International Conference on virtual execution Environments(Vee '09). ACM, New York, NY, USA, 51-60. (GS: 94)
By Michael R. hines wrote this paper in S. u. n.y. a Ph.D. student at Binghamton University, who graduated from IBM Watson Research Center, is interested in "Creating and analyzing experimental, networked systems ". He participated in the "IBM cloud rapid experimentation and analysis tool" project, which is an open-source project and mainly serves: "a framework that automates IAAs cloud Benchmarking through the running of controlled experiments ". Currently, the supported IAAs platforms include Amazon EC2, openstack, ibm scp and ibm sce.
Pre-copy is the default method for VM dynamic migration in xen. This article proposes the post-copy method and compares it with the pre-copy method in various aspects (pages transferred, total migration time and network overhead ).
Post-copy migration: "defers the transfer of a VM'sMemeoryContents until after itsProcessor stateHas been sent to the target host ."
The following is an excerpt from the paper:
1. The post-copy method ensures that the memory page is transmitted at most once. However, the post-copy method in this article does not have fault tolerance when the desination node fails. (S1)
2. Adaptive Pre-paging is used to reduce "network page faults (major faults)" (S1)
* About minor/major faults, http://en.wikipedia.org/wiki/page_fault:
- Minor faults: If the page isLoaded in memoryAt the time the fault is generated, but is not marked in the memory management unit as being loaded in memory, then it is called a minor or soft page fault.
- Major faults: If the page isNot loaded in memoryAt the time the fault is generated, then it is called a major or hard page fault.
3. Dynamic Self-ballooning (DSB) is used to process free pages to reduce the total migration time (S1)
In this article, DSB is triggered every five seconds, "responds directly to OS memory allocation requests ".
* Ballooning: allows a guest OS to reduce its memory footprint by releasing its free memory pages back to the hypervisor.
* Memory Footprint (memory usage): The amount of main memory that a program uses or references while running.
4. (S2) describes the related work, which is divided into the following directions:
- Process Migration: the post-Copy technology has been extensively studied in Process Migration.
- Pre-Paging: Also known as "Adaptive prefetching" or "Adaptive remote paging"
- Live VM migration: supports hypervisor-based approaches, OS-level approaches, and wide-area migration.
- Non-live VM migration: non-dynamic migration
- DSB: ballooing technology is widely used in VM memory resizing
The author mentioned that the closest thing to this article is snowflock.
5. (S3) describes the design of the Post-copy method in this article, combining four technologies:
- Demand-Paging: ensure that each page is transmitted only once
- Active push: when the major fault occurs, pages near the fault page will also be transmitted (using spatial locality) (S3.2)
- Pre-Paging: VM page access mode, which predicts the occurrence of major fault, so that the relevant pages can be transmitted in advance
- DSB: Reduce the number of free pages transmitted (s.3)
6. (S4) discusses several implementation issues:
(1) "trap page faults at the target VM": three methods (shadow paging, page tracking, and pseudo-Paging) are introduced. This article uses the third method, this method achieves the fastest (s4.1)
(2) DSB implementation (s4.2)
The experiment platform is based on xen 3.2.1 and para-starter laized Linux 2.6.18.8.
7. (s6) The conclusion section mentions the areas for improvement.
- Investigate shadow paging based page fault detection
- Handle destinationnode failure during post-copy migration
- Implement a hybrid pre/post copy approach