• DocumentCode
    262272
  • Title

    Efficient Pre-copy Live Migration with Memory Compaction and Adaptive VM Downtime Control

  • Author

    Guangyong Piao ; Youngsup Oh ; Baegjae Sung ; Chanik Park

  • Author_Institution
    Dept. of Comput. Sci. & Eng., Pohang Univ. of Sci. & Eng., Pohang, South Korea
  • fYear
    2014
  • fDate
    3-5 Dec. 2014
  • Firstpage
    85
  • Lastpage
    90
  • Abstract
    Virtual machine (VM) live migration is an important feature of the virtualization technique. Pre-copy method is typically used to support live migration in most virtualization environments. It is observed that pre-copy may not work in some situation where memory-intensive applications are running in KVM. Moreover, pre-copy live migration does not consider high duplication ratio between memory and root file system as well as between memory snapshots. This paper proposes two techniques to improve pre-copy live migrations: 1) memory compaction technique based on disk cache and memory snapshot, 2) adaptive downtime control scheme based on the history of VM´s memory update information called WWS (Writable Working Set). We have successfully implemented the proposed method in KVM. It is shown that the proposed method can support live migration under any circumstances with little overhead.
  • Keywords
    cache storage; disc storage; virtual machines; virtualisation; VM memory update information; WWS; adaptive downtime control scheme; disk cache; duplication ratio; memory compaction technique; memory snapshots; precopy live migration; precopy method; root file system; virtual machine live migration; virtualization environments; virtualization technique; writable working set; Compaction; Encoding; History; Linear regression; Memory management; Radio access networks; Servers; Downtime Control; KVM; Memory Compaction; VM Live Migration;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Big Data and Cloud Computing (BdCloud), 2014 IEEE Fourth International Conference on
  • Conference_Location
    Sydney, NSW
  • Type

    conf

  • DOI
    10.1109/BDCloud.2014.57
  • Filename
    7034770