Full Program »
Virtual Machine Migration by Google Cloud's Migration Service
The migration of virtual machines running on legacy platforms is essential to bring many applications to the cloud computing environments. The migration time can vary due to the form of a machine. A machine's operating system, defined CPU size, disk size, and data written to the disk are just some potential variables that could impact the time of migration.
These variables can have a negligible or significant effect on the time of migration. A solid understanding of the effect of each one could provide sufficient information to plan a migration. A strategy can be developed to mitigate the effects of the large factors of migration, which can become quite important when multiple machines are migrated in parallel.
The Google Cloud Platform allows the creation of virtual machines with large amounts of variance. Google Cloud's VM Migration Service, powered by Cloud Endure, allows virtual machine migration and disaster recovery with relative ease. Cloud Endure effectively has two phases of migration: the initial data replication phase which uses block-level data-replication on the source machine bringing it to Cloud Endure, and the launch phase where the machine on Cloud Endure is launched on the target location.
We have conducted a series experiments by creating virtual machines on Google Cloud of varying types and using Cloud Endure to migrate these machines. For the initial data replication phase, the disk size of the virtual machine was found to be a large factor of migration time. Data on disk was also found to impact the migration time. Operating system and CPU size were negligible in their affects. For the launch phase, machine variables had little impact on the operation, with test results being consistent among machines.