Virtual Machine Migration via On-Demand Storage Block Transfer
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Legacy techniques for migrating virtual machines between clusters are slow and costly due to the need to move entire storage assets, introducing significant latency and bandwidth consumption, even though virtual machines often only require a small portion of the data to function.
Innovation Solution
The technique involves migrating only the metadata of virtual disks to the new cluster, allowing the virtual machine to start up quickly, and then satisfying storage asset demands on-demand by moving specific blocks as needed, based on usage patterns and access histories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the entire storage asset is moved from the original cluster to the new cluster before starting the virtual machine, then the virtual machine can access all its data, but the migration process consumes significant network bandwidth and introduces unacceptable latency
Solution Approach 1:
The storage asset is divided into multiple blocks, and only the metadata describing these blocks is moved initially. The virtual machine can access the metadata and retrieve specific data blocks on-demand from the original cluster, avoiding the need to transfer the entire storage asset upfront.
Solution Approach 2:
The metadata is moved and made available to the virtual machine before the actual data blocks are transferred. This preliminary action allows the virtual machine to start up and begin operations immediately, while data blocks are transferred in the background as needed.
2Reliability
If the entire storage asset is moved from the original cluster to the new cluster, then the virtual machine has full data access, but the network bandwidth consumption becomes significant and costly
Solution Approach 1:
Only the essential metadata is extracted and moved to the new cluster, while the bulk data blocks remain in the original cluster. The virtual machine accesses data blocks on-demand through the metadata, eliminating the need to transfer unnecessary data and reducing network bandwidth consumption.
Solution Approach 2:
Instead of moving the complete storage asset, only the necessary metadata is moved initially. Additional data blocks are transferred partially and on-demand based on actual access requirements, avoiding excessive data transfer and associated costs.
3Loss of energy
If data compression is applied to reduce the size of moved data, then the bandwidth consumption is slightly reduced, but the migration process still requires moving the entire storage asset and introduces significant latency
Solution Approach 1:
The storage asset is segmented into metadata and data blocks. Only the compact metadata is moved initially, enabling the virtual machine to start immediately. This segmentation approach is more effective than compression because it eliminates the need to transfer the entire data set rather than just reducing its size.
Data Source
AI summary
A method commences upon receiving a request to migrate a subject virtual machine from a first computing machine to a second computing machine. An agent determines which storage assets are used by the subject virtual machine. The agent collects volume metadata pertaining to the storage assets and clones volume metadata from the first computing machine to the second computing machine. The data blocks comprising the volumes are not moved to the second computing machine until after migrating the subject virtual machine from the first computing machine to the second computing machine. After starting the migrated instance of the subject virtual machine, the migrated instance mounts volumes using the cloned metadata. The first computing machine and the second computing machine cooperate by processing storage asset block requests issued by the migrated virtual machine. Data blocks comprising the volume are moved on demand. The migrated virtual machine can be migrated yet again.