Virtual Machine Image Replication via Block Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image distribution systems in network environments face inefficiencies in starting virtual machines due to high network traffic and I/O operations, as they require transferring entire images or redundant data blocks, which increases time and resource usage.
Innovation Solution
A method and system that replicate images by determining similarity between images, reconstructing new images using equivalent blocks from local or donor devices, and minimizing network link usage through a distributed virtual server optimizer, which generates an image dependency matrix to identify identical blocks and locate necessary blocks on local or donor hosts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If entire image is copied from storage server to local storage, then VM can be started, but the time required is equal to the time to copy the entire image and the amount of data transferred is equal to the image size
Solution Approach 1:
The patent extracts only the necessary blocks from the image rather than copying the entire image. The system identifies and retrieves only those blocks required to start the VM, significantly reducing the amount of data that needs to be transferred while still enabling VM provisioning.
Solution Approach 2:
The patent applies partial action by copying only a portion of the image (specifically, only the blocks that are actually needed to start the VM) rather than the complete image. This partial copying approach reduces transfer time and network usage while maintaining sufficient functionality for VM operation.
2Loss of time
If image blocks are streamed from storage server to target host, then VM can be started sooner, but the amount of data transferred remains the same and redundant blocks are transferred multiple times
Solution Approach 1:
The patent performs preliminary analysis to identify which blocks are needed before the streaming process begins. By pre-determining the necessary blocks and checking for their availability on target hosts or donor hosts, the system avoids redundant transfers and optimizes network usage before the actual VM start process.
Solution Approach 2:
The patent uses copying from local storage on target hosts or donor hosts instead of repeated network streaming. Once blocks are identified as needed, they are copied from available local sources, eliminating redundant network transfers and reducing overall network bandwidth consumption.
3Loss of energy
If read blocks are also copied to local storage, then redundant network transfers are eliminated, but the total amount of data transferred increases to include all blocks
Solution Approach 1:
The patent applies local quality by making storage decisions specific to each block and each host. Rather than uniformly copying all blocks to all hosts, the system determines on a per-block basis whether copying is necessary based on local availability and actual usage needs, optimizing both network usage and storage utilization.
4Reliability
If streaming rate is dynamically adjusted to network conditions, then network congestion is avoided, but the total time to transfer all blocks is significantly increased
Solution Approach 1:
The patent performs preliminary identification of needed blocks and checks for local availability before initiating network transfers. This advance preparation allows the system to minimize network transfers and reduce the impact of dynamic rate adjustments on total transfer time.
Data Source
AI summary
A method and system include replicating an image representing a sequence of bytes on a local storage medium on a target device by determining a similarity between images and reconstructing a new image using equivalent blocks from one or more similar images locally available on the target device or available on donor devices to reduce network link usage and transfer time in replicating the image.