Cloned Block Device Startup to Avoid VM Lazy-Load Read Hotspots
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In cloud computing, the lazy load process for virtual machine starting in batch cloning scenarios leads to read hotspots in block devices, causing performance degradation due to simultaneous data access by multiple devices, which overwhelms the storage cluster.
Innovation Solution
A method and apparatus that involves receiving a lazy load request, acquiring target snapshot data, downloading multiple snapshot copies, connecting each block device to corresponding snapshot copies to obtain log files, and forming a cloned block device to start the virtual machine, thereby distributing data access and reducing read hotspot pressure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If lazy load technology is used to download block data during virtual machine starting, then the starting speed of virtual machines is improved, but read hotspots are formed in block devices causing performance degradation
Solution Approach 1:
The patent divides the block device into multiple independent block devices, each responsible for specific data ranges. This segmentation distributes the read load across multiple devices, preventing read hotspots while maintaining fast data access during virtual machine starting and operation.
Solution Approach 2:
The patent introduces a new dimension of data distribution by organizing block data across multiple block devices in a distributed manner. Instead of concentrating all data in a single block device, the system spreads data across multiple devices, transforming the storage architecture from centralized to distributed and thereby eliminating read hotspots.
2Productivity
If multiple block devices access data simultaneously during batch cloning, then the cloning efficiency is improved, but the storage cluster is overwhelmed causing bandwidth limitations
Solution Approach 1:
The patent segments the data access workload across multiple block devices, where each device handles specific data ranges independently. This segmentation allows parallel access without overwhelming the storage cluster, as the total bandwidth consumption is distributed across multiple devices rather than concentrated on a single device.
Solution Approach 2:
The patent performs preliminary data preparation by pre-loading or pre-positioning data in appropriate block devices before batch cloning operations. This preliminary action reduces the immediate bandwidth demand during cloning by having data ready in advance, thereby improving cloning efficiency without overwhelming the storage cluster.
Data Source
AI summary
Embodiments of the present disclosure provide a method for controlling virtual machine starting based on a cloned block device, and an apparatus and an electronic device. The method includes: receiving a lazy load request sent by a virtual machine to a block device of elastic block storage, acquiring target snapshot data, and sending a snapshot download request to an image center to download at least two snapshot copies; connecting each of the block devices to a corresponding snapshot copy to obtain log files corresponding to the at least two snapshot copies respectively; and forming a cloned block device by loading, through the block device, a log file corresponding to the snapshot copy to, and starting the virtual machine based on the cloned block device.