Distributed Storage Server Booting with Virtual Disk Failover
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing server booting systems in distributed storage face challenges in fault tolerance, particularly when the system disk fails, leading to server unavailability and increased downtime due to the need for backup, reinstallations, and synchronization, which are inconvenient and costly.
Innovation Solution
Implementing a system where the Operating System (OS) is hosted on a virtual disk stored on distributed storage, allowing boot components to be started from multiple physical block devices, enabling access to the virtual disk for booting, and configuring boot components to manage the OS installation and services, utilizing iSCSI for networking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the OS is installed on a single system disk, then the server can boot quickly and simply, but the fault tolerance is poor and the server becomes unavailable when the system disk fails
Solution Approach 1:
The patent divides the boot system into two independent parts: boot components (stored on physical block devices) and OS images (stored on virtual disks in distributed storage). This segmentation allows the boot process to start from multiple physical locations while the OS itself is distributed, resolving the contradiction between fault tolerance and simplicity by separating the single-point-of-failure OS from the multi-location boot components.
Solution Approach 2:
The patent introduces a virtual disk as an intermediary layer between the physical block devices and the OS. The boot components on physical devices access the virtual disk, which in turn loads the OS image from distributed storage. This intermediary decouples the boot process from any single physical disk, enabling fault tolerance without requiring complex changes to the OS installation itself.
2Reliability
If a backup system disk is maintained, then fault tolerance is improved, but storage space is consumed and synchronization complexity increases
Solution Approach 1:
Instead of maintaining a complete backup system disk that requires synchronization, the patent uses copying at the OS image level in distributed storage. Multiple nodes can share the same OS image through the distributed storage system, eliminating the need for separate backup disks and their associated synchronization overhead while consuming minimal additional storage space.
Solution Approach 2:
The distributed storage system serves multiple functions simultaneously: it stores OS images, provides fault tolerance through replication, and enables multiple servers to share the same OS image. This universal approach replaces the dedicated backup disk approach, reducing total storage consumption while maintaining or improving fault tolerance across the system.
3Ease of repair
If the OS is reinstalled after system disk failure, then the server can be restored, but significant downtime and manual intervention are required
Solution Approach 1:
The patent prepares OS images in advance in the distributed storage system, so that when a physical block device fails, the boot components can immediately access a pre-prepared OS image from the distributed storage without requiring reinstallation. This preliminary preparation eliminates the time-consuming reinstallation process while maintaining ease of repair through automated boot component selection.
Solution Approach 2:
The boot component is designed to automatically detect failures and self-select alternative OS images from the distributed storage system without requiring manual intervention for reinstallation. This self-service capability reduces both downtime and the complexity of repair operations, as the system automatically restores functionality by loading from alternative sources.
4Reliability
If multiple physical block devices are made bootable, then fault tolerance is improved, but the boot process complexity increases
Solution Approach 1:
Instead of having multiple boot components across physical devices point to multiple different OS installations (which would increase complexity), the patent inverts the approach: all boot components on multiple physical devices point to the same virtual disk and OS image in distributed storage. This inversion maintains boot process simplicity while achieving fault tolerance, as the complexity is shifted to the distributed storage layer rather than the boot process.
Data Source
AI summary
Disclosed herein are systems and method for booting servers in a distributed storage to improve fault tolerance. In one aspect, an exemplary method comprises, during a booting of a hardware server belonging to the distributed storage, starting a boot component from on any one physical block device from at least two or more physical block devices of the hardware server, wherein the at least two or more physical block devices of the hardware server are bootable and each physical block device stores the boot component, wherein an Operating System (OS) of the hardware server is installed on a virtual disk, and wherein the virtual disk is stored on the distributed storage, and by the boot component, accessing the distributed storage to obtain access to the virtual disk, mounting the virtual disk as a root file system of the OS, and booting the OS from the mounted virtual disk.