Distributed Key-Value VM Availability During Control Plane Outages
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cluster computing systems with cloud-based control planes are vulnerable to temporary unavailability due to network failures, security attacks, or natural disasters, leading to loss of virtual machine high availability (VM HA) when the control plane is down.
Innovation Solution
Implementing a distributed key-value store (DKVS) on on-premise nodes to manage VM HA independently from the cloud-based control plane, with leader and follower nodes coordinating to detect node unavailability and relocate VMs to surviving nodes, ensuring VM HA is maintained even during cloud service outages.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a cloud-based control plane is used to manage virtual machine high availability, then centralized control and management is improved, but system reliability deteriorates when network failures or security attacks occur
Solution Approach 1:
The patent divides the centralized control function into distributed components by deploying a key-value store across multiple nodes. Each node maintains a local copy of the state database, enabling independent operation during outages. This segmentation allows the system to continue functioning without requiring all nodes to communicate with a central controller.
Solution Approach 2:
The patent implements replicated copies of the state database across multiple nodes in the key-value store. Each node has a local copy that can be read and written independently, allowing the system to maintain availability during outages. The replication mechanism ensures that data is distributed across multiple nodes rather than stored centrally.
2Reliability
If a distributed key-value store is deployed on on-premise nodes, then system reliability is improved during cloud outages, but device complexity increases
Solution Approach 1:
The patent makes the on-premise nodes multi-functional by enabling them to serve both as compute nodes for virtual machines and as data nodes for the key-value store. This allows a single node to perform multiple functions (computing and data storage), reducing the need for separate dedicated components and thereby reducing overall system complexity.
Solution Approach 2:
The patent changes the operational parameters of existing on-premise nodes by enabling them to participate in the key-value store protocol. Nodes transition from purely compute-oriented to both compute and data storage functions, allowing them to handle high availability tasks locally without requiring completely new hardware architectures.
3Speed
If nodes autonomously manage VM relocation without cloud control plane, then response speed is improved, but coordination complexity increases
Solution Approach 1:
The patent implements a feedback mechanism where nodes continuously monitor the operational status of other nodes and automatically trigger VM relocation when failures are detected. The key-value store provides real-time feedback about node availability, enabling autonomous responses without requiring constant cloud control plane intervention. This feedback loop enables fast local decision-making while maintaining coordination through the distributed database.
Data Source
AI summary
Availability of a virtual machine is managed using a distributed key-value store. The distributed key-value store includes a first entry and a second entry. The first entry represents a definition of the virtual machine, and the second entry represents that a first node of the cluster hosts the virtual machine. Managing availability of the virtual machine includes detecting unavailability of the virtual machine. Managing availability of the virtual machine includes, responsive to detecting unavailability of the virtual machine, writing a task entry to the distributed key-value store to cause a second node of the cluster to create the virtual machine on the second node. Managing availability of the virtual machine includes rewriting the second entry so that the second entry represents that the second node hosts the virtual machine.