Container Image Distribution for Node Failure Recovery
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In container-based systems, when a node fails, quickly starting containers from the failed node on another node is challenging due to resource constraints and the need for downloading container images, which reduces system availability.
Innovation Solution
A method where a processor identifies the node with the largest number of containers and determines if it can start these containers on another node; if not, it instructs a different node to download the container image in advance, ensuring quick startup without wasting resources or time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If container images are downloaded when a node fails and containers need to be restarted, then containers can be started on other nodes, but the startup time increases due to the download process
Solution Approach 1:
The patent applies preliminary action by pre-distributing container images to multiple nodes before failure occurs. When a node fails, the system can immediately start containers on other nodes that already have the required images, eliminating the download wait time and maintaining high availability
Solution Approach 2:
The patent segments container images across multiple nodes rather than centralizing them. Each node stores a subset of container images, creating a distributed cache system that enables fast local startup while reducing the burden on any single node
2Speed
If container images are pre-distributed to all nodes, then container startup is fast, but storage resources are wasted on nodes that don't need all images
Solution Approach 1:
The patent applies local quality by making each node's storage composition specific to its role and needs. Nodes store container images based on their workload requirements and historical failure patterns, rather than uniformly distributing all images to all nodes
Solution Approach 2:
The patent uses partial action by distributing only the necessary subset of container images to each node based on prediction algorithms, rather than distributing the complete set of all possible images. This balances startup speed with storage efficiency
3Reliability
If container images are stored on multiple nodes, then fault tolerance improves, but network bandwidth is consumed during image distribution
Solution Approach 1:
The patent performs image distribution as a preliminary action during normal operation periods when network resources are abundant, rather than during failure recovery when quick action is needed. This separates the distribution phase from the recovery phase, optimizing both scenarios
Data Source
AI summary
A container image arrangement method executed by a processor included in a computer to execute a process, the computer being connected to each of a plurality of nodes, the process includes, identifying a first node that has a first storage storing a container image and has a largest number of containers started from the container image among the plurality of nodes, determining whether the container operating in the first node is capable of starting in a second node among the plurality of nodes other than the first node, where the second has a second storage storing the container image, and storing the container image in a third storage included in a third node different from each of the first node and the second node among the plurality of the nodes when it is determined that the container is not capable of starting in the second node.