Cluster Management via Request Transfer Pods
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional cluster switching methods are slow and incomplete due to the difficulty in simultaneously updating DNS records across multiple DNS servers, leading to incomplete switching and inability to perform application-level switching.
Innovation Solution
A cluster management method that transfers requests from a target program to a substitute program in a substitute cluster, allowing for immediate switching by deploying transfer pods and updating DNS records to direct client terminals to the substitute cluster.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If DNS records are updated across multiple DNS servers to switch clusters, then cluster switching can be performed, but the switching process becomes slow and incomplete due to the difficulty in simultaneously updating all DNS servers
Solution Approach 1:
The patent introduces a load balancer as an intermediary component that sits between client terminals and the cluster. The load balancer maintains a mapping relationship with worker nodes and can redirect requests to substitute worker nodes in a substitute cluster without requiring DNS record updates. This mediator approach allows cluster switching to be performed at the application layer rather than at the network infrastructure layer, eliminating the need to update multiple DNS servers simultaneously.
Solution Approach 2:
The patent segments the cluster into multiple independent units called worker nodes, each capable of being substituted independently. When switching clusters, only the necessary worker nodes need to be substituted rather than requiring system-wide DNS changes. This segmentation allows partial switching to proceed while other parts of the system remain operational, reducing both time loss and improving completeness.
2Speed
If DNS records are updated to switch from first cluster to second cluster, then routing can be directed to the new cluster, but switching is not completed immediately because client terminals cache DNS records
Solution Approach 1:
The load balancer acts as an intermediary that directly controls request routing without relying on DNS resolution. Since the load balancer maintains its own mapping between domain names and worker node IP addresses, it can immediately redirect requests to substitute worker nodes without waiting for DNS cache invalidation. This eliminates the cache invalidation time delay that would otherwise prevent immediate switching.
3Adaptability or versatility
If conventional cluster switching methods are used, then DNS records can be updated, but application-level switching cannot be performed and switching is slow
Solution Approach 1:
The patent implements dynamic substitution of worker nodes based on real-time needs. The load balancer can dynamically redirect requests from active worker nodes to substitute worker nodes in the substitute cluster without requiring predefined DNS changes or system-wide reconfiguration. This dynamic approach enables application-level switching to be performed efficiently by substituting only the necessary worker nodes rather than performing complete cluster reconfiguration.
Data Source
AI summary
A cluster includes a plurality of nodes including a storage unit that stores a program to be executed in response to a request from a client terminal connected to a network, and a processor that executes the program, in which, in a case where at least one substitute program stored in a substitute cluster is executed as a substitute for at least one target program stored in the cluster, when a request to the target program transmitted from the client terminal is acquired, a request transfer process of transferring the request to the target program to the substitute cluster is executed such that the substitute program is executed in response to the request to the target program.