Cluster-Wide View for High Availability Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computer system management tools are inadequate in efficiently managing large data centers with numerous virtual machines (VMs) and virtual input/output servers (VIOSes), leading to reduced high availability (HA) effectiveness and increased burdens on control nodes, which limits response times and recovery options during disasters or outages.
Innovation Solution
Implementing a control node that communicates with VIOSes to monitor the health and status of VMs, allowing the control node to manage only the VIOSes while the VIOSes manage the VMs, and utilizing a shared storage pool cluster for geographically dispersed resiliency to automate disaster recovery operations and improve HA.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the control node monitors and manages all VMs directly, then comprehensive monitoring coverage is achieved, but the control node's workload increases significantly
Solution Approach 1:
The system divides the monitoring function into two segments: VIOS-level monitoring agents that collect VM status locally, and a control node that aggregates this data. This segmentation reduces the control node's direct monitoring burden while maintaining comprehensive coverage across all VMs through hierarchical data collection.
Solution Approach 2:
VIOSes act as intermediary components between VMs and the control node. The VIOS monitoring agents collect detailed VM status information and present it to the control node, which then manages HA decisions. This intermediary layer filters and aggregates data, reducing the control node's processing workload while preserving monitoring precision.
2Speed
If more control node resources are allocated for monitoring, then response time improves, but system cost increases
Solution Approach 1:
The VIOS monitoring agents perform self-service by autonomously collecting VM status data and maintaining local records. This eliminates the need for the control node to actively query each VM, reducing response time for status detection without requiring additional control node resources or increasing system cost.
Solution Approach 2:
The system implements continuous monitoring through VIOS agents that maintain up-to-date VM status records without requiring periodic control node interventions. This continuous action ensures fast response times while keeping resource allocation constant, avoiding the need to increase system cost for improved speed.
3Ease of operation
If the control node manages a large number of VMs directly, then centralized control is maintained, but recovery options during disasters are limited
Solution Approach 1:
The system segments control authority between VIOS-level agents that manage local VM status and the control node that handles high-level HA decisions. This segmentation enables the control node to maintain centralized control policies while VIOS agents provide distributed knowledge of local conditions, expanding disaster recovery options without sacrificing centralized coordination.
Solution Approach 2:
The VIOS monitoring agents provide local quality by maintaining detailed, VM-specific status records at the source. This local knowledge enables more informed disaster recovery decisions while the control node maintains centralized control policies. The combination of local detailed knowledge and centralized policy management expands recovery options while preserving ease of operation.
Data Source
AI summary
A method of administering a shared pool of computer system resources by a control node by monitoring a first cluster with a first and second node and a second cluster with a third and fourth node. The method also includes maintaining, by the first cluster, a record of parameters and maintaining, by the second cluster, a record of parameters. The method also includes monitoring, by the control node, a database of the records. The method also includes determining, by the control node, whether the parameter associated with write traffic between the first and third node exceeds a threshold value, and responsive to the parameter exceeding the threshold value, instructing, by the control node, the second node to be a responding node. The method also includes notifying a user to check the computer system resources when the parameter exceeds the threshold value.