Conditional Resource Reclamation for High Availability Failover
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High availability systems face inefficiencies in hardware and power utilization due to pre-reserved computing resources for failover being unused and not being dynamically utilized during host failures, leading to suboptimal resource allocation and utilization.
Innovation Solution
Implementing a method that designates computing resources into two sets: one set for immediate availability and another set as conditionally available, allowing low-priority virtual machines to be powered off during host failures to allocate resources to higher-priority virtual machines, thereby enabling efficient resource reclamation and reallocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If computing resources are pre-reserved for failover, then system reliability is improved, but hardware utilization deteriorates
Solution Approach 1:
The patent implements dynamic resource allocation by transitioning computing resources from a static pre-reserved state to a dynamic state where resources are conditionally reserved and can be automatically allocated during failover events. The system monitors host health and dynamically adjusts resource allocation based on actual failover needs, allowing resources to be used by low-priority VMs during normal operation and automatically reclaimed when host failures occur.
Solution Approach 2:
The system changes the reservation parameter of computing resources from 'fully reserved' to 'conditionally reserved'. This parameter change allows the same physical resources to serve dual purposes: supporting low-priority virtual machines during normal operation while maintaining the capability to immediately support failover of high-priority virtual machines when needed, thus resolving the contradiction between reliability and resource utilization.
2Reliability
If spare computing resources are allocated beforehand, then failover capability is improved, but productivity deteriorates
Solution Approach 1:
The patent makes computing resources universal by enabling them to serve multiple functions: during normal operation, conditionally reserved resources are utilized by low-priority virtual machines to contribute to overall productivity, while during failover events, the same resources are automatically reallocated to high-priority virtual machines needing recovery. This multi-functionality resolves the contradiction by allowing resources to be productive in both normal and failure states.
Solution Approach 2:
The system implements self-service through automated failover detection and resource reallocation. When host failures are detected, the system automatically identifies which conditionally reserved resources should be reclaimed and reallocates them to affected high-priority virtual machines without manual intervention, thus maintaining both failover capability and continuous resource utilization.
3Ease of operation
If admission control prevents use of spare resources, then resource allocation control is improved, but hardware utilization deteriorates
Solution Approach 1:
The patent replaces static admission control with dynamic conditional reservation. Instead of permanently blocking resources from being used, the system dynamically adjusts resource availability based on host health status. During normal operation, resources are effectively 'unblocked' for use by low-priority VMs, while automatic monitoring and failover mechanisms ensure resources are quickly reclaimed when needed, thus improving both control and utilization.
Data Source
AI summary
In one embodiment, a method determines a first set of virtual machines and a second set of virtual machines. The first set of virtual machines is associated with a first priority level and the second set of virtual machines is associated with a second priority level. A first set of computing resources and a second set of computing resources are associated with hosts. Upon determining a failure of a host, the method performs: generating a power off request for one or more of the second set of virtual machines powered on the second set of computing resources and generating a power on request for one or more virtual machines from the first set of virtual machines that were powered on the failed host, the power on request powering on the one or more virtual machines from the first set of virtual machines on the second set of computing resources.