Reserved Instance Allocation via Gale-Shapley Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Cloud-based computational systems face inefficiencies and high costs due to prepaid, discounted resource reservations, where consumers are locked into fixed resource allocations for a period, limiting flexibility and potential cost savings, especially when resource utilization varies.
Innovation Solution
A management system that identifies and ranks reserved instances based on cost savings, affinity, and usage constraints to optimally allocate computational resources across multiple virtual machines, using a Gale-Shapley algorithm to match resources with VMs, ensuring maximum cost savings while maintaining performance targets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If prepaid resource reservations are used to reduce costs, then cost savings are improved, but flexibility and adaptability deteriorate because consumers are locked into fixed resource allocations
Solution Approach 1:
The patent applies dynamics by enabling flexible reconfiguration of reserved computational resources among multiple virtual machines based on changing utilization needs. The system allows dynamic reallocation where VMs can borrow resources from underutilized reserved instances while maintaining cost savings through coordinated resource sharing and return mechanisms.
2Loss of energy
If fixed resource allocations are reserved for a period, then cost savings are improved, but resource utilization efficiency deteriorates when utilization varies
Solution Approach 1:
The patent implements universality by creating a shared resource pool where reserved computational resources serve multiple VMs simultaneously. The system enables one reserved instance to be dynamically allocated to different VMs based on demand, allowing the same physical resources to fulfill multiple functions and serve different workloads at different times.
Solution Approach 2:
The patent applies parameter changes by dynamically adjusting resource allocation parameters (CPU, memory, storage) among VMs based on utilization metrics. The system monitors resource usage and automatically reconfigures allocations to optimize both cost savings and utilization efficiency, changing allocation parameters in response to varying workload demands.
3Loss of energy
If local optimization is used to choose cheapest instance types, then cost savings are improved, but the ability to handle demand fluctuations and maintain performance deteriorates
Solution Approach 1:
The patent introduces an intermediary resource management layer that mediates between cost optimization and performance requirements. This intermediary system coordinates resource borrowing and returning among VMs, ensuring that performance targets are met while maintaining cost efficiency through intelligent resource allocation and reallocation decisions.
Data Source
AI summary
Various approaches for allocating resources to multiple virtual machines include identifying multiple reserved instances, each specifying a quantity of one or more computational resources compatible with the feasible resource template for the VMs; computationally generating, for each of the VMs, an instance-ranking list indicating a ranking order of the reserved instances having templates feasible for the VM; computationally generating, for each of the reserved instances, a VM-ranking list indicating a ranking order of the VMs to which the resources specified by the reserved instance may be allocated; and based at least in part on the instance-ranking list and the VM-ranking list, computationally mapping each of the VMs to one of the reserved instances.