VM Placement via Infrastructure Utilization Metrics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing distributed computing systems face inefficiencies in deploying and utilizing physical host machines, leading to substantial costs due to the lack of effective resource management, as they require pre-estimation of application types and machine numbers, and load balancers do not sufficiently optimize machine deployment.
Innovation Solution
A resource management node determines operational resources needed for guest VMs, generates placement scenarios for physical host machines within cabinets, and initiates placements based on defined rules to optimize infrastructure usage, ensuring efficient deployment and utilization of resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If pre-estimation of application types and machine numbers is used for deployment, then deployment planning can be performed in advance, but the system cannot adapt to actual infrastructure utilization requirements, leading to inefficient resource allocation
Solution Approach 1:
The system continuously monitors actual infrastructure utilization metrics (power consumption, cooling requirements, space usage) and uses this feedback to dynamically adjust placement decisions. The resource management node receives utilization data, compares it against placement scenarios, and modifies deployments to match actual needs rather than pre-estimates, resolving the contradiction between adaptability and deployment efficiency
Solution Approach 2:
The system changes the parameters used for deployment decisions from static pre-estimated values to dynamic actual utilization metrics. By monitoring real-time infrastructure parameters (power, cooling, space) and using these actual values to drive placement scenarios, the system achieves both efficient resource allocation and adaptability to changing conditions
2Productivity
If load balancers are used to direct requests to physical host machines, then system utilization can be balanced, but the efficiency of physical host machine deployment and usage is not sufficiently improved
Solution Approach 1:
The system performs preliminary placement scenario generation that includes both VM-to-host mapping and host-to-cabinet placement decisions before deployment occurs. By pre-determining optimal placements based on actual infrastructure utilization metrics rather than relying on post-deployment load balancing, the system improves host machine utilization efficiency and reduces the number of physical hosts needed
Solution Approach 2:
The resource management node acts as an intermediary between infrastructure monitoring systems and deployment systems. It receives actual utilization data from infrastructure monitors, processes this information to generate optimized placement scenarios, and translates these scenarios into deployment actions, thereby improving overall host machine utilization efficiency
3Reliability
If more physical host machines are deployed to support applications, then application requirements can be met, but infrastructure costs increase substantially
Solution Approach 1:
The system changes the metrics used for deployment decisions from generic application requirements to specific actual infrastructure utilization parameters (power consumption, cooling needs, space requirements). By matching the number and type of physical hosts to actual measured infrastructure usage rather than application estimates, the system maintains application availability while significantly reducing infrastructure costs and energy consumption
Solution Approach 2:
The system uses feedback from actual infrastructure utilization monitoring to adjust deployment strategies. By continuously comparing actual infrastructure usage against deployment configurations, the system can optimize the balance between maintaining application service availability and reducing infrastructure costs, eliminating waste from over-provisioning
Data Source
AI summary
A resource management node includes a processor and a memory coupled to the processor. The memory includes computer readable program code that when executed by the processor causes the processor to perform operations. The operations can include, for each of a plurality of guest virtual machines (VMs), determining operational resources of physical host machines available in a distributed computing system that are needed to provide the guest VM. A placement scenario for placing the guest VMs on the physical host machines and placing the physical host machines in cabinets of a distributed computing system is generated. An amount of infrastructure of the distributed computing system used by the placement scenario is determined. A determination is made whether the placement scenario satisfies a defined rule for how much infrastructure of the distributed computing system can be used. Placement of the physical host machines in the cabinets and placement of the guest VMs on the physical host machines is initiated based on the placement scenario satisfying the defined rule.