VM Allocation via HDD IOPS Prediction to Reduce Interference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Cloud computing systems face significant challenges in efficiently managing hard disk drive resources for virtual machines, leading to underutilization of infrastructure, which results in substantial monetary costs due to low CPU utilization.
Innovation Solution
A resource management system that includes a prediction engine and a virtual machine allocator, leveraging historical data to predict future VM and tenant behaviors, allowing for informed decisions on VM placement and resource allocation, such as grouping VMs based on predicted resource utilization and lifetime, and optimizing cluster selection to minimize interference and maximize resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If virtual machines are allocated without prediction of resource usage patterns, then allocation speed is fast, but resource utilization efficiency deteriorates
Solution Approach 1:
The system performs preliminary actions by collecting historical resource usage data and training prediction models before actual VM allocation occurs. The prediction engine pre-analyzes resource usage patterns, enabling informed allocation decisions that improve utilization efficiency without adding complexity to the real-time allocation process
Solution Approach 2:
A prediction engine is introduced as an intermediary component between the VM allocation system and historical data. This mediator analyzes past resource usage patterns and provides predictions to the allocation system, enabling efficient resource utilization without requiring the allocation system itself to become more complex
2Loss of energy
If multiple virtual machines are co-located on the same host, then infrastructure cost is reduced, but performance interference increases
Solution Approach 1:
The system applies local quality by making allocation decisions specific to each host's characteristics and the particular VM's resource usage patterns. The prediction engine analyzes local conditions at each host and matches them with VM requirements, enabling efficient co-location while avoiding performance interference through localized optimization
Solution Approach 2:
The system changes parameters by using prediction models that analyze multiple resource usage parameters (CPU, memory, storage, network) to determine optimal VM placement. By dynamically adjusting allocation based on predicted parameter interactions, the system reduces infrastructure costs through co-location while maintaining performance stability
3Adaptability or versatility
If cloud infrastructure is continuously updated with newer hardware, then system capability is improved, but underutilization cost increases
Solution Approach 1:
The system applies dynamics by implementing adaptive allocation that continuously learns from resource usage patterns and adjusts VM placement strategies accordingly. The prediction engine dynamically updates its models based on changing workloads and hardware performance, ensuring optimal utilization of updated infrastructure while maintaining system capability
Solution Approach 2:
A feedback mechanism is implemented where the prediction engine continuously monitors actual resource usage against predictions, learns from discrepancies, and refines allocation strategies. This feedback loop ensures that newly added hardware is efficiently utilized by adjusting VM placement to match actual usage patterns, reducing underutilization costs while preserving system capability
Data Source
Figure 1~2
Figure 3
Figure 4~6
AI summary
A system receives a request to deploy a virtual machine on a node from a plurality of nodes running a plurality of virtual machines in a cloud computing system. The system selects one of the plurality of nodes having a hard disk drive (HDD) input output operations per second (IOPS) value less than an observed HDD IOPS value for the plurality of nodes running the plurality of virtual machines. The system receives a predicted HDD IOPS value for the virtual machine and determines a new HDD IOPS value for the selected node based on the HDD IOPS value for the selected node and the predicted HDD IOPS value for the virtual machine. The system instantiates the virtual machine on the selected node when the new HDD IOPS value for the selected node is less than or equal to the observed HDD IOPS value for the plurality of nodes.