Hybrid VM Allocation Using Demand Forecast and Eviction Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems struggle to determine optimal allocation plans for virtual machines (VMs) in cloud computing platforms that consider multiple allocation types with different pricing models and user satisfaction, failing to adapt to rapid demand changes and complex workload scenarios.
Innovation Solution
A data processing system using a causal inference model to correlate user experience with VM utilization, combined with demand forecasting and spot VM eviction modeling, to optimize VM allocation through mixed-integer optimization, considering various allocation types and user satisfaction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple VM allocation types with different pricing models are used, then cost efficiency and flexibility are improved, but allocation plan complexity increases
Solution Approach 1:
The patent segments the VM allocation problem by separating different allocation types (spot VMs, reserved VMs, on-demand VMs) into distinct categories, each with its own pricing model and characteristics. This segmentation allows the system to handle each type independently while optimizing the overall allocation plan across all types simultaneously.
Solution Approach 2:
The patent changes parameters by introducing multiple pricing model parameters (spot pricing, reserved pricing, on-demand pricing) and allocation type parameters. The optimization system dynamically adjusts these parameters based on demand forecasts, historical data, and current allocation states to determine the optimal mix of different VM allocation types.
2Reliability
If user satisfaction is considered in allocation optimization, then service quality is improved, but computational complexity increases
Solution Approach 1:
The patent introduces an intermediary component that translates user satisfaction requirements into quantifiable metrics and constraints for the optimization algorithm. This intermediary layer processes qualitative service quality requirements and converts them into mathematical parameters that can be incorporated into the mixed-integer optimization model without directly increasing computational complexity.
Solution Approach 2:
The patent replaces manual or rule-based service quality assessment with an automated optimization system that uses machine learning models and mathematical programming. This substitution transforms the complex task of balancing user satisfaction with resource allocation into a computationally tractable optimization problem with defined objective functions and constraints.
3Productivity
If demand forecasting and dynamic optimization are implemented, then responsiveness to demand changes is improved, but system complexity increases
Solution Approach 1:
The patent implements preliminary action by performing demand forecasting in advance using historical data and machine learning models. The system predicts future demand patterns and pre-computes allocation strategies before actual demand occurs, enabling proactive rather than reactive resource allocation. This preliminary analysis reduces the complexity of real-time decision-making.
Solution Approach 2:
The patent maintains continuity of useful action by implementing dynamic optimization that continuously monitors demand changes and adjusts allocation plans in real-time. The system operates as an ongoing process rather than periodic batch processing, continuously refining allocation decisions based on new data while maintaining optimal resource utilization throughout the allocation period.
Data Source
AI summary
Systems and methods for determining an allocation plan for allocating virtual machines (VMs) to a hosted service utilizes a demand forecast that predicts future demand for VMs, a response curve that correlates user experience to VM utilization, and an estimated spot eviction rate to determine the allocation plan. The demand forecast, the response curve and the eviction rate are processed using mixed-integer optimization to determine the numbers of VMs of each allocation type that should be online at any given time to meet demand.