Cloud Resource Allocation via Predictive Forecasting and Failover Amortization

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Traditional auto-scaling strategies in cloud-based networks are inefficient due to their reactive nature, slow response to load changes, and reliance on fixed scaling rules, leading to overprovisioning costs and potential service latency or unavailability issues.

Innovation Solution

A proactive and reactive resource allocation system that uses predictive scheduling and failover amortization to dynamically adjust resource allocation based on forecasted metrics and performance count values, prioritizing resource allocation across multiple geographical regions to optimize resource usage and minimize downtime.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional reactive auto-scaling strategies are used, then resource allocation is adjusted after load changes are detected, but the response is slow and leads to overprovisioning costs

Engineering Contradiction:

Improveservice availabilityVSAvoidoverprovisioning costs

Core Design Contradiction:

ReliabilityVSLoss of energy

Solution Approach 1:

The system performs preliminary actions by using predictive algorithms to forecast future resource needs before actual load changes occur. The resource allocation system receives forecast data indicating predicted metric values and proactively adjusts resource allocation in advance, rather than reacting after load changes are detected. This eliminates the slow response time of traditional reactive strategies while avoiding overprovisioning costs.

Inventive Principle:

Principle #10Preliminary action

2Ease of manufacture

If fixed scaling rules are used, then resource allocation is simplified, but the system cannot adapt to dynamic load changes and causes service latency

Engineering Contradiction:

Improveallocation simplicityVSAvoidservice performance

Core Design Contradiction:

Ease of manufactureVSReliability

Solution Approach 1:

The system replaces fixed scaling rules with dynamic predictive algorithms that continuously learn from historical data and adapt to changing load patterns. The resource allocation system uses forecast data and performance count values to dynamically adjust resource allocation, enabling the system to adapt to dynamic load changes while maintaining service performance. The prioritization mechanism dynamically weighs different allocation strategies based on current conditions.

Inventive Principle:

Principle #15Dynamics

3Productivity

If proactive resource allocation based on forecast data is used, then resource usage is optimized, but the system complexity increases

Engineering Contradiction:

Improveresource allocation efficiencyVSAvoidsystem complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The system introduces an intermediary layer of predictive algorithms and forecast servers that simplify the complexity of proactive resource allocation. The resource allocation system receives pre-processed forecast data from forecast servers, reducing the complexity of processing raw historical data. The prioritization mechanism acts as an intermediary that consolidates multiple allocation strategies into a unified decision, making the system more manageable despite the advanced predictive capabilities.

Inventive Principle:

Principle #24Intermediary (Mediator)

4Reliability

If failover resource allocation is implemented, then service continuity is improved during failures, but the resource allocation amount increases

Engineering Contradiction:

Improveservice continuityVSAvoidresource allocation amount

Core Design Contradiction:

ReliabilityVSQuantity of substance

Solution Approach 1:

The system implements preliminary failover preparation by calculating failover resource amounts in advance based on predicted failures and historical data. The resource allocation system determines a failover resource amount to cover at least a portion of load in case of failure, rather than reacting after failures occur. This proactive approach ensures service continuity while optimizing the resource allocation amount by only allocating what is needed for potential failover scenarios.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS10873541B2Systems and methods for proactively and reactively allocating resources in cloud-based networks

Publication Date: 2020.12.22 MICROSOFT TECHNOLOGY LICENSING LLC

AI summary

A resource allocation system is provided and includes a processor, a memory, and an application including instructions configured to: receive forecast data from a forecast server computer indicating a predicted metric value corresponding to a cloud-based service for a first geographical region; determine an expected usage amount for the first geographical region based on the predicted metric value; determine a failover resource amount to cover a portion of a load in a second geographical region due to a failure; determine a predicted resource allocation value based on the expected usage and failover resource amounts; determine a reactive resource allocation value based on the predicted metric value or a parameter, where the parameter corresponds to access of cloud-based resources for the cloud-based service; prioritize the predicted and reactive resource allocation values; and adjust a resource allocation amount for the first geographical region over time based on a result of the prioritization.