ML-Based Cloud Request Routing Across Regions

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

In cloud computing environments, existing technologies face challenges in dynamically balancing user requests across regions, leading to performance degradation when one region experiences high demand while others have excess capacity, as requests are preassigned and cannot be dynamically rerouted to alleviate overload.

Innovation Solution

A machine learning model is employed to distribute user requests across cloud computing systems based on validity rules and resource utilization, continually trained to optimize routing and balance load across regions, allowing for dynamic redistribution of requests to improve overall performance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If user requests are preassigned to specific regions, then request routing is simple and deterministic, but load balancing performance deteriorates when demand fluctuates across regions

Engineering Contradiction:

Improverequest routing simplicityVSAvoidload balancing performance

Core Design Contradiction:

Ease of operationVSProductivity

Solution Approach 1:

The patent implements dynamic request routing by transitioning from static preassignment to a machine learning-based system that continuously adapts routing decisions based on real-time resource utilization metrics and historical patterns, enabling the system to respond dynamically to demand fluctuations across regions

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system incorporates feedback mechanisms by collecting resource utilization data from cloud computing systems across regions and using this feedback to continuously train and update the machine learning model, which then optimizes routing decisions based on learned patterns from historical performance data

Inventive Principle:

Principle #23Feedback

2Device complexity

If requests are statically routed to regions, then system complexity is reduced, but service level agreement compliance worsens during demand fluctuations

Engineering Contradiction:

Improvesystem complexityVSAvoidservice level agreement compliance

Core Design Contradiction:

Device complexityVSReliability

Solution Approach 1:

The patent introduces a machine learning model as an intermediary layer between request sources and cloud computing regions, which processes routing decisions based on learned patterns from historical data and real-time metrics, thereby managing complexity while improving SLA compliance through intelligent mediation

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system performs preliminary actions by continuously training the machine learning model on historical resource utilization data before actual demand fluctuations occur, enabling the model to predict and prepare optimal routing strategies in advance, thus improving SLA compliance during demand variations

Inventive Principle:

Principle #10Preliminary action

3Adaptability or versatility

If cloud computing systems operate independently in their assigned regions, then regional autonomy is maintained, but overall environment performance deteriorates due to inability to redistribute capacity

Engineering Contradiction:

Improveregional autonomyVSAvoidoverall environment performance

Core Design Contradiction:

Adaptability or versatilityVSProductivity

Solution Approach 1:

The patent implements universality by creating a centralized machine learning-based routing system that serves multiple regions simultaneously, enabling the system to dynamically allocate requests across different regions based on overall capacity rather than rigid regional assignments, thus improving overall environment performance while maintaining regional operational independence

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS11509746B2Distributing user requests to cloud computing systems across regions using a machine learning model

Publication Date: 2022.11.22 SALESFORCE INC

AI summary

Systems and methods are described for receiving a user request from a user computing system located in a first geographic region, generating a user request context for the user request, selecting a cloud computing system assigned to a second geographic region having more computing resources currently available to process the user request than a cloud computing system assigned to the first geographic region based at least in part on the user request context and a machine learning model including current utilizations of computing resources of cloud computing systems assigned to the first and second geographic regions, and sending the user request to the selected cloud computing system in the second geographic region. The systems and methods further include getting current utilizations of computing resources from cloud computing systems assigned to a plurality of regions, aggregating user request contexts and the current utilizations of computing resources; and updating the machine learning model with the aggregated user resource contexts and the aggregated current utilizations of computing resources of cloud computing systems.