AI Hotspot Scattering for Load Balancing Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing load balancing methods for hotspot URLs do not effectively predict and preprocess potential hotspots, leading to inefficient traffic distribution and increased burden on unified ports, especially with large file sizes, resulting in processing delays and poor user experience.
Innovation Solution
An intelligent hotspot scattering method using an artificial intelligence learning model to predict request quantities and perform proactive scattering operations, distributing requests to multiple cache and origin servers, and redistributing requests to alleviate burden on destination processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If centralized URLs are randomly or evenly distributed to back-end machines using existing technologies, then load balancing is achieved for hotspot URLs, but no judgment or preprocessing is made on the possible trend of hotspot URLs in the early stage, leading to waste of traffic on machines without caches
Solution Approach 1:
The patent applies preliminary action by predicting potential hotspot URLs before they actually become hotspots using an AI learning model. This allows the system to pre-distribute URLs to back-end machines in advance, so that when hotspot requests arrive, the machines already have the URLs in their caches, avoiding traffic waste and improving load balancing efficiency.
2Ease of operation
If all return-to-origin traffics are redirected back to a unified port by using the intra-group cache sharing scheme, then traffic is consolidated, but in the case of a large hotspot file size, the hotspot accumulation is not solved before the machines finish fetching the file, increasing the burden of the unified port
Solution Approach 1:
The patent applies segmentation by dividing the return-to-origin traffic into multiple segments and distributing them to different back-end machines based on AI prediction results. Instead of consolidating all traffic to a single unified port, the system segments the load across multiple machines, reducing the burden on any single port while maintaining efficient cache sharing.
3Productivity
If scattering operation is performed only after hotspot URLs appear in existing technologies, then dispersed distribution of requests is achieved, but processing delays occur and user experience is not optimized
Solution Approach 1:
The patent applies preliminary action by performing the scattering operation in advance based on AI prediction of potential hotspot URLs, rather than waiting for hotspots to appear. This proactive approach allows requests to be distributed to appropriate machines before the hotspots occur, eliminating processing delays and improving user experience.
Data Source
AI summary
An intelligent hotspot scattering method includes learning request quantity curves of a plurality of URLs based on an artificial intelligence learning model and performing request quantity prediction on the plurality of URLs, determining a first URL from the plurality of URLs, determining a second URL from the plurality of URLs, and performing a hotspot scattering operation on the URLs. A predicted request quantity of the first URL is greater than or equal to a first predetermined request quantity threshold corresponding to the first URL. A request quantity of the second URL is not predictable and an actual request quantity of the second URL is greater than or equal to a second predetermined request quantity threshold corresponding to the second URL.