Edge Server Page Personalization via ESI Feature Vector Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing edge computing methods for personalization rely heavily on centralized decision-making, leading to poor performance and high costs due to reduced caching capacity, which limits the improvement of end-user quality of service (QoS) in regions with low bandwidth and poor networking conditions.
Innovation Solution
Implementing a system that performs page personalization at the edge server by defining feature vectors for users and content modules, using a switch statement in Edge Side Includes (ESI) to select the closest matching content modules for caching and delivery, reducing dependency on central services and enabling edge caching of page structures and ESI commands.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If centralized decision-making is used for personalization, then personalization can be achieved, but the amount of cacheable content is greatly reduced and performance deteriorates
Solution Approach 1:
The patent segments the personalization decision-making process by introducing ESI (Edge Side Includes) commands that allow different parts of a webpage to be independently personalized at the edge server. Content modules are divided into separate includable units that can be cached and assembled dynamically, resolving the contradiction between personalization and caching capacity.
Solution Approach 2:
The patent introduces an intermediary layer (edge server with ESI processing) between the centralized personalization service and the end user. This intermediary enables local caching of content modules while maintaining personalized assembly, thus improving performance without sacrificing personalization capability.
2Reliability
If edge computing is employed to reduce network data volume, then QoS improves, but personalization services suffer due to reduced caching capacity
Solution Approach 1:
The patent implements nested ESI structures where ESI commands can be embedded within content modules that are themselves included via ESI. This nested architecture allows multi-level caching and assembly at the edge, enabling both QoS improvement through reduced network traffic and maintained personalization service quality through hierarchical content assembly.
3Speed
If more content is cached at the edge, then delivery speed improves, but the ability to provide personalized content decreases
Solution Approach 1:
The patent introduces dynamic content assembly at the edge server using ESI processing. While content modules are statically cached for fast delivery, their assembly and selection are dynamically determined based on user-specific parameters. This dynamic assembly process enables both fast delivery (through caching) and personalized content composition.
4Stability of the object's composition
If centralized services are used for page assembly, then content coherence is maintained, but time dependency increases and reliability decreases
Solution Approach 1:
The patent performs preliminary actions by pre-caching content modules and ESI command structures at the edge server before they are needed. The page assembly logic and matching rules are pre-configured in the ESI markup, enabling the edge server to autonomously assemble personalized pages without real-time centralized intervention, thus improving reliability while maintaining coherence.
Data Source
AI summary
A method is provided for performing page personalization at an edge server. In response to a page request, page mark-up is retrieved from a page server, the page mark-up including embedded edge server instructions. A plurality of content feature vectors are parsed from the embedded edge server instructions, the plurality of content feature vectors being respectively associated with a plurality of content modules. A user feature vector associated with the user is determined. The user feature vector is matched against the plurality of content feature vectors to determine a closest matching content feature vector to the user feature vector. A selected content module associated with the closest matching content feature vector is retrieved. The selected content module is combined with the page mark-up to define an edge-personalized page mark-up.