Adaptive CDN Selection via Score-Based Random Draw
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for selecting content delivery networks (CDNs) for data segment retrieval are inflexible and fail to adapt to changing transmission quality over time, leading to sub-optimal choices in the medium to long term.
Innovation Solution
A method that calculates a score for each CDN based on transmission quality, uses a random draw to select the CDN, and adjusts probabilities based on an exploration parameter, incorporating bandwidth and transmission error heuristics to ensure flexible and adaptive selection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a client device always selects the CDN with the highest current transmission quality score, then short-term transmission quality is optimized, but medium to long-term adaptability deteriorates because the CDN may become sub-optimal over time
Solution Approach 1:
The patent implements dynamic CDN selection by introducing an exploration parameter that adjusts the randomness of CDN selection over time. The selection process transitions from highly exploratory (trying multiple CDNs) to more exploitative (sticking with best-performing CDN) based on accumulated performance data, allowing the system to adapt to changing transmission conditions while maintaining quality
Solution Approach 2:
The system continuously monitors transmission quality from selected CDNs and uses this feedback to update scores and adjust the exploration parameter. This closed-loop feedback mechanism enables the client device to learn from past performance and adapt its selection strategy, resolving the contradiction between maintaining high quality and adapting to changing conditions
2Adaptability or versatility
If a client device uses a purely random selection among multiple CDNs, then adaptability is improved, but transmission quality deteriorates due to ignoring known performance differences
Solution Approach 1:
The patent changes the parameter of selection probability by introducing an exploration parameter that modulates the randomness. Instead of uniform random selection or always selecting the best, the system adjusts selection probabilities dynamically based on the exploration parameter and accumulated scores, balancing exploration and exploitation to maintain both adaptability and quality
3Adaptability or versatility
If a client device frequently switches between CDNs to find the best one, then adaptability is improved, but stability of data retrieval deteriorates due to frequent changes
Solution Approach 1:
The exploration parameter dynamically controls the frequency and randomness of CDN switching. When the parameter indicates high exploration, the system tries different CDNs; when it indicates low exploration, the system stabilizes on the current best-performing CDN. This dynamic control resolves the contradiction between adaptability and stability
Solution Approach 2:
The system accumulates performance scores from multiple CDNs beforehand, creating a buffer of knowledge about relative performance. This cushioning effect allows the system to make informed switching decisions rather than random switches, maintaining stability while preserving adaptability
Data Source
AI summary
There is proposed a method for obtaining a data segment by a client device capable of communicating with a plurality of content delivery networks (CDN1, CDN2), the method comprising the following steps: for each network of the plurality of networks, calculating (100) a score associated with the network indicative of a quality of transmission of data from the network to the client device; randomly drawing (102) a network in the plurality of networks, the random draw being adapted so that a probability of drawing a network of the plurality associated with a score indicative of a first quality of transmission is higher than a probability of drawing a network of the plurality associated with a score indicative of a second quality of transmission lower than the first quality of transmission; and requiring (104) the drawn network for the client device to obtain the data segment from the drawn network.