Predictive Content Routing via Client Flow Trajectories
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current content distribution systems face inefficiencies due to reliance on geo-location and network proximity methods, which often result in sub-optimal content delivery paths as they do not accurately reflect the client's proximity to content distribution points, leading to longer download times and increased latency.
Innovation Solution
The method involves predicting client flow trajectories to identify the most frequently used peering routers and correlating this information with content caches, using a service routing engine to rank and select the optimal content cache based on both network and geo-proximity, ensuring content is delivered through the nearest and most responsive path.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If geo-location proximity is used to determine content delivery point, then the client's physical location is considered, but the actual network delivery path may be sub-optimal as content may leave the area to find a peering point and then return
Solution Approach 1:
The patent introduces an analytics system as an intermediary that collects and analyzes actual network flow data between clients and content distribution points. This mediator bridges the gap between geo-location data and actual network paths by using observed routing behavior to predict optimal delivery paths, thereby resolving the contradiction between location-based selection and actual delivery efficiency
Solution Approach 2:
The system performs preliminary analysis of network flow trajectories before content delivery occurs. By pre-collecting and analyzing routing data from multiple peering points, the analytics system prepares predictive models that enable optimal content distribution point selection in advance, avoiding sub-optimal paths during actual delivery
2Device complexity
If network proximity from content distributor perspective is used, then the shortest topological path from content distribution nodes is determined, but asymmetric routing in the Internet causes this to differ from the client's optimal path
Solution Approach 1:
Instead of determining proximity from the content distributor's perspective (outbound path), the patent inverts the approach by analyzing inbound traffic flows to the content distribution points. By observing which peering points actually receive traffic from clients, the system determines proximity from the client's perspective, resolving the asymmetry between distributor and client routing paths
Solution Approach 2:
The analytics system implements feedback by continuously monitoring actual network traffic patterns and using this information to refine proximity measurements. The system collects data on real routing behavior and uses this feedback to improve the accuracy of optimal content distribution point selection over time
3Measurement precision
If the client polls hundreds of delivery nodes to assess which is most responsive, then the closest topological delivery node can be determined, but this process is onerous and intractable given the number of delivery nodes and clients
Solution Approach 1:
The patent extracts the complex task of proximity measurement from the client device and relocates it to the analytics system. By removing the polling burden from clients and centralizing the analysis in the analytics system that has access to aggregated flow data, the solution achieves accurate proximity measurement without imposing intractable complexity on individual client devices
Data Source
AI summary
System, method, and computer program product to perform an operation, the operation comprising capturing, at one or more peering routers, parameters for a plurality of data packets sent by a client device and specifying a destination network address, identifying which peering router captured the parameters for each of the plurality of data packets, determining, based on the identified peering routers, a first peering router nearest to the client, relative to the other peering routers, identifying a first content cache, of a plurality of content caches in a content distribution network, nearest to the first peering router, and fulfilling a content request from the client device using content stored on the first content cache.