GPU-Optimized User Profile Data Structures for Campaign Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Cloud-based applications face computational burdens when processing large-scale user profiles due to the complexity and number of user interests, making CPU-based processing inefficient.
Innovation Solution
Offload user profile processing tasks to GPUs using specialized data structures that facilitate parallel processing, including organizing user and campaign logic tree attributes in a manner consumable by GPUs, enabling efficient evaluation of user profiles against campaign logic trees.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If CPU-based processing is used for user profiles, then processing can be performed, but computational efficiency deteriorates due to the scale and complexity of big data user profiles
Solution Approach 1:
The patent segments user profile data into specialized GPU data structures including separate arrays for user IDs, interest tags, and campaign logic tree nodes. This segmentation allows the GPU to process different data elements in parallel, resolving the contradiction by organizing the large quantity of data into manageable, parallelizable units that maintain high computational efficiency.
Solution Approach 2:
The patent transitions from traditional CPU-based sequential processing to GPU-based parallel processing by reorganizing data into multi-dimensional structures suitable for GPU architecture. This includes creating interleaved arrays and hierarchical tree structures that enable simultaneous processing across multiple data dimensions, thereby maintaining productivity despite the increasing quantity of user profile data.
2Speed
If traditional data structures are used, then data can be stored, but processing speed deteriorates when evaluated by GPU parallel operations
Solution Approach 1:
The patent applies preliminary action by pre-processing and organizing user profile data into GPU-optimized data structures before processing. This includes pre-segmenting interest tags, pre-organizing campaign logic trees, and pre-interleaving data arrays to match GPU processing patterns. This preliminary organization enables the GPU to process data at maximum speed without being hindered by structural complexity during execution.
Solution Approach 2:
The patent changes the structural parameters of data storage from traditional formats to GPU-specific formats. This includes changing from hierarchical database structures to flat interleaved arrays, from single-format storage to multi-format specialized structures, and from CPU-optimized to GPU-optimized data layouts. These parameter changes enable faster processing speed while managing data structure complexity through systematic reorganization.
3Productivity
If CPU processing is used for campaign logic tree evaluation, then evaluation can be performed, but the computational load increases significantly with millions of user profiles
Solution Approach 1:
The patent extracts the computationally intensive campaign logic tree evaluation task from the CPU and relocates it to the GPU. This is achieved by creating separate GPU-optimized data structures for campaign logic trees and user profiles, allowing the GPU to handle the heavy computational load of evaluating millions of profiles against campaign criteria, thereby reducing CPU computational load and increasing overall evaluation throughput.
Solution Approach 2:
The patent creates copies of user profile data and campaign logic tree data in GPU-optimized formats. This includes copying data from traditional storage into specialized GPU data structures such as interleaved arrays and hierarchical trees that are tailored for parallel processing. These copies enable the GPU to perform evaluations independently and efficiently, significantly increasing throughput while reducing the computational burden on the CPU.
Data Source
AI summary
Disclosed are techniques for processing user profiles using data structures that are specialized for processing by a GPU. More particularly, the disclosed techniques relate to systems and methods for evaluating characteristics of user profiles to determine whether to offload certain user profiles to the GPU for processing or to process the user profiles locally by one or more central processing units (CPUs). Processing user profiles may include comparing the interest tags included in the user profiles with logic trees, for example, logic trees representing marketing campaigns, to identify user profiles that match the campaigns.