Parallelized Graph Partitioning for Approximate Max-K Cut
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing graph database systems struggle with performance when analyzing large graphs or sub-graphs, as the machinery implementing analytics algorithms is not sufficiently performant and difficult to optimize.
Innovation Solution
A graph database system that performs parallelized graph partitioning using multiple threads to execute approximate max-k cut processing, employing greedy random construction, local search, and path relinking to maximize the weight of edges across partitions, with parallel computation to reduce processing time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If sequential graph partitioning algorithms are used, then algorithm simplicity is maintained, but processing time increases significantly for large graphs
Solution Approach 1:
The graph partitioning algorithm is divided into independent tasks that can be executed in parallel. Each thread handles a subset of nodes and edges, allowing concurrent processing. The system segments the computational workload to achieve parallel execution, significantly reducing processing time for large graphs while maintaining algorithm correctness through coordinated task execution.
Solution Approach 2:
The patent introduces a temporal dimension to the processing by executing partitioning operations concurrently across multiple threads. Instead of sequential execution, the system leverages parallel time execution to process graph partitions simultaneously, transforming a single-dimensional sequential algorithm into a multi-dimensional parallel system that achieves faster processing speeds.
2Loss of time
If approximate max-k cut processing is performed, then processing time is reduced, but partition accuracy may be compromised
Solution Approach 1:
The algorithm performs partial max-k cut processing by executing approximate computations that yield sufficiently accurate results without requiring exhaustive optimization. The parallel processing framework enables the system to perform enough computational actions to achieve acceptable partition accuracy while significantly reducing the time required compared to exact methods. The excessive action principle is applied by allowing threads to perform computations independently and collecting results that collectively provide accurate partitions.
Solution Approach 2:
Multiple threads create independent copies of the partitioning process, each working on different subsets of the graph. These parallel copies execute the same algorithmic logic simultaneously on different data partitions, allowing the system to achieve accurate results faster by distributing the computational workload across multiple independent execution paths that converge on the final partitioning solution.
3Adaptability or versatility
If graph analytics algorithms are implemented, then analytical capability is improved, but system performance deteriorates
Solution Approach 1:
The parallel processing framework is designed to be universally applicable to various graph analytics algorithms. The same parallel execution infrastructure supports different analytical operations (max-k cut, community detection, graph traversal, etc.), allowing the system to maintain high analytical capability while achieving improved performance through parallelization. The framework's multi-functionality enables it to handle diverse algorithms without requiring separate optimization for each.
Solution Approach 2:
The system dynamically adjusts the parallel processing configuration based on the specific algorithm being executed and the characteristics of the input graph. The framework can adapt the number of threads, task distribution strategy, and parallelization degree to optimize performance for different analytical operations. This dynamic adaptation allows the system to maintain high performance across various graph analytics tasks while preserving full analytical capability.
Data Source
AI summary
A method and high-performance apparatus for creating an approximate maximum k-cut of a graph. In various embodiments, nodes and weighted edges in a graph are used to compute a partitioning of a graph such that the edge weight between partitions is maximized. In various embodiments, the method and apparatus use a greedy random construction of cuts for its first approximation, whose solutions are then subject to local search for better solutions, and local path-relinking for the final level of refinement. In various embodiments, one or more of the greedy random construction, local search, and path relinking are parallelized. A novel parallel algorithm is disclosed for perturbing the state of a cut to see if a better one is cheaply available.