Optical Network Effect Graphs for Parallel Simulation Scaling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing optical network simulation tools perform computations in a scalar-based, sequential manner, leading to high latency and inefficiency as network sizes grow, and are not adapted for High Performance Computing techniques.
Innovation Solution
A directed graph representation of the optical network is used to model optical effects at vertices, allowing for parallel computation and batch processing of optical effects, reducing latency and increasing scalability through pre-built static network graphs and parallel processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If scalar-based sequential computation is used for optical network simulation, then implementation simplicity is maintained, but computation time increases significantly and scalability deteriorates
Solution Approach 1:
The patent segments the optical network simulation into independent computational tasks represented as vertices in a directed acyclic graph (DAG). Each vertex corresponds to a specific optical effect calculation (e.g., attenuation, dispersion, amplifier gain) that can be executed independently. This segmentation enables parallel processing across multiple CPU cores or GPUs, reducing total computation time while maintaining implementation clarity through the graph structure.
Solution Approach 2:
The patent transitions from sequential one-dimensional computation to parallel multi-dimensional computation by representing the simulation as a DAG where vertices can be processed simultaneously across different time steps and computational dimensions. This allows batch processing of multiple optical effects and network configurations concurrently, dramatically reducing computation time for large networks.
2Device complexity
If scalar-based sequential computation is used for optical network simulation, then algorithm simplicity is maintained, but scalability with network size deteriorates
Solution Approach 1:
The algorithm segments the network simulation into independent vertices representing different optical effects and uses a DAG structure to manage dependencies. This segmentation allows the simulation to scale linearly with network size, as each additional network element simply adds more vertices to the graph without fundamentally changing the computational approach. The segmented structure enables efficient parallel processing that maintains performance as network size increases.
Solution Approach 2:
The DAG-based framework provides a universal structure that can handle various network configurations, topologies, and optical effects through a single unified algorithm. The same core algorithm processes different network sizes and types by simply varying the graph structure, making the system highly scalable and adaptable without requiring algorithm redesign for different network scenarios.
3Adaptability or versatility
If existing optical network simulation tools are used, then compatibility with current standards is maintained, but adaptability to High Performance Computing techniques deteriorates
Solution Approach 1:
The patent implements dynamic adaptability by designing the simulation framework to automatically adjust its execution model based on available computational resources. The DAG structure can be executed sequentially on single-CPU systems for simplicity, or automatically parallelized across multi-core CPUs, GPUs, or distributed clusters when higher performance is needed. This dynamic execution capability maintains compatibility with existing standards while embracing HPC techniques when beneficial.
Solution Approach 2:
The framework changes execution parameters such as parallelization degree, data granularity, and computation distribution based on the target platform. On HPC systems, the same DAG can be executed with fine-grained parallelism across thousands of cores, while on conventional systems it runs with coarser grain. This parameter adaptability allows the tool to leverage HPC capabilities without requiring fundamental architectural changes or sacrificing compatibility with existing optical network simulation standards.
Data Source
AI summary
Systems and methods include obtaining a processing graph having aggregated vertices and edges from a network graph representing an optical network, wherein one or more vertex and edge in the processing graph represents a grouping of vertices and edges from the network graph based on a given optical effect from a plurality of optical effects; and tracing the processing graph to perform a simulation of one or more optical effects of the plurality of optical effects. Prior to the obtaining, the systems and methods can include forming the network graph based on a topology of the optical network, wherein each vertex in the network graph represent an optical effect in a given component in the optical network and each edge represents propagation between components; tracing the network graph to determine a type and a sequence number; and forming the processing graph based on the traced network graph.