Vector Space NSP Mapping for SDI Service Orchestration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The complexity of managing Software Defined Infrastructures (SDIs) increases due to the high number of Network and Service Primitives (NSPs) that need to be interconnected and continuously updated, resulting in a heavy computational burden for management, control, and orchestration processes.
Innovation Solution
A system that translates SDI service requests into a corresponding service chain of NSPs, maps these NSPs onto a multidimensional vector space using a space matrix, and employs artificial intelligence learning methods to predict optimized sequences of NSPs, reducing complexity by leveraging geometric operations and recurrent neural networks for efficient resource management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If Network and Service Primitives (NSPs) are used to model SDI services, then service flexibility and programmability are improved, but system complexity and computational burden increase
Solution Approach 1:
The patent introduces a Service Space Module as an intermediary between the NSP translation module and the control module. This module maps NSPs onto a multidimensional vector space and performs geometric operations to predict optimized service chains, thereby reducing the direct complexity of managing numerous NSPs while preserving service flexibility
Solution Approach 2:
The patent replaces traditional mechanical/combinatorial methods of NSP interconnection with artificial intelligence learning methods. By using recurrent neural networks to predict service chains in vector space, the system substitutes complex rule-based orchestration with machine learning-based prediction, reducing computational burden
2Adaptability or versatility
If a high number of NSPs are interconnected to implement network functions, then service functionality is improved, but computational burden and management complexity increase
Solution Approach 1:
The Service Space Module predicts only the necessary NSPs for a given service request by performing geometric operations on vectors in multidimensional space. Instead of evaluating all possible NSP combinations, the system uses AI learning to predict the optimal subset, thereby reducing computational burden while maintaining full service functionality
Solution Approach 2:
The patent transforms the discrete NSP selection problem into a continuous vector space problem. By representing NSPs as vectors and using geometric operations (addition, subtraction, normalization), the system changes the parameter space from combinatorial to continuous, enabling more efficient computation through gradient-based optimization and neural network inference
Data Source
Figure 1
AI summary
A system comprising: - a provider module (110) configured to serve a Software Defined Infrastructure (SDI) service request generated by a consumer (105), such SDI service request being formulated in a first language; - an NSP translation module (140) configured to translate such SDI requests from said first language into a corresponding service chain comprising a set of Network and Service Primitives (NSP) comprising at least one NSP; - a service space module (150) configured to map each NSP of a service chain onto a multidimensional vector space by exploiting a space matrix, said space matrix being a data structure providing the positions within the multidimensional vector space of vectors NSP(i) corresponding to any possible NSP; and wherein - said service space module (150) is also configured to query said space matrix with a sequence SNSP of vectors NSP(i) so as to retrieve a further sequence SNSP' of vectors NSP'(i) by carrying out on the vectors NSP(i) of the sequence SNSP geometric operations in order to find a level of similarity between such vectors and the vectors of the V-dimensional service space SS so as to carry out at least one among: - predict a new NSP in the composition of the further sequence SNSP' starting from a set of neighbors NSPs composing the sequence SNSP; and - predict a set of neighbors NSPs composing the further sequence SNSP', starting from a NSP of the sequence SNSP.