Automated Legacy System Migration via Input-Output Vector Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for migrating a starting computation system to a target computation system are time-consuming and costly, especially when the source code is not available and the amount of training data sets is limited.
Innovation Solution
A system and method that utilize input-output vector pairs to determine a parametrised subset of building blocks, which are then used to implement a target computation system that replicates the behavior of the starting computation system, without relying on neural networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If neural networks are used to model the starting computation system, then the migration can be automated and accelerated, but a huge amount of training data and calculation time is required
Solution Approach 1:
The patent extracts only the essential input-output vector pairs from the starting computation system's execution traces, rather than requiring comprehensive training data. By selecting and utilizing only the necessary pairs that capture the system's behavioral patterns, the method achieves automated migration without needing massive datasets, thus resolving the contradiction between migration speed and data quantity requirements.
Solution Approach 2:
The patent applies partial action by using a subset of execution traces and input-output vector pairs sufficient for capturing the computation system's behavior, rather than processing all available data. This partial approach enables automated migration at acceptable speed while avoiding the excessive computational resources and data processing requirements that would be needed for complete training.
2Productivity
If neural networks are used to model the starting computation system, then the migration can be automated, but a huge amount of calculation time is required
Solution Approach 1:
The patent extracts only the essential input-output vector pairs from the starting computation system's execution traces, rather than requiring comprehensive training data. By selecting and utilizing only the necessary pairs that capture the system's behavioral patterns, the method achieves automated migration without needing massive datasets, thus resolving the contradiction between migration speed and data quantity requirements.
Solution Approach 2:
The patent applies partial action by using a subset of execution traces and input-output vector pairs sufficient for capturing the computation system's behavior, rather than processing all available data. This partial approach enables automated migration at acceptable speed while avoiding the excessive computational resources and data processing requirements that would be needed for complete training.
3Reliability
If the source code is available, then the migration can be performed by reprogramming, but the process involves many dozens of man-years and is expensive
Solution Approach 1:
The patent implements self-service by enabling the computation system to migrate itself through automated analysis of its execution traces and generation of input-output vector pairs. The system uses its own operational data to create the models needed for migration, eliminating the need for manual reprogramming by experts and significantly reducing migration time while maintaining accuracy.
Solution Approach 2:
The patent replaces the mechanical process of manual code analysis and reprogramming with an automated computational system that processes execution traces and generates migration models. This substitution of manual expert work with automated algorithms eliminates the time-consuming human intervention required in traditional migration approaches.
4Reliability
If the starting computation system is extensive and complex, then the migration ensures accurate replication, but the process becomes extremely time-consuming
Solution Approach 1:
The patent segments the complex computation system into discrete input-output vector pairs extracted from execution traces. By breaking down the system's behavior into these atomic units, the method can systematically process and replicate the system's functionality without being overwhelmed by its overall complexity, thus achieving accurate replication efficiently.
Solution Approach 2:
The patent replaces the mechanical process of manual code analysis and reprogramming with an automated computational system that processes execution traces and generates migration models. This substitution of manual expert work with automated algorithms eliminates the time-consuming human intervention required in traditional migration approaches.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system (10) for migrating a starting computation system (1) to a target computation system (3) is disclosed. After migration (2), the target computation system (3) replicates a behaviour of the starting computation system (1), wherein the behaviour of the starting computation system (1) is represented by a plurality of input output vector pairs (8), each of the input output vector pairs (8) comprising an input vector (6) and an output vector (7) being generated by the starting computation system (1) in reaction to the input vector (6) of the same pair (8). The system (10) comprises an input output vector interface (11), a matching unit (12) and an implementation unit (13). The input output vector interface (11) is configured to receive a plurality of input output vector pairs (8). The matching unit (12) is communicatively coupled to the input output vector interface (11), thus being able to receive a plurality of input output vector pairs (8) via the input output vector interface (11). Additionally, the matching unit (12) is configured to determine a parametrised subset of building blocks (16) by matching the parametrised subset of building blocks (16) with the received input output vectors (8). Various ways of this kind of matching are disclosed. The implementation unit (13) is communicatively coupled with the matching unit (12) to receive the parametrised subset of building blocks (16), wherein the implementation unit (13) is configured to implement the target computation system (3) based on the parametrised subset of building blocks (16).