Bottleneck Detection via Correlation Model Correction

The patent creates a prediction model that copies the correlation relationships from the migration-source environment to predict bottlenecks in the migration-destination environment. Instead of directly operating the migration-destination system, the patent uses benchmark performance data from both environments to generate a predicted correlation model that replicates the behavior patterns of the original system, enabling bottleneck identification without actual operation of the target system.

The patent performs preliminary actions by collecting benchmark performance data from both migration-source and migration-destination environments before migration occurs. The prediction model is built in advance using this pre-collected data, allowing bottleneck prediction to be performed without requiring actual operation of the migration-destination system during the migration process itself.

The patent employs a universal benchmarking approach where the same benchmark program is executed on both migration-source and migration-destination systems to collect performance data. This multi-functional benchmarking process serves multiple purposes: characterizing the source system, characterizing the destination system, and providing the data foundation for correlation model correction, thereby reducing overall data collection complexity.