Storage Inter-Volume Copy Time Prediction via Pool Segmentation

The patent segments the storage system into multiple resource pools, each with its own copy speed prediction model. Instead of using a single fixed copy speed for all volumes, the system divides the storage resources into distinct pools (e.g., pool 0, pool 1, etc.) and predicts copy speeds separately for each pool based on their specific resource usage patterns. This segmentation allows accurate prediction even when multiple volumes execute IO and copy processing simultaneously, as each pool's dynamics are independently modeled.

The patent performs preliminary analysis of resource pool dynamics before executing the copy operation. The system collects historical data on resource usage (CPU, memory, I/O operations) for each resource pool and pre-calculates copy speed prediction models. When a copy operation is initiated, the system quickly applies the pre-established model to predict copy time, rather than calculating resource dynamics in real-time during the copy process. This preliminary action enables fast prediction while maintaining accuracy by accounting for resource overload conditions.

The patent implements dynamic resource allocation based on predicted copy speeds for each resource pool. Instead of using fixed copy speeds, the system continuously monitors resource pool states (resource usage rates, queue depths, etc.) and adjusts copy speeds dynamically according to the predicted values. For example, when resource pool 0 shows high CPU usage, the system automatically reduces the copy speed for volumes in that pool to prevent resource overload. This dynamic approach maintains system stability during simultaneous IO and copy processing while keeping resource allocation manageable through pool-based grouping.