In certain embodiments, a method and system and computer program product are provided for utilizing a neural network and machine learning mechanisms for improving the performance of a host interface (e.g., a host bus adapter) of a storage subsystem by dynamically allocating TCBs to each port of the host interface. In certain embodiments, an optimal number of resources (TCBs) are calculated to allocate to each port of the host bus adapter based on a plurality of input factors. Adjustments is made to the number of resources to allocate to each port based on I/O queues and the availability or non-availability of resources. The embodiments dynamically allocate TCBs to each port by utilizing a machine learning module.

FIG. 1 illustrates a block diagram of a computing environment 100 for training and using a machine learning module for load balancing of shared resources in a storage controller, in accordance with certain embodiments. The computing environment 100 comprises a storage controller 102 in which a load balancing application 104 and a machine learning module 106 are implemented, in accordance with certain embodiments.