In the embodiments shown in FIG. 11, the margin of error is computed by calculating the deviation of the actual output from the expected output to adjust weights and biases in the machine learning module 106 via back propagation (reference numeral 1118).

FIG. 12 illustrates a block diagram 1200 that shows an example for adjustment of weights via back propagation by computing a margin of error during training of the machine learning module 106 based on global queuing 1008, in accordance with certain embodiments.

In FIG. 12, in one example, the following are the values of certain parameters:

(i) The actual output of the machine learning module=N (reference numeral 1202);

(ii) The number of I/O operations queued in a global queue=Y (reference numeral 1204);

(iii) Number of free TCBs in a global TCB pool=X. (reference numeral 1106).

For this example, the expected output 1208 of the machine learning module 106 is calculated as N?Y+X That is output should have been less for a port when more requests are queued globally and should have been more when less requests queued globally.

In the embodiments shown in FIG. 12, the margin of error is computed by calculating the deviation of the actual output 1202 from the expected output 1208 to adjust weights and biases in the machine learning module 106 via back propagation (reference numeral 1210).