Method and system to optimize cloud cost by analyzing pricing models

A system, comprising:

• a memory storing instructions;
• one or more communication interfaces;
• and
• one or more hardware processors coupled to the memory via the one or more communication interfaces, wherein the one or more hardware processors are configured by the instructions to:
• obtain an optimal band from utilization of one or more metrics specific to one or more set of resources using a sliding window approach by:
• (i) sorting a set of datapoints U corresponding to a timeseries U(t) over T timesteps, in an ascending order;
• (ii) selecting one or more datapoints (n) from the sorted set of datapoints that covers a pre-defined percentage of the set of datapoints U, as a window size;
• and
• (iii) applying the sliding window approach on the sorted set of datapoints U for the selected window size for obtaining the optimal band by:
• (i) creating an initial band with the selected the one or more datapoints (n), wherein a first datapoint in the selected the one or more datapoints (n) is considered as a lower bound of the initial band ULower and a nth datapoint in the selected one or more datapoints (n) is considered as a upper bound of the initial band UUpper;
• (ii) calculating a first width of the created initial band as a minimum width Wmin by finding a difference between the upper bound of the initial band Uupper and the lower bound of the initial band ULower as UUpper−ULower;
• and
• (iii) iteratively performing until a penultimate datapoint from the set of datapoints U is indexed by:
• a) obtaining a new lower bound of the initial band UNewLower and a corresponding new upper bound of the initial band UNewUpper by adding (n) to a starting point index, for each new starting point comprised in the lower bound of the initial band ULower;
• b) computing a second width W by finding a difference between the associated new upper bound of the initial band UNewUpper and the new lower bound of the initial band UNewLower as UNewUpper−UNewLower;
• c) comparing the second width W and the minimum width Wmin;
• d) updating the minimum width Wmin based on the comparison and saving the second width W as the optimal band;
• and
• e) selecting an interval [UUpper, ULower] that has minimum width Wmin, UUpper−ULower;
• calculate a total outlier factor to determine the set of datapoints U outside the optimal band and assessing a proximity of the set of datapoints U to the optimal band, using a Root mean square deviation (RMSD) technique by:
• (i) calculating a lower band penalty for each datapoint comprised in the set of datapoints U below the lower bound of the optimal band by determining a Root mean square deviation of the set of datapoints U with respect to the lower bound of the initial band;
• (ii) calculating an upper band penalty for each datapoint set of datapoints U above the upper bound of the optimal by determining the Root mean square deviation of the set of datapoints U with respect to the upper bound of the band;
• and
• (iii) calculating the total outlier factor based on the lower band penalty and the upper band penalty;
• and
• generate one or more recommendations for optimizing a cloud cost based on the obtained optimal band and the total outlier factor comprising:
• (i) providing the one or more set of resources for a long-term reserve instance (RI) pricing model, if the width of the optimal band is less than a first pre-defined threshold;
• (ii) providing the one or more set of resources for the long-term reserve instance (RI) pricing model, if the total outlier factor is less than a second pre-defined threshold;
• and
• (iii) providing the one or more set of resources for a short-term reserve instance (RI) pricing model, if the total outlier factor is greater than a third pre-defined threshold.

One or more non-transitory machine-readable information storage mediums comprising one or more instructions which when executed by one or more hardware processors cause:

• obtaining an optimal band from utilization of one or more metrics specific to one or more set of resources using a sliding window approach by:
• (i) sorting a set of datapoints U corresponding to a timeseries U(t) over T timesteps, in an ascending order;
• (ii) selecting one or more datapoints (n) from the sorted set of datapoints that covers a pre-defined percentage of the set of datapoints U, as a window size;
• and
• (iii) applying the sliding window approach on the sorted set of datapoints U for the selected window size for obtaining the optimal band by:
• (i) creating an initial band with the selected the one or more datapoints (n), wherein a first datapoint in the selected the one or more datapoints (n) is considered as a lower bound of the initial band ULower and a nth datapoint in the selected one or more datapoints (n) is considered as a upper bound of the initial band UUpper;
• (ii) calculating a first width of the created initial band as a minimum width Wmin by finding a difference between the upper bound of the initial band Uupper and the lower bound of the initial band ULower as UUpper−ULower;
• and
• (iii) iteratively performing until a penultimate datapoint from the set of datapoints U is indexed by:
• a) obtaining a new lower bound of the initial band UNewLower and a corresponding new upper bound of the initial band UNewUpper by adding (n) to a starting point index, for each new starting point comprised in the lower bound of the initial band ULower;
• b) computing a second width W by finding a difference between the associated new upper bound of the initial band UNewUpper and the new lower bound of the initial band UNewLower as UNewUpper−UNewLower;
• c) comparing the second width W and the minimum width Wmin;
• d) updating the minimum width Wmin based on the comparison and saving the second width W as the optimal band;
• and
• e) selecting an interval [UUpper, ULower] that has minimum width Wmin, UUpper−ULower;
• calculating a total outlier factor to determine the set of datapoints U outside the optimal band and assessing a proximity of the set of datapoints U to the optimal band, using a Root mean square deviation (RMSD) technique by:
• (i) calculating a lower band penalty for each datapoint comprised in the set of datapoints U below the lower bound of the optimal band by determining a Root mean square deviation of the set of datapoints U with respect to the lower bound of the initial band;
• (ii) calculating an upper band penalty for each datapoint set of datapoints U above the upper bound of the optimal by determining the Root mean square deviation of the set of datapoints U with respect to the upper bound of the band;
• and
• (iii) calculating the total outlier factor based on the lower band penalty and the upper band penalty;
• and
• generating one or more recommendations for optimizing a cloud cost based on the obtained optimal band and the total outlier factor comprising:
• (i) providing the one or more set of resources for a long-term reserve instance (RI) pricing model, if the width of the optimal band is less than a first pre-defined threshold;
• (ii) providing the one or more set of resources for the long-term reserve instance (RI) pricing model, if the total outlier factor is less than a second pre-defined threshold;
• and
• (iii) providing the one or more set of resources for a short-term reserve instance (RI) pricing model, if the total outlier factor is greater than a third pre-defined threshold.