Database control using machine learning based prediction

1 . A method, comprising:

accessing a first set of designated database metrics characterizing one or more active sessions of at least one database workload for one or more designated time intervals, wherein the one or more active sessions utilize one or more database resources for performing one or more database operations;

accessing a second set of designated database metrics characterizing one or more waiting sessions of the at least one database workload for the one or more designated time intervals, wherein the one or more waiting sessions execute in response to an occurrence of one or more designated events, wherein the second set of designated database metrics is distinct from the first set of designated database metrics;

evaluating, by at least one processing device, an amount of work performed by the one or more active sessions and the one or more waiting sessions of the at least one database workload within at least one designated time interval by aggregating the first and second sets of designated database metrics within each of the at least one designated time interval;

applying, by the at least one processing device, historical workload data of the at least one database workload to one or more machine learning models to obtain a forecast of an amount of work performed by the at least one database workload for one or more designated time intervals within at least one subsequent designated time period; and

initiating, by the at least one processing device, at least one automated action using the forecast of the amount of work performed by the at least one database workload for the one or more designated time intervals, wherein the at least one automated action comprises rescheduling an execution of one or more scheduled jobs of the at least one database workload based at least in part on the obtained forecast of the amount of work performed by the at least one database workload for the one or more designated time intervals within the at least one subsequent designated time period;

wherein the method is performed by at least one processing device comprising a processor coupled to a memory.

2 . The method of claim 1 , wherein the aggregating the first and second sets of designated database metrics within each of the at least one designated time interval comprises summing a database workload completion time for a plurality of database workload sessions within the at least one designated time interval divided by a duration of the at least one designated time interval.

3 . The method of claim 1 , further comprising characterizing (i) the amount of work performed by the at least one database workload for the one or more designated time intervals within a designated time period relative to a total amount of work performed by the at least one database workload in the designated time period; and (ii) the amount of work performed by the at least one database workload for the one or more designated time intervals within a designated repeating timeframe.

4 . The method of claim 3 , further comprising grouping the characterizations of the amount of work for each of the designated time intervals according to the designated repeating timeframe; determining a tolerance band of the grouped characterizations; and determining at least one dynamic anomalous workload threshold for anomaly detection for each of the designated time intervals using the tolerance band of the corresponding grouped characterizations.

5 . The method of claim 4 , further comprising identifying one or more anomalous workloads by comparing the at least one dynamic anomalous workload threshold to one or more of measured workload data and forecasted workload data.

6 . The method of claim 1 , further comprising applying two or more of: a database resource allocation specification, a database resource utilization specification, a database workload specification for one or more active workloads and a database workload specification for one or more waiting workloads to at least one supervised machine learning regression model to determine a predicted workload breaking point.

7 . The method of claim 6 , wherein the at least one supervised machine learning regression model is trained using historical data associated with the two or more of: the database resource allocation specification, the database resource utilization specification, the database workload specification for the one or more active workloads and the database workload specification for the one or more waiting workloads.

8 . The method of claim 6 , further comprising generating one or more notifications in response to one or more database workloads exceeding the predicted workload breaking point.

9 . The method of claim 1 , wherein the historical workload data of the at least one database workload comprises the first and second sets of designated database metrics.

10 . An apparatus comprising:

at least one processing device comprising a processor coupled to a memory;

the at least one processing device being configured to implement the following steps:

accessing a first set of designated database metrics characterizing one or more active sessions of at least one database workload for one or more designated time intervals, wherein the one or more active sessions utilize one or more database resources for performing one or more database operations;

accessing a second set of designated database metrics characterizing one or more waiting sessions of the at least one database workload for the one or more designated time intervals, wherein the one or more waiting sessions execute in response to an occurrence of one or more designated events, wherein the second set of designated database metrics is distinct from the first set of designated database metrics;

evaluating, by at least one processing device, an amount of work performed by the one or more active sessions and the one or more waiting sessions of the at least one database workload within at least one designated time interval by aggregating the first and second sets of designated database metrics for within each of the at least one designated time interval;

applying, by the at least one processing device, historical workload data of the at least one database workload to one or more machine learning models to obtain a forecast of an amount of work performed by the at least one database workload for one or more designated time intervals within at least one subsequent designated time period; and

initiating, by the at least one processing device, at least one automated action using the forecast of the amount of work performed by the at least one database workload for the one or more designated time intervals, wherein the at least one automated action comprises rescheduling an execution of one or more scheduled jobs of the at least one database workload based at least in part on the obtained forecast of the amount of work performed by the at least one database workload for the one or more designated time intervals within the at least one subsequent designated time period.

11 . The apparatus of claim 10 , wherein the aggregating the first and second sets of designated database metrics within each of the at least one designated time interval comprises summing a database workload completion time for a plurality of database workload sessions within the at least one designated time interval divided by a duration of the at least one designated time interval.

12 . The apparatus of claim 10 , further comprising characterizing (i) the amount of work performed by the at least one database workload for the one or more designated time intervals within a designated time period relative to a total amount of work performed by the at least one database workload in the designated time period; and (ii) the amount of work performed by the at least one database workload for the one or more designated time intervals within a designated repeating timeframe; grouping the characterizations of the amount of work for each of the designated time intervals according to the designated repeating timeframe; determining a tolerance band of the grouped characterizations; and determining at least one dynamic anomalous workload threshold for anomaly detection for each of the designated time intervals using the tolerance band of the corresponding grouped characterizations.

13 . The apparatus of claim 10 , further comprising applying two or more of: a database resource allocation specification, a database resource utilization specification, a database workload specification for one or more active workloads and a database workload specification for one or more waiting workloads to at least one supervised machine learning regression model to determine a predicted workload breaking point and generating one or more notifications in response to one or more database workloads exceeding the predicted workload breaking point.

14 . The apparatus of claim 13 , wherein the at least one supervised machine learning regression model is trained using historical data associated with the two or more of: the database resource allocation specification, the database resource utilization specification, the database workload specification for the one or more active workloads and the database workload specification for the one or more waiting workloads.

15 . The apparatus of claim 10 , wherein the historical workload data of the at least one database workload comprises the first and second sets of designated database metrics.

16 . A non-transitory processor-readable storage medium having stored therein program code of one or more software programs, wherein the program code when executed by at least one processing device causes the at least one processing device to perform the following steps:

accessing a first set of designated database metrics characterizing one or more active sessions of at least one database workload for one or more designated time intervals, wherein the one or more active sessions utilize one or more database resources for performing one or more database operations;

accessing a second set of designated database metrics characterizing one or more waiting sessions of the at least one database workload for the one or more designated time intervals, wherein the one or more waiting sessions execute in response to an occurrence of one or more designated events, wherein the second set of designated database metrics is distinct from the first set of designated database metrics;

evaluating, by at least one processing device, an amount of work performed by the one or more active sessions and the one or more waiting sessions of the at least one database workload within at least one designated time interval by aggregating the first and second sets of designated database metrics for within each of the at least one designated time interval;

applying, by the at least one processing device, historical workload data of the at least one database workload to one or more machine learning models to obtain a forecast of an amount of work performed by the at least one database workload for one or more designated time intervals within at least one subsequent designated time period; and

initiating, by the at least one processing device, at least one automated action using the forecast of the amount of work performed by the at least one database workload for the one or more designated time intervals, wherein the at least one automated action comprises rescheduling an execution of one or more scheduled jobs of the at least one database workload based at least in part on the obtained forecast of the amount of work performed by the at least one database workload for the one or more designated time intervals within the at least one subsequent designated time period.

17 . The non-transitory processor-readable storage medium of claim 16 , further comprising characterizing (i) the amount of work performed by the at least one database workload for the one or more designated time intervals within a designated time period relative to a total amount of work performed by the at least one database workload in the designated time period; and (ii) the amount of work performed by the at least one database workload for the one or more designated time intervals within a designated repeating timeframe; grouping the characterizations of the amount of work for each of the designated time intervals according to the designated repeating timeframe; determining a tolerance band of the grouped characterizations; and determining at least one dynamic anomalous workload threshold for anomaly detection for each of the designated time intervals using the tolerance band of the corresponding grouped characterizations.

18 . The non-transitory processor-readable storage medium of claim 16 , further comprising applying two or more of: a database resource allocation specification, a database resource utilization specification, a database workload specification for one or more active workloads and a database workload specification for one or more waiting workloads to at least one supervised machine learning regression model to determine a predicted workload breaking point and generating one or more notifications in response to one or more database workloads exceeding the predicted workload breaking point.

19 . The non-transitory processor-readable storage medium of claim 18 , wherein the at least one supervised machine learning regression model is trained using historical data associated with the two or more of: the database resource allocation specification, the database resource utilization specification, the database workload specification for the one or more active workloads and the database workload specification for the one or more waiting workloads.

20 . The non-transitory processor-readable storage medium of claim 16 , wherein the historical workload data of the at least one database workload comprises the first and second sets of designated database metrics.