Self-maintained tablespace
A system, method, and computer program product are configured to: provide a tablespace comprising an original partition having a lower boundary L, an upper boundary U, a split percentage P, and a split ratio R; and in response to the original partition reaching a fullness of P, automatically split the original partition to a first progeny partition and a second progeny partition, wherein the first progeny partition has a lower boundary L 1 that is the same as the lower boundary L of the original partition and an upper boundary U 1 that is determined based on whether the insert is a sequential or random insert, and wherein the second progeny partition has an upper boundaries U 2 that is the same as the upper boundary U of the original partition and a lower boundary L 2 that is determined based on whether the insert is a sequential or random insert.
1. A computer-implemented method, comprising
training, by a processor set, a machine learning model using a machine learning classification algorithm which uses an existing predicted split percentage and an existing predicted split ratio of a tablespace, wherein the machine learning model is continuously trained after a split to determine a predicted split percentage and a predicated split ratio for a next split event;
providing, by the processor set, the tablespace comprising an original partition, the original partition having a lower boundary L, an upper boundary U, a split percentage P, and a split ratio R using the trained machine learning model;
monitoring, by the processor set, a capacity of partitions in the tablespace; and
in response to the original partition reaching a fullness of P based on the monitored capacity of partitions in the tablespace, automatically splitting, by the processor set, the original partition to a first progeny partition and a second progeny partition,
wherein the first progeny partition has a lower boundary L 1 that is the same as the lower boundary L of the original partition and an upper boundary U 1 that is determined based on whether an insert is a sequential insert or a random insert, and
wherein the second progeny partition has an upper boundary U 2 that is the same as the upper boundary U of the original partition and a lower boundary L 2 that is determined based on whether the insert is the sequential insert or the random insert,
the machine learning model further comprises a support vector machine (SVM),
the machine learning model is trained using partition size, history of insert volume, and frequency of insert for determining and outputting the predicted split percentage, and
the machine learning model is trained using history of insert distribution and current key distribution for determining and outputting the predicted split ratio.
2. The computer-implemented method of claim 1 , further comprising collecting, by the processor set, data history of the tablespace, wherein the data history comprises historical input values for queries, previous workload, previous tablespace extension data, and previous split data, wherein the split percentage P and the split ratio R are determined by the trained machine learning model.
3. The computer-implemented method of claim 2 , wherein the trained machine learning model considers the partition size, the history of insert volume, the frequency of insert on the original partition, and history of elapsed time of the split to determine the split percentage P, and the trained machine learning model further comprises an artificial neural network.
4. The computer-implemented method of claim 2 , wherein the trained machine learning model considers the history of insert distribution and the current key distribution of the original partition to determine the split ratio R.
5. The computer-implemented method of claim 2 , wherein the first progeny partition and the second progeny partition retain the split percentage P and the split ratio R.
6. The computer-implemented method of claim 1 , wherein in response to the insert being the random insert, the upper boundary U 1 of the first progeny partition is (U−1)/2−1, and the lower boundary L 2 of the second progeny partition is (U−1)/2.
7. The computer-implemented method of claim 1 , wherein in response to the insert being the sequential insert and the original partition being a middle partition, the upper boundary U 1 of the first progeny partition is R*U, and the lower boundary L 2 of the second progeny partition is R*U+1.
8. The computer-implemented method of claim 1 , wherein in response to the insert being an ascending sequential insert and the original partition being a first partition or last partition, the upper boundary U 1 of the first progeny partition is an index i of a last insert recorded in the original partition, and the lower boundary L 2 of the second progeny partition is i+1.
9. The computer-implemented method of claim 1 , wherein in response to the insert being a descending sequential insert and the original partition being a first partition or last partition, the lower boundary L 2 of the second progeny partition is an index i of a last insert recorded in the original partition, and upper boundary U 1 of the first progeny partition is i−1.
10. A computer program product comprising one or more computer readable storage media having program instructions collectively stored on the one or more computer readable storage media, the program instructions executable to:
train a machine learning model using a machine learning classification algorithm which uses an existing predicted split percentage and an existing predicted split ratio of a tablespace, wherein the machine learning model is trained after a split to determine a predicted split percentage and a predicated split ratio for a next split event;
provide the tablespace comprising an original partition, the original partition having a lower boundary L, an upper boundary U, a split percentage P, and a split ratio R using the trained machine learning model;
monitor a capacity of partitions in the tablespace; and
in response to the original partition reaching a fullness of P based on the monitored capacity of partitions in the tablespace, automatically split the original partition to a first progeny partition and a second progeny partition,
wherein the first progeny partition has a lower boundary L 1 that is the same as the lower boundary L of the original partition and an upper boundary U 1 that is determined based on whether an insert is a sequential insert or a random insert, and
wherein the second progeny partition has an upper boundaries U 2 that is the same as the upper boundary U of the original partition and a lower boundary L 2 that is determined based on whether the insert is the sequential insert or the random insert,
the machine learning model is trained using partition size, history of insert volume, and frequency of insert for determining and outputting the predicted split percentage, and
the machine learning model is trained using history of insert distribution and current key distribution for determining and outputting the predicted split ratio.
11. The computer program product of claim 10 , wherein the program instructions are further executable to collect data history of the tablespace, and the data history comprises historical input values for queries, previous workload, previous tablespace extension data, and previous split data, wherein in response to the insert being the random insert, the upper boundary U 1 of the first progeny partition is (U−1)/2−1, and the lower boundary L 2 of the second progeny partition is (U−1)/2.
12. The computer program product of claim 10 , wherein in response to the insert being the sequential insert and the original partition being a middle partition, the upper boundary U 1 of the first progeny partition is R*U, and the lower boundary L 2 of the second progeny partition is R*U+1, and the trained machine learning model further comprises an artificial neural network.
13. The computer program product of claim 10 , wherein in response to the insert being an ascending sequential insert and the original partition being a first partition or last partition, the upper boundary U 1 of the first progeny partition is an index i of a last insert recorded in the original partition, and the lower boundary L 2 of the second progeny partition is i+1.
14. The computer program product of claim 10 , wherein in response to the insert being a descending sequential insert and the original partition being a first partition or last partition, the lower boundary L 2 of the second progeny partition is an index i of a last insert recorded in the original partition, and upper boundary U 1 of the first progeny partition is i−1.
15. A system comprising a processor set, one or more computer readable storage media, and program instructions collectively stored on the one or more computer readable storage media, the program instructions executable to:
train a machine learning model using a machine learning classification algorithm which uses an existing predicted split percentage and an existing predicted split ratio of a tablespace, wherein the machine learning model is continuously trained after a split to determine a predicted split percentage and a predicated split ratio for a next split event;
provide the tablespace comprising an original partition, the original partition having a lower boundary L, an upper boundary U, a split percentage P, and a split ratio R using the trained machine learning model;
monitor a capacity of partitions in the tablespace; and
in response to the original partition reaching a fullness of P based on the monitored capacity of partitions in the tablespace, automatically split the original partition to a first progeny partition and a second progeny partition, and
the machine learning model is trained using partition size, history of insert volume, and frequency of insert for determining and outputting the predicted split percentage.
16. The system of claim 15 , further comprising collecting, by the processor set, data history of the tablespace, and the data history comprises historical input values for queries, previous workload, previous tablespace extension data, and previous split data,
wherein the first progeny partition has a lower boundary L 1 that is the same as the lower boundary L of the original partition and an upper boundary U 1 that is determined based on whether an insert is a sequential insert or a random insert,
wherein the second progeny partition has an upper boundaries U 2 that is the same as the upper boundary U of the original partition and a lower boundary L 2 that is determined based on whether the insert is the sequential insert or the random insert, and
wherein when the insert is a random insert, the upper boundary U 1 of the first progeny partition is (U−1)/2−1, and the lower boundary L 2 of the second progeny partition is (U−1)/2.
17. The system of claim 15 , wherein the first progeny partition has a lower boundary L 1 that is the same as the lower boundary L of the original partition and an upper boundary U 1 that is determined based on whether an insert is a sequential insert or a random insert,
wherein the second progeny partition has an upper boundaries U 2 that is the same as the upper boundary U of the original partition and a lower boundary L 2 that is determined based on whether the insert is the sequential insert or the random insert, and
wherein in response to the insert being a sequential insert and the original partition being a middle partition, the upper boundary U 1 of the first progeny partition is R*U, and the lower boundary L 2 of the second progeny partition is R*U+1.
18. The system of claim 15 , wherein the first progeny partition has a lower boundary L 1 that is the same as the lower boundary L of the original partition and an upper boundary U 1 that is determined based on whether an insert is a sequential insert or a random insert,
wherein the second progeny partition has an upper boundaries U 2 that is the same as the upper boundary U of the original partition and a lower boundary L 2 that is determined based on whether the insert is the sequential insert or the random insert, and
wherein in response to the insert being an ascending sequential insert and the original partition being a first partition or last partition, the upper boundary U 1 of the first progeny partition is an index i of a last insert recorded in the original partition, and the lower boundary L 2 of the second progeny partition is i+1.
19. The system of claim 15 , wherein the first progeny partition has a lower boundary L 1 that is the same as the lower boundary L of the original partition and an upper boundary U 1 that is determined based on whether an insert is a sequential insert or a random insert,
wherein the second progeny partition has an upper boundaries U 2 that is the same as the upper boundary U of the original partition and a lower boundary L 2 that is determined based on whether the insert is the sequential insert or the random insert, and
wherein in response to the insert being an descending sequential insert and the original partition being a first partition or last partition, the lower boundary L 2 of the second progeny partition is an index i of a last insert recorded in the original partition, and upper boundary U 1 of the first progeny partition is i−1.
20. The system of claim 15 , wherein the split percentage P and the split ratio R are determined by the trained machine learning model, and the trained machine learning model further comprises an artificial neural network.