Methods and systems for controlling data backup
Implementations of the subject matter described herein provide methods and systems of controlling data backup, and a computer readable storage medium. The method comprises: obtaining test data for performing data backup with a plurality of backup solutions, the plurality of backup solutions being associated with each set of values amongst a plurality of sets of values of a plurality of factors affecting the data backup; generating a training sample set based on comparison of the test data; converting the training sample set into a decision model for controlling data backup; and optimizing the decision model based on the configuration parameters obtained by performing data backup with the decision model.
1. A method of controlling data backup, the method comprising:
obtaining test data for performing data backup with a plurality of backup solutions, the plurality of backup solutions being associated with each set of preset values amongst a plurality of sets of preset values for a plurality of factors affecting the data backup, wherein obtaining the test data includes generating the test data by performing, based on each set of preset values for the plurality of factors, data backup procedures between a first data storage system and a second data storage system with each one of the plurality of backup solutions, and wherein the test data includes time for performing each of the data backup procedures between the first data storage system and the second data storage system;
generating a training sample set based on comparison of the test data, wherein generating the training sample set includes:
for each set of preset values, comparing time for performing the data backup between the first data storage system and the second data storage system with different backup solutions in the test data,
selecting a backup solution with which the time for performing the data backup between the first data storage system and the second data storage system is lower than a threshold value, and
generating a training sample of the training sample set based on the selected backup solution and the associated set of values for the plurality of factors;
converting the training sample set into a decision model for controlling data backup by a client and storing the decision model into a shared storage unit;
optimizing the decision model based on configuration parameters obtained at least in part by performing, by the client, data backup with the decision model, wherein optimizing the decision model includes performing incremental sample training based only on the configuration parameters, and wherein performing the incremental sample training includes:
generating an incremental training sample based only on the configuration parameters, and
training the incremental training sample based on a machine learning algorithm that enables multi-variable system classification and supports incremental learning of newly-added data; and
wherein the configuration parameters obtained by performing data backup with the decision model are obtained by:
performing, by the client, the data backup online with the decision model stored in the shared storage unit, and
collecting, by a storage management server, the configuration parameters of the client online.
2. The method according to claim 1 , wherein the factors include at least one of the following: a size of data to be backed up, a proportion of new bytes relative to last backup data, an average storage segment size, a minimum storage segment size, a maximum storage segment size, a total number of storage segments, the number of files to be backed up, a bandwidth of a network between a device performing backup and a device to be backed up, and a network round-trip time (RTT) between the device performing the backup and the device to be backed up.
3. The method according to claim 1 , wherein converting the training sample set into a decision model for controlling data backup comprises:
converting, based on a Very Fast Decision Tree (VFDT), the training sample set into a decision tree with the factors as attributes.
4. The method according to claim 3 , wherein an internal node of the decision tree represents one of the factors, a leaf node represents a backup solution for controlling data backup, and a branch from the internal node to the leaf node represents a class based on a range of values for backup environment variables.
5. The method according to claim 1 , wherein the decision model is a decision tree with the factors as attributes; and
wherein training the incremental training sample comprises:
determining splitting of an internal node in the decision tree according to a Hoeffding boundary by calculating an information gain for an attribute of the incremental training sample.
6. The method according to claim 1 , wherein performing data backup procedures between the first data storage system and the second data storage system with each one of the plurality of backup solutions comprises performing real data backup procedures between the first data storage system and the second data storage system with all of the backup solutions.
7. A system for controlling data backup, comprising:
a storage unit configured to store one or more programs and a decision model for controlling selection of a data backup solution; and
a processor coupled to the storage unit and configured to execute the one or more programs to cause the system to perform acts including:
obtaining test data for performing data backup with a plurality of backup solutions, the plurality of backup solutions being associated with each set of preset values amongst a plurality of sets of preset values for a plurality of factors affecting the data backup, wherein obtaining the test data includes generating the test data by performing, based on each set of preset values for the plurality of factors, data backup procedures between a first data storage system and a second data storage system with each one of the plurality of backup solutions, and wherein the test data includes time for performing each of the data backup procedures between the first data storage system and the second data storage system;
generating a training sample set based on comparison of the test data, wherein generating the training sample set includes:
for each set of preset values, comparing time for performing the data backup between the first data storage system and the second data storage system with different backup solutions in the test data,
selecting a backup solution with which the time for performing the data backup between the first data storage system and the second data storage system is lower than a threshold value, and
generating a training sample of the training sample set based on the selected backup solution and the associated set of values for the plurality of factors;
converting the training sample set into a decision model for controlling data backup by a client and storing the decision model into a shared storage unit;
optimizing the decision model based on configuration parameters obtained at least in part by performing, by the client, data backup with the decision model, wherein optimizing the decision model includes performing incremental sample training based only on the configuration parameters, and wherein performing the incremental sample training includes:
generating an incremental training sample based only on the configuration parameters, and
training the incremental training sample based on a machine learning algorithm that enables multi-variable system classification and supports incremental learning of newly-added data; and
wherein the configuration parameters obtained by performing data backup with the decision model are obtained by:
performing, by the client, the data backup online with the decision model stored in the shared storage unit, and
collecting, by a storage management server, the configuration parameters of the client online.
8. The system according to claim 7 , wherein the factors include at least one of the following: a size of data to be backed up, a proportion of new bytes relative to last backup data, an average storage segment size, a minimum segment size, a maximum storage segment size, a total number of storage segments, the number of files to be backed up, a bandwidth of a network between a device performing backup and a device to be backed up, and a network round-trip time (RTT) between the device performing the backup and the device to be backed up.
9. The system according to claim 7 , wherein converting the training sample set into a decision model for controlling data backup comprises:
converting, based on a Very Fast Decision Tree (VFDT), the training sample set into a decision tree with the factors as attributes.
10. The system according to claim 9 , wherein an internal node of the decision tree represents one of the factors, a leaf node represents a backup solution for controlling data backup, and branch from the internal nodes to the leaf node represents a class based on a range of values for backup environment variables.