Machine learning for database migration source

Patent metadata

Field	Value
Publication number	US-9286571-B2
Application number	US-201213879799-A
Country	US
Kind code	B2
Filing date	Apr 1, 2012
Priority date	Apr 1, 2012
Publication date	Mar 15, 2016
Grant date	Mar 15, 2016

Abstract

Technologies are generally provided for maintaining performance level of a database being migrated between different cloud-based service providers employing machine learning. In some examples, data requests submitted to an original data store/database may be submitted to a machine learning-based filter for recording and analysis. Based on the results of the data requests and the filter analyses, new key value structures for a new data store/database may be created. The filter may assign performance scores to the original data requests (made to the original data store) and data requests made to the newly-created key value structures. The filter may then compare the performance scores associated with the created key value structures to each other and to performance scores associated with the original data requests and may select the created key value structures with performance scores that are at least substantially equal to those of the original data requests for the new data store.

First claim

What is claimed is:

1. A method to maintain data store performances upon transfer between cloud computing environments, the method comprising:
submitting data requests to an original data store at a source datacenter;
submitting the data requests to a filter to record and analyze;
creating a new key value structure at a new data store at a target datacenter based on results of the requests to the original data store and analyses by the filter that employs machine learning;
assigning a score to each original data store performance and each new data store performance;
collecting scores of original data store performances and scores of new data store performances for a query and a corresponding key value structure at the new data store;
comparing the scores of the original data store performances and the scores of the new data store performances; and
in response to a determination that the scores of the original data store performances are not substantially equal to the scores of the new data store performances, discarding the corresponding key value structure.

2. The method according to claim 1 , further comprising:
in response to a determination that the scores of the new data store performances are substantially equal to or better than the scores of the original data store performances, recording the scores of the new data store performances for the corresponding key value structure.

3. The method according to claim 2 , further comprising:
selecting a key value structure with a highest score among a plurality of possible key value structures corresponding to an abstracted query at the new data store.

4. The method according to claim 1 , wherein performances include one or more of a processing time, a reading time, and a writing time.

5. The method according to claim 1 , wherein the original data store is a relational database.

6. The method according to claim 1 , further comprising:
receiving a basic table at a beginning of a training process for key value structures at the new data store.

7. A computing device to maintain data store performances upon transfer between cloud computing environments, the computing device comprising:
a memory configured to store instructions; and
a processing unit configured to execute a migration module in conjunction with the instructions, wherein the migration module is configured to:
submit data requests to an original data store at a source datacenter;
submit the data requests to a filter to record and analyze, wherein the filter is further configured to abstract each request to the original data store; and
create a new key value structure at a new data store at a target datacenter based on results of the requests to the original data store and analyses by the filter that employs machine learning;
assign a score to each original data store performance and each new data store performance;
collect scores of original data store performances and scores of new data store performances for a query and a corresponding key value structure at the new data store;
compare the scores of the original data store performances and the scores of the new data store performances; and
in response to a determination that the scores of the original data store performances are not substantially equal to the scores of the new data store performances, discard the corresponding key value structure.

8. The computing device according to claim 7 , wherein the filter is configured to determine the new key value structure based on a probability of types of the requests to the original data store.

9. The computing device according to claim 8 , wherein the filter is further configured to simulate and repeat a series of recorded requests to the new data store to train the new data store.

10. The computing device according to claim 7 , wherein the abstraction includes determination of a type of each request, a table associated with each request, a column associated with each request, and a comparison operation.

11. The computing device according to claim 7 , wherein an abstracted request includes a list of possible key value structures that meet the request to the original data store.

12. The computing device according to claim 11 , wherein the migration module is further configured to:
record the possible key value structures and commands associated with replacing the request to the original data store for subsequent testing.

13. The computing device according to claim 7 , wherein the original data store stores key values.

14. The computing device according to claim 7 , wherein the results of the requests to the original data store are stored in the new data store as user data.

15. The computing device according to claim 7 , wherein the requests to the original data store are one or more of a query, a write, and a transaction.

16. A non-transitory computer-readable storage medium having instructions stored thereon to maintain data store performances upon transfer between cloud computing environments, the instructions comprising:
submitting data requests to an original data store at a source datacenter;
submitting the data requests to a filter to record and analyze;
creating a new key value structure at a new data store at a target datacenter based on results of the requests to the original data store and analyses by the filter that employs machine learning; and
assigning a score to each original data store performance and each new data store performance;
collecting scores of original data store performances and scores of new data store performances for a query and a corresponding key value structure at the new data store;
comparing the scores of the original data store performances and the scores of the new data store performances; and
in response to a determination that the scores of the original data store performances are not substantially equal to the scores of the new data store performances, discarding the corresponding key value structure.

17. The non-transitory computer-readable storage medium according to claim 16 , wherein the instructions further comprise:
receiving a basic table at a beginning of a training process for key value structures at the new data store.

Assignees

• Cao Junwei
• Chen Wei
• Empire Technology Dev Llc

Inventors

• Cao Junwei
• Chen Wei

Classifications

• H04L67/1097

  for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS] · CPC title

• G06N99/005 Primary

  Physics · mapped topic

• G06F17/303

  Physics · mapped topic

• G06F16/119 Primary

  Details of migration of file systems (migration mechanisms in storage systems G06F3/0647) · CPC title

• G06N20/00

  Machine learning · CPC title