PARALLEL PROCESSING FOR ETL PROCESSES

• US 20080222634A1
• Filed: 03/06/2007
• Published: 09/11/2008
• Est. Priority Date: 03/06/2007
• Status: Abandoned Application

• Current Assignee: Oath Inc. (Verizon Communications Inc.)
• Original Assignee: Yahoo! Inc. (Apollo Global Management, Inc.)
• Inventors: Rustagi, Amit
• Application Number:
  US11/682,815
• Publication Number:
  US 20080222634A1
• Time in Patent Office: Days
• US Class Current: 718/100
• CPC Class Codes: G06F 2209/5018 Thread allocation
• CPC Class Codes: G06F 2209/5021 Priority
• CPC Class Codes: G06F 9/5038 considering the execution order of a plurality of tasks, e.g. taking priority or time dependency constraints into consideration scheduling strategies G06F9/4881 and subgroups

Abstract

A technique for parallel processing of data from a plurality of data sources in conjunction with an Extract-Transform-Load (ETL) process, the data being part of a related data set, which comprises the following: staging a unit of extracted data from each of the plurality of data sources, thereby generating a plurality of units of staged data; identifying a plurality of tasks relating to transforming the staged data; assigning a subset of the tasks to each of a plurality of child processes being managed by a master process, such that dependent tasks are assigned to a same child process; concurrently executing the subsets of tasks assigned to the child processes, thereby generating a plurality of units of transformed data from the plurality of units of staged data; and publishing the transformed data after all tasks are completely executed, thereby ensuring that the published data represent the related data set.

Claims

1. A computer-implemented method for parallel processing of data from a plurality of data sources in conjunction with an Extract-Transform-Load (ETL) process, the data being part of a related data set, comprising:

staging a unit of extracted data from each of the plurality of data sources, thereby generating a plurality of units of staged data;

identifying a plurality of tasks for transforming the staged data;

assigning a subset of the tasks to each of a plurality of child processes being managed by a master process, such that dependent tasks are assigned to a same child process;

concurrently executing the subsets of tasks assigned to the child processes, thereby generating a plurality of units of transformed data from the plurality of units of staged data; and

publishing the transformed data to at least one data store after all of the plurality of tasks are completed, thereby ensuring that the published data represent the related data set.

2. The method, as recited in claim 1, further comprising:

assigning the tasks to each subset of tasks such that each child process executes approximately a same number of tasks.

3. The method, as recited in claim 1, further comprising for each of the child processes:

monitoring execution of the tasks assigned to the child process; and

if the child process fails to execute all of the tasks assigned to the child process, assigning unexecuted ones of the tasks assigned to the child process to another one of the child processes.

4. The method, as recited in claim 1, further comprising for each of the plurality of child processes:

monitoring execution of the tasks assigned to the child process; and

if the child process fails to execute all of the tasks assigned to the child process, invoking a new child process to replace the child process and assigning unexecuted ones of the tasks assigned to the child process to the new child process.

5. The method, as recited in claim 1, further comprising:

if execution of the master process terminates before completion, terminating execution of all of the child processes.

6. The method, as recited in claim 1, wherein identifying the plurality of tasks begins after all the units of staged data are available.

7. The method, as recited in claim 1, wherein the related data set is defined with reference to at least one metric selected from the group consisting of a period of time, a geographic region, a business entity, a business analysis, a product, and a group of customers.

8. A system for parallel processing of data in conjunction with an ETL process, comprising:

a plurality of data sources operable to generate the data, the data being part of a related data set;

at least one data store operable to store published data generated by the ETL process; and

at least one computing device configured to;

stage a unit of extracted data from each of the plurality of data sources, thereby generating a plurality of units of staged data;

identify a plurality of tasks for transforming the staged data;

assign a subset of the tasks to each of a plurality of child processes being managed by a master process, such that dependent tasks are assigned to a same child process;

concurrently execute the subsets of tasks assigned to the child processes, thereby generating a plurality of units of transformed data from the plurality of units of staged data; and

publish the transformed data to the at least one data store after all of the plurality of tasks are completed, thereby ensuring that the published data represent the related data set.

9. The system, as recited in claim 8, wherein the at least one computing device is configured to identify the plurality of tasks after all the units of staged data are available.

10. The system, as recited in claim 8, wherein the related data set is defined with reference to at least one metric selected from the group consisting of a period of time, a geographic region, a business entity, a business analysis, a product, and a group of customers.

11. The system, as recited in claim 8, wherein the at least one computing device is configured to, for each of the child processes:

monitor execution of the tasks assigned to the child process; and

if the child process fails to execute all of the tasks assigned to the child process, assign unexecuted ones of the tasks assigned to the child process to another one of the child processes.

12. The system, as recited in claim 8, wherein the at least one computing device is configured to, for each of the child processes:

monitor execution of the tasks assigned to the child process; and

if the child process fails to execute all of the tasks assigned to the child process, invoke a new child process to replace the child process and assign unexecuted ones of the tasks assigned to the child process to the new child process.

13. A computer program product for parallel processing of data from a plurality of data sources in conjunction with an Extract-Transform-Load (ETL) process, the data being part of a related data set, the computer program product comprising a computer-readable medium having a plurality of computer program instructions stored therein, which are operable to cause at least one computer device to:

stage a unit of extracted data from each of the plurality of data sources, thereby generating a plurality of units of staged data;

identify a plurality of tasks for transforming the staged data;

assign a subset of the tasks to each of a plurality of child processes being managed by a master process, such that dependent tasks are assigned to a same child process;

concurrently execute the subsets of tasks assigned to the child processes, thereby generating a plurality of units of transformed data from the plurality of units of staged data; and

publish the transformed data to at least one data store after all of the plurality of tasks are completed, thereby ensuring that the published data represent the related data set.

14. The computer program product, as recited in claim 13, wherein the computer program instructions are further operable to cause the at least one computer device to:

assign the tasks to each subset of tasks such that each child process executes approximately a same number of tasks.

15. The computer program product, as recited in claim 13, wherein the computer program instructions are further operable to cause the at least one computer device to, for each of the child processes:

monitor execution of the tasks assigned to the child process; and

if the child process fails to execute all of the tasks assigned to the child process, assign unexecuted ones of the tasks assigned to the child process to another one of the child processes.

16. The computer program product, as recited in claim 13, wherein the computer program instructions are further operable to cause the at least one computer device to, for each of the plurality of child processes:

monitor execution of the tasks assigned to the child process; and

if the child process fails to execute all of the tasks assigned to the child process, invoke a new child process to replace the child process and assign unexecuted ones of the tasks assigned to the child process to the new child process.

17. The computer program product, as recited in claim 13, wherein the computer program instructions are further operable to cause the at least one computer device to:

if execution of the master process terminates before completion, terminate execution of all of the child processes.

18. The computer program product, as recited in claim 13, wherein the computer program instructions are operable to cause the at least one computer device to identify the plurality of tasks by identifying at least one task for transforming each unit of staged data when the unit of staged data becomes available.

19. The computer program product, as recited in claim 13, wherein the computer program instructions are operable to cause the at least one computer device to identify the plurality of tasks after all the units of staged data are available.

20. The computer program product, as recited in claim 13, wherein the related data set is defined with reference to at least one metric selected from the group consisting of a period of time, a geographic region, a business entity, a business analysis, a product, and a group of customers.