System and method for limiting the impact of stragglers in large-scale parallel data processing
원문보기
IPC분류정보
국가/구분
United States(US) Patent
등록
국제특허분류(IPC7판)
G06F-009/44
G06F-009/50
출원번호
US-0213188
(2016-07-18)
등록번호
US-9886325
(2018-02-06)
발명자
/ 주소
Malewicz, Grzegorz
Dvorsky, Marian
Colohan, Christopher B.
Thomson, Derek P.
Levenberg, Joshua Louis
출원인 / 주소
GOOGLE LLC
대리인 / 주소
Morgan, Lewis & Bockius LLP
인용정보
피인용 횟수 :
1인용 특허 :
33
초록▼
A large-scale data processing system and method including a plurality of processes, wherein a master process assigns input data blocks to respective map processes and partitions of intermediate data are assigned to respective reduce processes. In each of the plurality of map processes an application
A large-scale data processing system and method including a plurality of processes, wherein a master process assigns input data blocks to respective map processes and partitions of intermediate data are assigned to respective reduce processes. In each of the plurality of map processes an application-independent map program retrieves a sequence of input data blocks assigned thereto by the master process and applies an application-specific map function to each input data block in the sequence to produce the intermediate data and stores the intermediate data in high speed memory of the interconnected processors. Each of the plurality of reduce processes receives a respective partition of the intermediate data from the high speed memory of the interconnected processors while the map processes continue to process input data blocks an application-specific reduce function is applied to the respective partition of the intermediate data to produce output values.
대표청구항▼
1. A method of performing a large-scale data processing job, comprising: executing a plurality of processes on a plurality of interconnected processors, the plurality of processes including a plurality of map processes and a plurality of reduce processes;in each respective map process of the plurali
1. A method of performing a large-scale data processing job, comprising: executing a plurality of processes on a plurality of interconnected processors, the plurality of processes including a plurality of map processes and a plurality of reduce processes;in each respective map process of the plurality of map processes: executing a map program to retrieve a sequence of input data blocks assigned to the respective map process and to apply a map function to each input data block in the sequence to produce the intermediate data; andstoring the intermediate data in memory; andin each respective reduce process of the plurality of reduce processes: receiving a respective partition of the intermediate data assigned to the respective reduce process; andapplying a reduce function to the respective partition of the intermediate data to produce output values; andin a first respective reduce process: receiving multiple distinct partitions of the intermediate data; andprocessing the multiple partitions one at a time in succession; andidentifying the first respective reduce process as a reduce process that is delaying the data processing job while continuing to process intermediate data and, in response, dividing the intermediate data in a partition that is assigned to the first respective reduce process into a plurality of subpartitions and assigning at least one subpartition of the plurality of subpartitions to a reduce process that is not the first respective reduce process. 2. The method of claim 1, further comprising partitioning the intermediate data into a plurality of partitions of the intermediate data, the plurality of partitions including the multiple distinct partitions of the intermediate data received by the first respective reduce process. 3. The method of claim 2, wherein the data processing job is initiated by a user, and the intermediate data is sorted into the plurality of partitions based on a partition function selected by the user. 4. The method of claim 3, wherein the partition function is defined by the user. 5. The method of claim 1, wherein the data processing job is initiated by a user, and the map function and the reduce function are selected by the user. 6. The method of claim 5, wherein the map function and the reduce function are defined by the user. 7. The method of claim 1, wherein: producing the intermediate data includes producing a plurality of blocks of intermediate data, wherein each block of intermediate data includes all of the intermediate data produced by applying the map function to a respective input data block; andreceiving a respective partition of the intermediate data includes receiving a subset of the intermediate data in a first block of intermediate data that is associated with the respective partition while a second block of intermediate data is being produced, the second block of intermediate data including at least some intermediate data that is associated with the respective partition. 8. The method of claim 1, further comprising identifying a partition of the intermediate data that is likely to delay the data processing job using predefined criteria and performing a remedial action with respect to the identified partition of the intermediate data. 9. The method of claim 8, wherein identifying a partition of the intermediate data that is likely to delay the data processing job includes determining the size of the partition of the intermediate data relative to the size of other partitions of the intermediate data in the data processing job. 10. The method of claim 8, wherein remedial action comprises scheduling the partition of the intermediate data for processing on a high capacity reduce process. 11. The method of claim 1, wherein applying a reduce function to the respective partition of the intermediate data to produce output values includes: while continuing to receive a respective partition of the intermediate data: storing at least a subset of the intermediate data of the respective partition in memory associated with the respective reduce process;while the intermediate data is stored in the memory associated with the respective reduce process, applying a combiner function to produce combined intermediate data values; andapplying the reduce function to the combined intermediate data values to produce output values. 12. The method of claim 11, wherein the combiner function is the same function as the reduce function. 13. The method of claim 1, wherein receiving the respective partition of the intermediate data occurs while the map processes that produced the received intermediate data continue to process input data blocks. 14. A system for large-scale processing of data, comprising: memory;one or more processors; andone or more modules stored in the memory and executed by the one or more processors, the one or more modules including instructions to:execute a plurality of processes on a plurality of interconnected processors, the plurality of processes including a plurality of map processes and a plurality of reduce processes;in each respective map process of the plurality of map processes: execute a map program to retrieve a sequence of input data blocks assigned to the respective map process and to apply a map function to each input data block in the sequence to produce intermediate data; andstore the intermediate data in memory; andin each respective reduce process of the plurality of reduce processes: receive a respective partition of the intermediate data assigned to the respective reduce process; andapply a reduce function to the respective partition of the intermediate data to produce output values; andin a first respective reduce process: receive multiple distinct partitions of the intermediate data; andprocess the multiple partitions one at a time in succession; andidentify the first respective reduce process as a reduce process that is delaying the data processing job while continuing to process intermediate data and, in response, divide the intermediate data in a partition that is assigned to the first respective reduce process into a plurality of subpartitions and assign at least one subpartition of the plurality of subpartitions to a reduce process that is not the first respective reduce process. 15. The system of claim 14, wherein the one or more modules further include instructions to identify a partition of the intermediate data that is likely to delay the data processing job using predefined criteria and to perform a remedial action with respect to the identified partition of the intermediate data. 16. The system of claim 14, wherein: the instructions to produce the intermediate data include instructions to produce a plurality of blocks of intermediate data, wherein each block of intermediate data includes all of the intermediate data produced by applying the map function to a respective input data block; andthe instructions to receive a respective partition of the intermediate data include instructions to receive a subset of the intermediate data in a first block of intermediate data that is associated with the respective partition while a second block of intermediate data is being produced, the second block of intermediate data including at least some intermediate data that is associated with the respective partition. 17. The system of claim 14, wherein the instructions to apply a reduce function to the respective partition of the intermediate data to produce output values include instructions to: while continuing to receive a respective partition of the intermediate data: store at least a subset of the intermediate data of the respective partition in memory associated with the respective reduce process;while the intermediate data is stored in the memory associated with the respective reduce process, apply a combiner function to produce combined intermediate data values; andapply the reduce function to the combined intermediate data values to produce output values. 18. The system of claim 14, wherein receiving the respective partition of the intermediate data occurs while the map processes that produced the received intermediate data continue to process input data blocks. 19. A non-transitory computer readable storage medium storing one or more programs for execution by one or more processors of a computer system, the one or more programs comprising instructions to: execute a plurality of processes on a plurality of interconnected processors, the plurality of processes including a plurality of map processes and a plurality of reduce processes;in each respective map process of the plurality of map processes: execute a map program to retrieve a sequence of input data blocks assigned to the respective map process and to apply a map function to each input data block in the sequence to produce intermediate data; andstore the intermediate data in memory; andin each respective reduce process of the plurality of reduce processes: receive a respective partition of the intermediate data assigned to the respective reduce process; andapply a reduce function to the respective partition of the intermediate data to produce output values; andin a first respective reduce process: receive multiple distinct partitions of the intermediate data; andprocess the multiple partitions one at a time in succession; andidentify the first respective reduce process as a reduce process that is delaying the data processing job while continuing to process intermediate data and, in response, divide the intermediate data in a partition that is assigned to the first respective reduce process into a plurality of subpartitions and assign at least one subpartition of the plurality of subpartitions to a reduce process that is not the first respective reduce process. 20. The non-transitory computer readable storage medium of claim 19, wherein the one or more programs further include instructions to identify a partition of the intermediate data that is likely to delay the data processing job using predefined criteria and to perform a remedial action with respect to the identified partition of the intermediate data.
연구과제 타임라인
LOADING...
LOADING...
LOADING...
LOADING...
LOADING...
이 특허에 인용된 특허 (33)
Sprenger Jeff H. ; Gramley George W. ; Major Debbie A. ; Thompson Richard A. ; Hatcherson Rob, Apparatus and system for an adaptive data management architecture.
McMillen Robert J. ; Watson M. Cameron ; Chura David J., Computer system using a master processor to automatically reconfigure faulty switch node that is detected and reported.
Van Huben Gary Alan ; Mueller Joseph Lawrence ; Siegel Michael Steven ; Warnock Thomas Bernard ; McDonald Darryl James, Data management system and process.
Sprenger Jeff H. ; Gramley George W. ; Major Debbie A. ; Thompson Richard A. ; Hatcherson Rob, Method and apparatus for data management using an event transition network.
Ricard Gary Ross ; Rocheleau Richard Miles ; Sadecki Wayne Christopher, Method and computer program product for implementing highly concurrent record insertion in an ordinal number dependent database.
Ekanadham Kattamuri ; Moreira Jose Eduardo ; Naik Vijay Krishnarao, Method for resource control in parallel environments using program organization and run-time support.
Bookman,Lawrence A.; Blair,David Albert; Rosenthal,Steven M.; Krawitz,Robert Louis; Beckerle,Michael J.; Callen,Jerry Lee; Razdow,Allen M.; Mudambi,Shyam R., Segmentation and processing of continuous data streams using transactional semantics.
Malewicz, Grzegorz; Dvorsky, Marian; Colohan, Christopher B.; Thomson, Derek P.; Levenberg, Joshua Louis, System and method for limiting the impact of stragglers in large-scale parallel data processing.
Kazemi, Moslem; Panikulam, Jacob; Liu, Chenggang; Lee, Andy; Bradley, David McAllister; Hogg, III, Charles R., Automatic tuning of autonomous vehicle cost functions based on human driving data.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.