阅读说明：本技术 计算处理的动态并行化 (Dynamic parallelization of computational processing ) 是由 S·科拉查拉 许健午 T·宏 L·罗登贝瑞 黄登胜 M·科斯拉维 P·霍兰德 B·帕特 于 2018-04-27 设计创作，主要内容包括：一种系统,该系统用于通过将基于规则的表达式列表划分到多个任务单元中并将处于相同计算级别的所有独立任务单元重新配置到若干并行化任务组中,生成评估规则集合或基于规则的表达式列表的并行计算规划,使得每个任务组内的任务单元可以被调度用于跨处理节点集群的并行执行。并行化可以基于所生成的任务被动态地确定,但是可以基于将每个任务基于范围地划分到多个并行可执行子任务中而进一步进行附加层的并行化。最终的并行化计算规划可以包括基于利用关于每个任务组的并行化执行的信息的问题分区和逻辑依赖性的顺序排序的任务组集合。(A system for generating a parallel computation plan evaluating a set of rules or a rule-based expression list by dividing the rule-based expression list into a plurality of task units and reconfiguring all independent task units at the same computation level into a number of parallelized task groups such that the task units within each task group can be scheduled for parallel execution across a cluster of processing nodes. The parallelization may be determined dynamically based on the generated tasks, but may be further additional layers of parallelization based on scoping each task into multiple parallel executable sub-tasks. The final parallelized computation plan may include a set of task groups based on a sequential ordering of problem partitions and logical dependencies with information about parallelized execution of each task group.)

1. A non-transitory computer-readable medium having instructions stored thereon, which when executed by a processor, cause the processor to generate a parallel computation plan evaluating a set of rule-based expressions, the generating comprising:

dividing the rule-based expression set into a plurality of task units; and

arranging the plurality of task units into a sequential set of parallelizable task groups, wherein sequential ordering is determined by a logical dependency problem partition associated with the set of rule-based expressions; and

generating the parallel computing plan including a sequential execution ordering of parallelizable task groups and a parallel execution ordering of task units within each parallelizable task group.

2. The non-transitory computer-readable medium of claim 1, wherein the plurality of task units includes one or more initialization, aggregation, or computation operations required to evaluate each rule-based expression in the set of rule-based expressions.

3. The non-transitory computer readable medium of claim 1, further comprising partitioning one or more parallelizable task groups of the sequential set from the parallelizable task groups into a parallel set of parallelizable subtask groups.

4. The non-transitory computer-readable medium of claim 3, wherein the parallel set of parallelizable subtask groups is generated by a scope-based parallelization of one or more parallelizable task groups from the sequential set of parallelizable task groups.

5. The non-transitory computer readable medium of claim 1, wherein the parallel computing plan is generated by a compute engine and dispatched to an intermediate tier for scheduling and execution.

6. The non-transitory computer-readable medium of claim 1, wherein the parallel computing plan is scheduled for parallel execution across a cluster of computers.

7. The non-transitory computer readable medium of claim 1, wherein the parallel computing plan is generated by a compute engine of a Retail Predictive Application Server (RPAS).

8. A computer-implemented method for generating a parallel computation plan evaluating a set of rule-based expressions, the method comprising:

dividing the rule-based expression set into a plurality of task units;

arranging the plurality of task units into a sequential set of parallelizable task groups, wherein sequential ordering is determined by a problem partition and logical dependencies associated with the set of rule-based expressions; and

generating the parallel computing plan including a sequential execution order of parallelizable task groups and a parallel execution order of task units within each parallelizable task group.

9. The computer-implemented method of claim 8, wherein the plurality of task units includes one or more initialization, aggregation, or computation operations required to evaluate each rule-based expression in the set of rule-based expressions.

10. The computer-implemented method of claim 8, further comprising partitioning one or more parallelizable task groups of the sequential set from the parallelizable task groups into a parallel set of parallelizable subtask groups.

11. The computer-implemented method of claim 10, wherein the parallel set of parallelizable subtask groups is generated by range-based parallelization of one or more parallelizable task groups from the sequential set of parallelizable task groups.

12. The computer-implemented method of claim 8, wherein the parallel computing plan is generated by a compute engine and dispatched to an intermediate layer for scheduling and execution.

13. The computer-implemented method of claim 12, wherein the computational engine is part of a retail forecasting application server (RPAS).

14. The computer-implemented method of claim 8, wherein the parallel computing plan is scheduled for parallel execution across a cluster of computers.

15. A system for generating a parallel computation plan evaluating a set of rule-based expressions, comprising:

a first unit to divide one or more rule-based expression sets into a plurality of task units;

a second unit to arrange the plurality of task units into a sequential set of parallelizable task groups, wherein sequential ordering is determined by a problem partition and logical dependencies associated with the set of rule-based expressions; and

a third unit for generating a parallel computation plan comprising a sequential execution ordering of the parallelizable task groups and a parallel execution ordering of the task units within each parallelizable task group.

16. The system of claim 15, wherein the plurality of task units includes one or more initialization, aggregation, or computation operations required to evaluate each rule-based expression in the set of rule-based expressions.

17. The system of claim 15, wherein the second unit is further to divide one or more parallelizable task groups of the sequential set from the parallelizable task groups into a parallel set of parallelizable subtask groups.

18. The system of claim 17, wherein the parallel set of parallelizable subtask groups is generated by range-based parallelization of one or more parallelizable task groups from the sequential set of parallelizable task groups.

19. The system of claim 15, wherein the first unit and the second unit constitute a compute engine, and wherein the parallel compute plan generated by the compute engine is dispatched to a middle tier level for scheduling and execution.

20. The system of claim 19, wherein the compute engine is part of a Retail Predictive Application Server (RPAS), and wherein the parallel compute plan is scheduled for parallel execution across a cluster of computers.