{"title":"服装厂并行仿真中同步问题的解决方法","authors":"P. Khambekar, S. Dharmaraj","doi":"10.1145/98949.99143","DOIUrl":null,"url":null,"abstract":"A near-term computer simulation of an apparel in dustry shop-floor is an application of interest. The complexity and the natural parallelism of the applica tion suggest a parallel implementation. Parallelizing a simulation causes synchronization problems. These problems are analyzed, various solutions presented and short-comings or inapplicability to the application de scribed. A method which is an adaptation of a previ ously published method is then presented. This con servative method produces an accurate simulation. I n tr o d u c t io n Clemson University is currently developing a nearterm computer simulation of an apparel factory floor. Given inputs of employee assignments, arrival of new bundles of garment parts, priorities of bundles, etc. (i.e. a tentative plan), the flow of bundles being sewn during the course of a day is simulated and high-level performance information in the form of graphics is dis played to the plant manager (the user). The manager should be able to run the simulation, evaluate the per formance metrics and in case they are not satisfactory, roll-back the simulation to a specified point in time and rerun the simulation with a new plan all in a matter of minutes. This simulation will provide a valuable tool for achieving Just-In-Time manufacturing. In a typical plant, there are hundreds of ma chines, hundreds of employees and thousands of bun dles containing garment parts. The complexity of the application, large input data, large number of metrics, quick simulation requirement and natural parallelism of the application strongly suggest performing the sim ulation in parallel. Unfortunately, in a general-purpose parallel simulation two synchronization problems may occur: deadlock and no-progress. This paper describes the methods planned to address these problems. Peniiinnlon lo copy wllliout fee all or part of this material la granted provided that the copiea are not made or distributed for direct com mercial advantage, the ACM copyright notice and the title of llie publication and Its dale appear, and notice ia given that copying ja by pcmtlsaion of llie Association for Computing Machinery. To copy otherwise, or to republish, requires a fee and/or specific per mission. Section 2 describes the apparel manufacturing environment. The parallelization of the application is described in Section 3. Section 4 describes the syn chronization problems posed by the parallel simula tion. A survey of the methods given in the literature to solve the problems is provided in Section 5. The method chosen for this implementation is described in Section 6. T h e A p p a r e l F a c to r y S h o p -F lo o r On the factory floor bundles of garment parts are pro cessed according to a style flow graph, an example of which is given in Figure 1. Rectangles in the figure represent buffers which have work waiting for oper ations to be performed. Examples of operations are \"set pocket” and \"attach buttons”. There is a one-toone correspondence between buffers and the type of operation to be performed. Employees at sewing ma chines draw bundles from an assigned buffer, process them and send them to the next buffer in the style flow graph. An employee-machine combination, repre sented by a circle in the figure, is called a workstation. In general there may be more than one workstation in front of any given buffer as shown in Figure 2. In the style flow graph, different subparts such as the fronts and the backs of shirts need to merge at certain points. Such a merge is called an and-merge because further progress of the bundle depends on the availability of all the necessary subparts. Figure 3 is another example of a style flow graph. Bundles from a single buffer can be processed in alternate ways (called alternate paths). The output from these paths goes into a common buffer and continues on. This par ticular type of merge is called an or-inerge because a bundle from any of the paths can continue as soon as it appears in the buffer. As can be seen in Figure 2, after bundles from a buffer are processed on alternate workstations there is a “merge” into the next buffer. Such a “merge” is also an or-merge. In general there are many different products or styles being produced simultaneously on the factory floor. Various styles can share buffers. This is called “style intersection”. A style intersection also corre sponds to an or-merge.","PeriodicalId":409883,"journal":{"name":"ACM-SE 28","volume":"143 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"1990-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":"{\"title\":\"Approaches to solving synchronization problems in parallel simulation of an apparel plant\",\"authors\":\"P. Khambekar, S. Dharmaraj\",\"doi\":\"10.1145/98949.99143\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"A near-term computer simulation of an apparel in dustry shop-floor is an application of interest. The complexity and the natural parallelism of the applica tion suggest a parallel implementation. Parallelizing a simulation causes synchronization problems. These problems are analyzed, various solutions presented and short-comings or inapplicability to the application de scribed. A method which is an adaptation of a previ ously published method is then presented. This con servative method produces an accurate simulation. I n tr o d u c t io n Clemson University is currently developing a nearterm computer simulation of an apparel factory floor. Given inputs of employee assignments, arrival of new bundles of garment parts, priorities of bundles, etc. (i.e. a tentative plan), the flow of bundles being sewn during the course of a day is simulated and high-level performance information in the form of graphics is dis played to the plant manager (the user). The manager should be able to run the simulation, evaluate the per formance metrics and in case they are not satisfactory, roll-back the simulation to a specified point in time and rerun the simulation with a new plan all in a matter of minutes. This simulation will provide a valuable tool for achieving Just-In-Time manufacturing. In a typical plant, there are hundreds of ma chines, hundreds of employees and thousands of bun dles containing garment parts. The complexity of the application, large input data, large number of metrics, quick simulation requirement and natural parallelism of the application strongly suggest performing the sim ulation in parallel. Unfortunately, in a general-purpose parallel simulation two synchronization problems may occur: deadlock and no-progress. This paper describes the methods planned to address these problems. Peniiinnlon lo copy wllliout fee all or part of this material la granted provided that the copiea are not made or distributed for direct com mercial advantage, the ACM copyright notice and the title of llie publication and Its dale appear, and notice ia given that copying ja by pcmtlsaion of llie Association for Computing Machinery. To copy otherwise, or to republish, requires a fee and/or specific per mission. Section 2 describes the apparel manufacturing environment. The parallelization of the application is described in Section 3. Section 4 describes the syn chronization problems posed by the parallel simula tion. A survey of the methods given in the literature to solve the problems is provided in Section 5. The method chosen for this implementation is described in Section 6. T h e A p p a r e l F a c to r y S h o p -F lo o r On the factory floor bundles of garment parts are pro cessed according to a style flow graph, an example of which is given in Figure 1. Rectangles in the figure represent buffers which have work waiting for oper ations to be performed. Examples of operations are \\\"set pocket” and \\\"attach buttons”. There is a one-toone correspondence between buffers and the type of operation to be performed. Employees at sewing ma chines draw bundles from an assigned buffer, process them and send them to the next buffer in the style flow graph. An employee-machine combination, repre sented by a circle in the figure, is called a workstation. In general there may be more than one workstation in front of any given buffer as shown in Figure 2. In the style flow graph, different subparts such as the fronts and the backs of shirts need to merge at certain points. Such a merge is called an and-merge because further progress of the bundle depends on the availability of all the necessary subparts. Figure 3 is another example of a style flow graph. Bundles from a single buffer can be processed in alternate ways (called alternate paths). The output from these paths goes into a common buffer and continues on. This par ticular type of merge is called an or-inerge because a bundle from any of the paths can continue as soon as it appears in the buffer. As can be seen in Figure 2, after bundles from a buffer are processed on alternate workstations there is a “merge” into the next buffer. Such a “merge” is also an or-merge. In general there are many different products or styles being produced simultaneously on the factory floor. Various styles can share buffers. This is called “style intersection”. A style intersection also corre sponds to an or-merge.\",\"PeriodicalId\":409883,\"journal\":{\"name\":\"ACM-SE 28\",\"volume\":\"143 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"1990-04-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"2\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"ACM-SE 28\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1145/98949.99143\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"ACM-SE 28","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1145/98949.99143","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Approaches to solving synchronization problems in parallel simulation of an apparel plant
A near-term computer simulation of an apparel in dustry shop-floor is an application of interest. The complexity and the natural parallelism of the applica tion suggest a parallel implementation. Parallelizing a simulation causes synchronization problems. These problems are analyzed, various solutions presented and short-comings or inapplicability to the application de scribed. A method which is an adaptation of a previ ously published method is then presented. This con servative method produces an accurate simulation. I n tr o d u c t io n Clemson University is currently developing a nearterm computer simulation of an apparel factory floor. Given inputs of employee assignments, arrival of new bundles of garment parts, priorities of bundles, etc. (i.e. a tentative plan), the flow of bundles being sewn during the course of a day is simulated and high-level performance information in the form of graphics is dis played to the plant manager (the user). The manager should be able to run the simulation, evaluate the per formance metrics and in case they are not satisfactory, roll-back the simulation to a specified point in time and rerun the simulation with a new plan all in a matter of minutes. This simulation will provide a valuable tool for achieving Just-In-Time manufacturing. In a typical plant, there are hundreds of ma chines, hundreds of employees and thousands of bun dles containing garment parts. The complexity of the application, large input data, large number of metrics, quick simulation requirement and natural parallelism of the application strongly suggest performing the sim ulation in parallel. Unfortunately, in a general-purpose parallel simulation two synchronization problems may occur: deadlock and no-progress. This paper describes the methods planned to address these problems. Peniiinnlon lo copy wllliout fee all or part of this material la granted provided that the copiea are not made or distributed for direct com mercial advantage, the ACM copyright notice and the title of llie publication and Its dale appear, and notice ia given that copying ja by pcmtlsaion of llie Association for Computing Machinery. To copy otherwise, or to republish, requires a fee and/or specific per mission. Section 2 describes the apparel manufacturing environment. The parallelization of the application is described in Section 3. Section 4 describes the syn chronization problems posed by the parallel simula tion. A survey of the methods given in the literature to solve the problems is provided in Section 5. The method chosen for this implementation is described in Section 6. T h e A p p a r e l F a c to r y S h o p -F lo o r On the factory floor bundles of garment parts are pro cessed according to a style flow graph, an example of which is given in Figure 1. Rectangles in the figure represent buffers which have work waiting for oper ations to be performed. Examples of operations are "set pocket” and "attach buttons”. There is a one-toone correspondence between buffers and the type of operation to be performed. Employees at sewing ma chines draw bundles from an assigned buffer, process them and send them to the next buffer in the style flow graph. An employee-machine combination, repre sented by a circle in the figure, is called a workstation. In general there may be more than one workstation in front of any given buffer as shown in Figure 2. In the style flow graph, different subparts such as the fronts and the backs of shirts need to merge at certain points. Such a merge is called an and-merge because further progress of the bundle depends on the availability of all the necessary subparts. Figure 3 is another example of a style flow graph. Bundles from a single buffer can be processed in alternate ways (called alternate paths). The output from these paths goes into a common buffer and continues on. This par ticular type of merge is called an or-inerge because a bundle from any of the paths can continue as soon as it appears in the buffer. As can be seen in Figure 2, after bundles from a buffer are processed on alternate workstations there is a “merge” into the next buffer. Such a “merge” is also an or-merge. In general there are many different products or styles being produced simultaneously on the factory floor. Various styles can share buffers. This is called “style intersection”. A style intersection also corre sponds to an or-merge.