Pub Date : 1994-04-28DOI: 10.1109/WPDRTS.1994.365633
K. Toda, Kyosuke Nishida, E. Takahashi, N. Michell, Y. Yamaguchi
A single-chip VLSI implementation of a 4 by 4 prioritized router for multistage real-time interconnection networks is presented. The chip employs packet switching and facilitates 32-bit priority arbitration by means of a priority forwarding scheme that prevents priority inversion and which provides accurate priority control in a network. The packets are of fixed size, having three 38-bit segments: a header and two bodies. Each input port has an 8-packet priority queue for simultaneous input and output, enabling virtual cut-through routing. The chip is pipelined with a 25-ns pitch and reduces the number of stages to two by overlapping the arbitration and priority queue stages. Hence, its data transmission rate is 190 MByte/s per port. The end-to-end delay of an s-stage network is 25/spl times/(2s+1) ns.<>
{"title":"Implementation of a priority forwarding router chip for real-time interconnection networks","authors":"K. Toda, Kyosuke Nishida, E. Takahashi, N. Michell, Y. Yamaguchi","doi":"10.1109/WPDRTS.1994.365633","DOIUrl":"https://doi.org/10.1109/WPDRTS.1994.365633","url":null,"abstract":"A single-chip VLSI implementation of a 4 by 4 prioritized router for multistage real-time interconnection networks is presented. The chip employs packet switching and facilitates 32-bit priority arbitration by means of a priority forwarding scheme that prevents priority inversion and which provides accurate priority control in a network. The packets are of fixed size, having three 38-bit segments: a header and two bodies. Each input port has an 8-packet priority queue for simultaneous input and output, enabling virtual cut-through routing. The chip is pipelined with a 25-ns pitch and reduces the number of stages to two by overlapping the arbitration and priority queue stages. Hence, its data transmission rate is 190 MByte/s per port. The end-to-end delay of an s-stage network is 25/spl times/(2s+1) ns.<<ETX>>","PeriodicalId":275053,"journal":{"name":"Second Workshop on Parallel and Distributed Real-Time Systems","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129632266","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1994-04-28DOI: 10.1109/WPDRTS.1994.365631
H.T. Olnowich, D. Kirk
This paper presents a proposed multi-stage switching network for real-time systems which features: priority scheduling scalability, high bandwidth, low latency, fault tolerance, and breaking of lower priority connections. The fault-tolerant approach presented gives continuous availability in the presence of many failures, and the correction time is so rapid that real-time messages can be delivered before their deadlines expire. This is possible because the ALLNODE-RT Network contains numerous alternate paths between any two nodes. The ALLNODE-RT hardware searches for an available, non-failed path and guarantees the meeting of a deadline by dynamically increasing priority as the deadline approaches and breaking lower priority connections. The network is based on the existing ALLNODE Switch concept. The modifications required to evolve to a real-time switch are described, as well as the fault tolerance concepts.<>
{"title":"ALLNODE-RT: a real time, fault tolerant network","authors":"H.T. Olnowich, D. Kirk","doi":"10.1109/WPDRTS.1994.365631","DOIUrl":"https://doi.org/10.1109/WPDRTS.1994.365631","url":null,"abstract":"This paper presents a proposed multi-stage switching network for real-time systems which features: priority scheduling scalability, high bandwidth, low latency, fault tolerance, and breaking of lower priority connections. The fault-tolerant approach presented gives continuous availability in the presence of many failures, and the correction time is so rapid that real-time messages can be delivered before their deadlines expire. This is possible because the ALLNODE-RT Network contains numerous alternate paths between any two nodes. The ALLNODE-RT hardware searches for an available, non-failed path and guarantees the meeting of a deadline by dynamically increasing priority as the deadline approaches and breaking lower priority connections. The network is based on the existing ALLNODE Switch concept. The modifications required to evolve to a real-time switch are described, as well as the fault tolerance concepts.<<ETX>>","PeriodicalId":275053,"journal":{"name":"Second Workshop on Parallel and Distributed Real-Time Systems","volume":"47 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133619002","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1994-04-28DOI: 10.1109/WPDRTS.1994.365653
Guohui Yu
To improve the performance of programs composed of abstract data types (ADTs), a parallel execution model is used to exploit concurrency at the method level. Typically, an ADT instance is used to manage more than one variable, and there may be contention for getting access to the ADT instance if multiple variables need to be accessed concurrently. To resolve the contention, the ADT instance is cloned, and the copies are placed in different processors so that the multiple copies can be used concurrently. To exploit automatically the concurrency available via ADT instance cloning, we analyze program dependence relations. Not only control and data dependence but also instance dependence is analyzed to reveal dependence relations among method calls and to determine the clonability of each ADT instance. The focus of this paper is the use of our ADT cloning techniques to help ease the task of off-line scheduling in hard-real-time systems. Cloning and ARPCs are applied repeatedly to increase concurrency within processes missing deadlines, thus reducing their execution times.<>
{"title":"Use of concurrency enhancement in off-line schedule construction","authors":"Guohui Yu","doi":"10.1109/WPDRTS.1994.365653","DOIUrl":"https://doi.org/10.1109/WPDRTS.1994.365653","url":null,"abstract":"To improve the performance of programs composed of abstract data types (ADTs), a parallel execution model is used to exploit concurrency at the method level. Typically, an ADT instance is used to manage more than one variable, and there may be contention for getting access to the ADT instance if multiple variables need to be accessed concurrently. To resolve the contention, the ADT instance is cloned, and the copies are placed in different processors so that the multiple copies can be used concurrently. To exploit automatically the concurrency available via ADT instance cloning, we analyze program dependence relations. Not only control and data dependence but also instance dependence is analyzed to reveal dependence relations among method calls and to determine the clonability of each ADT instance. The focus of this paper is the use of our ADT cloning techniques to help ease the task of off-line scheduling in hard-real-time systems. Cloning and ARPCs are applied repeatedly to increase concurrency within processes missing deadlines, thus reducing their execution times.<<ETX>>","PeriodicalId":275053,"journal":{"name":"Second Workshop on Parallel and Distributed Real-Time Systems","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115395120","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1994-04-28DOI: 10.1109/WPDRTS.1994.365647
G. Ramanathan
Work on scheduling concurrent transactions in real-time databases must address two issues: (i) synchronization of the corresponding tasks' accesses to shared data items (ii) guaranteeing timing requirements of the transactions. In this paper, first, we present a concurrency control protocol for real-time databases in a uniprocessor system. In this protocol, we consider the system characteristics to be dynamic. This is in contrast to the priority ceiling protocol and most of the work in scheduling theory where the system workload is assumed to be static and predetermined. Priorities are assigned dynamically to transactions using the well-known strategy Earliest Deadline first. The protocol is proven to avoid deadlocks. The blocking duration arising from mutual exclusion of shared resources is bounded under this protocol. Schedulability analysis for dynamically occurring transactions is provided. Next, we extend the protocol for distributed databases in a shared memory multiprocessor system. The protocol for distributed databases is shown to have the properties of the uniprocessor protocol.<>
{"title":"Scheduling transactions in real-time distributed databases","authors":"G. Ramanathan","doi":"10.1109/WPDRTS.1994.365647","DOIUrl":"https://doi.org/10.1109/WPDRTS.1994.365647","url":null,"abstract":"Work on scheduling concurrent transactions in real-time databases must address two issues: (i) synchronization of the corresponding tasks' accesses to shared data items (ii) guaranteeing timing requirements of the transactions. In this paper, first, we present a concurrency control protocol for real-time databases in a uniprocessor system. In this protocol, we consider the system characteristics to be dynamic. This is in contrast to the priority ceiling protocol and most of the work in scheduling theory where the system workload is assumed to be static and predetermined. Priorities are assigned dynamically to transactions using the well-known strategy Earliest Deadline first. The protocol is proven to avoid deadlocks. The blocking duration arising from mutual exclusion of shared resources is bounded under this protocol. Schedulability analysis for dynamically occurring transactions is provided. Next, we extend the protocol for distributed databases in a shared memory multiprocessor system. The protocol for distributed databases is shown to have the properties of the uniprocessor protocol.<<ETX>>","PeriodicalId":275053,"journal":{"name":"Second Workshop on Parallel and Distributed Real-Time Systems","volume":"71 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126406304","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1994-04-28DOI: 10.1109/WPDRTS.1994.365630
J. Rexford, J. Dolter
Real-time communication typically consists of guaranteed packets that must satisfy their delivery deadlines and best-effort packets that can tolerate occasional deadline misses for improved average latency. This paper presents hardware techniques for supporting the coexistence of these two traffic classes in real-time point-to-point networks. A careful selection of routing and switching techniques, coupled with fine-grain arbitration between traffic classes, can allow network adapters to support the diverse performance requirements of best-effort and guaranteed communication. Cycle-level simulations of SPIDER (Scalable Point-to-point Interface DrivER), a network adapter for point-to-point distributed systems, demonstrate the utility of supporting multiple low-level communication policies for different classes of traffic.<>
{"title":"Hardware support for controlled interaction of guaranteed and best-effort communication","authors":"J. Rexford, J. Dolter","doi":"10.1109/WPDRTS.1994.365630","DOIUrl":"https://doi.org/10.1109/WPDRTS.1994.365630","url":null,"abstract":"Real-time communication typically consists of guaranteed packets that must satisfy their delivery deadlines and best-effort packets that can tolerate occasional deadline misses for improved average latency. This paper presents hardware techniques for supporting the coexistence of these two traffic classes in real-time point-to-point networks. A careful selection of routing and switching techniques, coupled with fine-grain arbitration between traffic classes, can allow network adapters to support the diverse performance requirements of best-effort and guaranteed communication. Cycle-level simulations of SPIDER (Scalable Point-to-point Interface DrivER), a network adapter for point-to-point distributed systems, demonstrate the utility of supporting multiple low-level communication policies for different classes of traffic.<<ETX>>","PeriodicalId":275053,"journal":{"name":"Second Workshop on Parallel and Distributed Real-Time Systems","volume":"26 19","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132273751","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1994-04-28DOI: 10.1109/WPDRTS.1994.365641
S. Oh, G. MacEwen
A compound service comprising one or more task, occurring unpredictably and constrained by a single deadline may not be successfully scheduled. A goal for managing such services as to maximize the success ratio SR, i.e. the percentage of successfully scheduled services. This paper describes an adaptive run-time scheduling method for improving the SR of a real-time distributed system. Simulation is used to obtain performance results.<>
{"title":"Compound service scheduling with run-time adaptation in real-time multiprocessor distributed systems","authors":"S. Oh, G. MacEwen","doi":"10.1109/WPDRTS.1994.365641","DOIUrl":"https://doi.org/10.1109/WPDRTS.1994.365641","url":null,"abstract":"A compound service comprising one or more task, occurring unpredictably and constrained by a single deadline may not be successfully scheduled. A goal for managing such services as to maximize the success ratio SR, i.e. the percentage of successfully scheduled services. This paper describes an adaptive run-time scheduling method for improving the SR of a real-time distributed system. Simulation is used to obtain performance results.<<ETX>>","PeriodicalId":275053,"journal":{"name":"Second Workshop on Parallel and Distributed Real-Time Systems","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130891769","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1994-04-28DOI: 10.1109/WPDRTS.1994.365643
C. Puchol, A. Mok
The control structure of many real-time applications are naturally described by state machines. However the state transitions of these machines are governed not only by the state of the system and the occurrence of events, but also by their time of occurrence. The control structure of a real-time system determines what computation to perform and the set of timing constraints in effect at all times. The computation performed by the system can then be modeled by a directed graph, where the nodes denote transformations on the data and the edges denote data flow. In this paper we discuss some research issues that arise from the integration of control flow with data flow.<>
{"title":"The integration of control and dataflow structures in distributed hard real-time systems","authors":"C. Puchol, A. Mok","doi":"10.1109/WPDRTS.1994.365643","DOIUrl":"https://doi.org/10.1109/WPDRTS.1994.365643","url":null,"abstract":"The control structure of many real-time applications are naturally described by state machines. However the state transitions of these machines are governed not only by the state of the system and the occurrence of events, but also by their time of occurrence. The control structure of a real-time system determines what computation to perform and the set of timing constraints in effect at all times. The computation performed by the system can then be modeled by a directed graph, where the nodes denote transformations on the data and the edges denote data flow. In this paper we discuss some research issues that arise from the integration of control flow with data flow.<<ETX>>","PeriodicalId":275053,"journal":{"name":"Second Workshop on Parallel and Distributed Real-Time Systems","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133942589","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1994-04-28DOI: 10.1109/WPDRTS.1994.365644
N. Kosugi, K. Takashio, M. Tokoro
Many researches on dynamic scheduling techniques for real-time systems have focused on changes in the number of tasks. We propose a dynamic real-time scheduling technique, called MART, in which it is also possible to change task characteristics. When MART is used in systems where task characteristics need to be and can be modified, it enables the system to be adaptable to changing environments dynamically and flexibly. By using MART, systems can accept on-line requirements such as changing task timing characteristics, and adding or deleting tasks. If a task set becomes unschedulable as a result of such modifications, MART can determine a schedulable task set. Though MART must adjust timing characteristics of other tasks to guarantee the schedulability of all tasks, the need for such adjustments is reduced. We also argue the importance of predictability in real-time computing, then improve the predictability of MART by introducing the notion of importance into each task. MART uses "importance" as a factor for task characteristic adjustment.<>
{"title":"Modification and adjustment of real-time tasks with rate monotonic scheduling algorithm","authors":"N. Kosugi, K. Takashio, M. Tokoro","doi":"10.1109/WPDRTS.1994.365644","DOIUrl":"https://doi.org/10.1109/WPDRTS.1994.365644","url":null,"abstract":"Many researches on dynamic scheduling techniques for real-time systems have focused on changes in the number of tasks. We propose a dynamic real-time scheduling technique, called MART, in which it is also possible to change task characteristics. When MART is used in systems where task characteristics need to be and can be modified, it enables the system to be adaptable to changing environments dynamically and flexibly. By using MART, systems can accept on-line requirements such as changing task timing characteristics, and adding or deleting tasks. If a task set becomes unschedulable as a result of such modifications, MART can determine a schedulable task set. Though MART must adjust timing characteristics of other tasks to guarantee the schedulability of all tasks, the need for such adjustments is reduced. We also argue the importance of predictability in real-time computing, then improve the predictability of MART by introducing the notion of importance into each task. MART uses \"importance\" as a factor for task characteristic adjustment.<<ETX>>","PeriodicalId":275053,"journal":{"name":"Second Workshop on Parallel and Distributed Real-Time Systems","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122448869","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1994-04-28DOI: 10.1109/WPDRTS.1994.365640
J. Apsel, K. Ecker, W. Halang
In this paper we consider the problem of scheduling hard real time tasks on a set of identical processors. Tasks are assumed to arrive for service at previously unknown points of time at the system. A decision has to be made whether a new task is rejected or accepted for processing. Once a task is accepted the scheduling policy assures that the task will be processed on time.<>
{"title":"Throwforward task scheduling strategy in real time systems","authors":"J. Apsel, K. Ecker, W. Halang","doi":"10.1109/WPDRTS.1994.365640","DOIUrl":"https://doi.org/10.1109/WPDRTS.1994.365640","url":null,"abstract":"In this paper we consider the problem of scheduling hard real time tasks on a set of identical processors. Tasks are assumed to arrive for service at previously unknown points of time at the system. A decision has to be made whether a new task is rejected or accepted for processing. Once a task is accepted the scheduling policy assures that the task will be processed on time.<<ETX>>","PeriodicalId":275053,"journal":{"name":"Second Workshop on Parallel and Distributed Real-Time Systems","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123743478","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1994-04-28DOI: 10.1109/WPDRTS.1994.365654
St, SchloB Birlinghoven
In this paper an on-line scheduling approach called TaskPair-Scheduling is presented which merges the concepts of guaranteeing (an activity) and exception handling (due to time-outs) such that the guarantee may enclose the execution of the exception handler. An on-line scheduling algorithm is shown which plans guaranteed try-except constructs as a pair of tasks (MainTask, ExceptTask) by considering the ExceptTask as a hard real-time task (task of maximum importance) and the MainTask by its relative importance (this may depend on arbitrary criteria). This leads to timely predictable behaviour even if the MainTask of the activity has an arbitrary execution time.<>
{"title":"TaskPair-scheduling: an approach for dynamic real-time systems","authors":"St, SchloB Birlinghoven","doi":"10.1109/WPDRTS.1994.365654","DOIUrl":"https://doi.org/10.1109/WPDRTS.1994.365654","url":null,"abstract":"In this paper an on-line scheduling approach called TaskPair-Scheduling is presented which merges the concepts of guaranteeing (an activity) and exception handling (due to time-outs) such that the guarantee may enclose the execution of the exception handler. An on-line scheduling algorithm is shown which plans guaranteed try-except constructs as a pair of tasks (MainTask, ExceptTask) by considering the ExceptTask as a hard real-time task (task of maximum importance) and the MainTask by its relative importance (this may depend on arbitrary criteria). This leads to timely predictable behaviour even if the MainTask of the activity has an arbitrary execution time.<<ETX>>","PeriodicalId":275053,"journal":{"name":"Second Workshop on Parallel and Distributed Real-Time Systems","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128369671","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：