Assignment and pre-run-time scheduling of object-based, parallel real-time processes

Abstract

The use of object-based programming techniques h.elps to reduce the cost of software development and maindenance, due to the benefits of reuse, information h.idin.g and encapsulation. This is especially helpful in complex, real-time systems that are highly parallel a,nd distributed, due to thear magnitude. The paper presents a semi-preemption execution. model of objectbased real-time systems th,at simplifies reasoning about th.e quality of particular-process-to-processor assignments. Th,e m,odel is used t o define system properties such. as in,ter-process parallelism, processor utilization and inter-processor communication. Additionally, we present an innovative assignment algorithm that incorporates feasibility constraints, and is guided b y an objective th.at balances minimum communication agnin.st maxim.um. parallelism. This work is part of a complete nssignin.ent an.d pre-run-tim.e scheduling approach for distributed real-time systems. IiitroductionDuring the last decade, dependability of computer systems has become ail increasingly important topic. In addition t,o correct functionality, dependability includes tim.eliness, i.e., real-time constraints imposed on t,he t,asks of the system must be satisfied. A condit,ion often required to guarantee timeliness is that applicat,ions and the environment have to be fully, or almost fully, predict>able. Otherwise, due to the overhead of scheduling a transient burst of new tasks at run-time, it might become impossible to satisfy the HRT constraints of cert,ain tasks. For instance, simu1a.tion results of an evduation of the run-time scheduling strategy described in [7] show that timeliness cannot be guaranteed in all circumstances. Given such a model, schedulability analysis techniques can be used t,o analyze an application [lo]. However, we prefer to schedule a, hard real-time system pre-run-time, such that. all t,iming constraint,s are satisfied and consistency of all resources, including processors, communication media and devices, is guaranteed [12]; with this approach, the run-time system simply dispatches and executes tasks according to the schedule constructed pre-run-time.We consider a distributed memory MIMD system that consists of a set of homogeneous processors P R , connected by a set of communication media CM. A communication medium is either a bus that connects a set of usually more than two processors or a link that connects two processors directly. The only assumption made about the interconnection configuration is that there is a set routes consisting of links, busses and processors between each pair of different processors. Each route is a sequence of triples sprj E PR, cmj E C M , rpri E PR, that indicates via which processors, busses and links, a communication from processor sprl to rprr has to be routed. Additionally, the system contains several shared memory modules and a set of physical devices DV, of different types such as sensor, actuator, and disk. Each device is assigned and physically connected to a particular processor after assignment of ...

Cocoon

“…While this paper focuses on the partitioning and assignment toolset, related papers emphasize other aspects of the process. Concurrency analysis techniques are pre-~zd in [4,8,9] and assignment algorithms are detailed in [5,9].…”

Metrics for assessment of designs and implementations during reengineering of computer-based systems

“…Software components are optimized for the execution paradigm provided by the hardware-operating system platform. The optimized software components are partitioned into tightly coupled clusters, which are assigned [Verhoosel et al 1994;Welch et aL 1993;] to the hardware platform in a way that (1) satisfies the new system requirements and (2) considers the new system objectives. Figure 2.…”

“…An obvious preemption point is when an activity switches to another process by performing a call to a method of an object in that process, because a context switch is inevitable. For the same reasons, a return from such a 3) A complete discussion of activities and beads is beyond the scope of this paper; for a comprehensive treatment of our representation of tasks/activities and beads, see [Verhoosel et al 1995].…”

“…loose coupling of software components) [Sitaraman et al 1993], abstraction (highly cohesive modules), and ease of understandability. Another benefit of layering is the simplification of analyses for concurrency and timing properties [Welch 1993[Welch , 1994Verhoosel et al 1993Verhoosel et al , 1995Stoyenko et al 1994] and dependability [Marlowe et al 1994].…”