Die Beobachtung (das Monitoring) der inneren Ablaufe in Multiprozessor- und Multicomputer-Systemen ist ein hervorragendes Werkzeug, um die Ursachen fur die extern festgestellte Gesamtleistung solcher Systeme zu analysieren und damit die Voraussetzungen fur eine Leistungsverbesserung zu schaffen. Im vorliegenden Papier wird ein Monitorsystem vorgestellt, das speziell fur (auch raumlich verteilte) Multiprozessor- und Multicomputer-Systeme entworfen wurde. Seine wesentlichen Eigenschaften sind die Kombination von verteilter und zentraler Instrumentierung des Objektsystems, die systemweite Ermittlung von Ereignisreihenfolgen und die quellbezogene Auswertung von Ereignisspuren. Neben dem globalen Konzept werden vertiefende Untersuchungen dargestellt, und es wird von ersten Meserfahrungen berichtet.

ZÄHLMONITOR 4: Ein Monitorsystem für das Hardware- und Hybrid-Monitoring von Multiprozessor- und Multicomputer-Systemen

The authors propose a centralized solution to physical clock synchronization in real-time distributed systems. The solution is dedicated to systems based on the factory instrumentation protocol (FIP) field bus. It is a fault-tolerant solution, it does not use dynamic election of a privileged station, and it allows minimization of the synchronization error, which is independent of the communication delay, to obtain a good synchronization accuracy. The initial states of clocks are not necessarily coherent. It is shown how characteristics of the FIP field bus enable such a solution to be developed. >

http://ieeexplore.ieee.org/iel2/315/3750/00138308.pdf

Clock synchronization in real-time distributed systems based on FIP field bus

The purpose of this report is threefold: (1) to identify communication functions which are considered useful within a communication system; (2) to document the extent to which these functions are incorporated in extant protocols; and (3) to identify those communication functions which are potentially useful for a real-time communication system. To accomplish these goals, section two makes some observations about the environment in which the real-time communication system is to serve — namely, real-time distributed systems. Section three identifies the principles of real-time communication as reported in the archival literature. To provide a framework for the eventual comparison of protocols, section four briefly discusses the ISO OSI Reference Model and its suitability as a paradigm for real-time communications. Section five identifies those functions which are commonly found in transport protocols. In section six we discuss eight specific network protocols and characterize each by identifying the presense or absence of our identified transport functionalities. Section seven evaluates the protocols and comments on how well each would support real-time communication. Our conclusions are in section eight. A real-time distributed system consists of some number of application programs (processes) which are physically distributed over a multiplicity of processing elements (processors), interconnected by a communications server (e.g., a local area network and its protocols). What makes the system real-time is that processes must complete their computations within a known and bounded time. The processes may be periodic or aperiodic (interrupt-driven), and the resources they require may be either local (on the same processor) or remote (on another processor). The purpose of the communications subsystem is to provide 2 fast, transparent, and reliable access to remote resources without jeopardizing the time constraints of the application programs. To appreciate the importance of the underlying communications server, consider the range of implementation possibilities. At one end of the spectrum, all processes could be assigned to a single processor. This would minimize the communications delay because all message passing would occur in the memory of the local processor, but would maximize computational delay because parallelism could not exist. At the other end of the spectrum, every process could be assigned to a different processor. Now computational delay is minimized because all modules execute in parallel, but communication delays are maximized because every process needs the communications server for every transaction. Since remote (processor-to-processor) communication can be orders of magnitude more "expensive" (in terms of delay) than process-to-process communication …

Evaluation of Transport Protocols for Real-Time Communication

Analyzing Tradeoffs between Temporary Consistency and Concurrency with Rollbacks and Aborts.

Distributed systems demonstrate characteristics that will provide benefits, such as reliability, availability, ease of modularity, incremental growth, configuration flexibility, and high system performance. However, the increased activity in distributed systems have yielded a substantial lack in methods and tools for monitoring and measuring these systems to verify these features.

A Tool for measuring and Monitoring Distributed Systems During Operation

The coordination of flexible manufacturing systems (FMS) in an automated factory requires that synchronization amongst the manufacturing processes be based on a common clock. The synchronization requirements on the factory floor are described, and several clock synchronization algorithms, their theoretical bounds, and the results of the authors' work are discussed. Measurement results based on the implementation of such synchronization algorithms on local area networks (LAN) are presented. For hierarchical LANs, an algorithm is developed and its behavior simulated. >

https://ieeexplore.ieee.org/iel1/41/176/00003109.pdf

S.K. Tripathi

Papers

Clock synchronization on the factory floor (FMS)