Redistribution in Distributed Ada

ثبت نشده
چکیده

In this paper we will provide a model using Ada and the Distributed Annex for relocating concurrent objects in a distributed dataaow application. This model will be used to provide passive and active error tolerance as well as the basis for a load balancing superstructure. Care will be taken to demonstrate how errors are trapped and propagated across partitions containing multiple threads of execution. The following is intended as an extended abstract. If accepted, the nal paper will contain pictures, code segments, examples, and execution speciics running under a network of workstations operating Solaris using a GNATTGLADE implementation. 2 The Model We rst summarize a model presented in a prior paper 3. A dataaow graph is deened as a directed acyclic graph with data passing through the directed edges, or queues, and dynamics at the nodes. Each queue is connected to exactly two nodes: an upstream node attached to its tail and a downstream node attached to its head data ows downstream. Each node possesses a set of input queues entering it, and a set of output queues leaving it, either set possibly being empty. Each queue contains a threshold and a buuer size and receives its data in a rst-in-rst-out fashion from its upstream node. Queues are said to be linked to nodes at ports, thus there are both output ports and input ports. When a queue accumulates a threshold of data, it is said to have reached threshold. When every upstream queue of a node has reached threshold, the node is ready to re. Upon ring, a node reads and consumes data from each of its upstream queues, processes this data, and writes results to its downstream queues. All nodes can function concurrently, and conceptually begin processing as soon as they reach a re condition. To allow for maximal concurrency in a distributed system, we implement queues as protected types, nodes as tasks, and provide an additional protected object a mailbox so that queues may send asynchronously to nodes. gure Each node has three mutually exclusive accept states Figure 1: 1. waiting to be initialized 2. waiting for the next command from its mailbox processing 3. waiting to be nalized 1

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

The Distributed Ada Run-time System DARTS

A distributed Ada run-time system, DARTS, is presented. DARTS is used with a source code transformation method for pre-partitioning, and it also allows a late configuration. A single program can be partitioned to run on a loosely coupled multiprocessor system. The distributed units are tasks, task objects, packages, variables, procedures and functions. Task objects can by dynamically distribute...

متن کامل

P-142: Adenosine Deaminase Activity during Menses, Follicular and Luteal Phases of Menstrual Cycle

Background: In recent years, numerous regulators of gonadal function have been studied. Adenosine deaminase(ADA) is widely distributed throughout human tissues and may contribute in the regulation of menstrual cycle. The purpose of this study was to determine the plasma activities of total adenosine deaminase (ADAT), and its isoenzymes, ADA1 and ADA2, and ADA1/ADA2 ratio during the menses, foll...

متن کامل

Supporting Ada 95 Passive Partitions in a Distributed Environment

Ada 95 passive partitions, containing passive library units, provide the means to distribute data within a network of workstations. This paper shows how passive partitions can be implemented via distributed shared virtual memory (DSM). DSM provides the logical view of a portion of memory shared between physically distributed workstations within a network. In this paper, we relate design issues ...

متن کامل

Secure Communication in Distributed Ada

This document describes an implementation of secure communication for distributed applications in Ada 95. It explains the basics of cryptography and introduces the problem of key exchange. After a brief description of distributed systems in Ada 95, the modular structure of the implementation and the way it handles key distribution is presented.

متن کامل

Fault Tolerance by Transparent Replication for Distributed Ada 95

This paper presents the foundations of RAPIDS (“Replicated Ada Partitions In Distributed Systems”), an implementation of the Distributed Systems Annex E incorporating the transparent replication of partitions in distributed Ada 95 applications. RAPIDS is a replication manager for semi–active replication based on a piecewise deterministic computation model. It guarantees replica consistency for ...

متن کامل

Chorus: a Support for Distributed and Reconfigurable Ada Software

In the ESA Columbus Project, including in particular the European Space Station, applications are written in Ada and are distributed; moreover, the long lifetime of the space elements requires flexibility in order to support smooth software evolution. CHORUS is a distributed operating system, developed and supported by Chorus systèmes, which provides a basis for supporting efficiently distribu...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 1999