Enterprise Information Systems can be described according to the Open distributed processing Reference Model (RM-ODP), where five different viewpoints are specified analyzing various aspects of the system. Configurati...
详细信息
Enterprise Information Systems can be described according to the Open distributed processing Reference Model (RM-ODP), where five different viewpoints are specified analyzing various aspects of the system. Configuration issues are explored in the Engineering Viewpoint of RM-ODP. In practice, configuration issues are explored in discrete stages, supported by autonomous software tools, each of which adopts its own metamodel for system representation. We propose a platform independent framework, which focuses on the Engineering Viewpoint of Enterprise Information Systems managing application configuration and network design issues independently of application development progress. In order to apply this framework using existing tools, model exchangeability and tool coordination must be supported by standard, open methods. Thus, a common metamodel is proposed to generate enterprise information system models, which are stored in XML. A UML 2.0 profile is defined to visualize these models, facilitate the designer to interact with them and coordinate specific tool invocation.
Although ISDN and OSI erupted in the world of telecommunications a dozen years ago, it is fair to recognize that the marriage has not completely been consummated between telephony and data communication. This is parti...
详细信息
Although ISDN and OSI erupted in the world of telecommunications a dozen years ago, it is fair to recognize that the marriage has not completely been consummated between telephony and data communication. This is particularly true for ISDN network signaling, in the CCITT 7 environment where the habits and customs of telephony still prevail. Circuit switching is today, within the network, the backbone of ISDN services. Tomorrow we can expect new services that go beyond what we experience and imagine now. The design of an integrated, intelligent and flexible signaling system to allow easy evolution of services and flexible distribution of intelligence within the network, is a key to this future. This paper discusses the concepts which span the study of such a new signaling system.
In current scientific workflows, the computational needs of tasks might not be known when it is submitted to a system for execution. Current resource management (RM) systems and workflow managers (WFMs) provide limite...
详细信息
ISBN:
(数字)9798350364606
ISBN:
(纸本)9798350364613
In current scientific workflows, the computational needs of tasks might not be known when it is submitted to a system for execution. Current resource management (RM) systems and workflow managers (WFMs) provide limited support for dynamic resource allocation in HPC systems, thus the common approach is to request the maximum resources needed for a maximum time, potentially wasting resources. However, in some cases, maximum resources may not be estimated a priori, as a result, a workflow may be completed after the deadline, or in cases, the task may terminated by the resource manager. A combination of workflow manager and resource management system that can accommodate a fine-grain elastic resource allocation during the execution of a workflow would alleviate this problem. This paper presents a dynamic elastic resource management framework based on the Parsl workflow manager and PMIx-enabled SLURM and reports the early evaluation of the framework using two workflow applications.
Federated Learning is a distributed model training system that uses the computing resources and private data of participants to collaboratively train Machine Learning models. In order to incentivize data holders to ac...
详细信息
ISBN:
(数字)9798331509712
ISBN:
(纸本)9798331509729
Federated Learning is a distributed model training system that uses the computing resources and private data of participants to collaboratively train Machine Learning models. In order to incentivize data holders to actively participate in Federated Learning, a crucial issue is how to fairly assess each participant’s contribution. One class of the mainstream methods estimates client contributions by calculating the correlation between local and global model parameters, substantially reducing the reliance on large computing resources and test datasets. However, these methods overlook an implicit key factor, leading to inaccuracy in assessing the contribution of the participants. We propose FedCA, which provides a fairness assessment of the client’s contribution based on the performance gain of the global model in each communication round. The experimental results demonstrate that FedCA accurately identifies the contributions of clients with varying data qualities and effectively approximates the optimal fairness method of Shapley values. The source code for this paper is available at https://***/paper-liu/***.
Node architectures of extreme-scale systems are rapidly increasing in complexity. Emerging homogenous and heterogeneous designs provide massive multi-level parallelism, but developing efficient runtime systems and mid...
详细信息
Node architectures of extreme-scale systems are rapidly increasing in complexity. Emerging homogenous and heterogeneous designs provide massive multi-level parallelism, but developing efficient runtime systems and middleware that allow applications to efficiently and productively exploit these architectures is extremely challenging. Moreover, current state-of-the-art approaches may become unworkable once energy consumption, resilience, and data movement constraints are added. The goal of this workshop is to attract the international research community to share new and bold ideas that will address the challenges of design, implementation, deployment, and evaluation of future runtime systems and middleware.
A new learning algorithm for multilayer neural networks is introduced. The algorithm merges the gradient technique and least squares technique. The presented algorithm is compared with backpropagation learning rule on...
详细信息
A new learning algorithm for multilayer neural networks is introduced. The algorithm merges the gradient technique and least squares technique. The presented algorithm is compared with backpropagation learning rule on the parity problem. Simulation results are presented.
This paper presents the performance evaluations of GMP-based applications on Web-based Volunteer Computing (VC) systems. Web-based VC is expected to gather many volunteer participants (workers) by allowing workers to ...
详细信息
This paper presents the performance evaluations of GMP-based applications on Web-based Volunteer Computing (VC) systems. Web-based VC is expected to gather many volunteer participants (workers) by allowing workers to execute a computation program (job) on Web browsers. On the other hand, the job execution performance on workers degraded because jobs are executed on Web browsers. To reveal the actual performance of Web-based VC, we convert practical applications which use GMP, that is a multi-precision library for scientific computations, and evaluate its performance. The experimental results show that the performance degradation is negligibly small in some cases, e.g. a short bit-length of arguments. This paper also shows a potential for the performance improvement of Web-based VC by substituting GMP functions.
HERON is a platform for object-oriented distributed computing in an open systems environment. We try to achieve a degree of distribution transparency previously known only from special distributed programming systems,...
详细信息
HERON is a platform for object-oriented distributed computing in an open systems environment. We try to achieve a degree of distribution transparency previously known only from special distributed programming systems, while at the same time accommodating heterogeneous, autonomous computer systems. distributed programs are written in Eiffel. The Eiffel language system is not modified: HERON employs proxies for remote object invocation and a flexible configuration procedure for building servers and distributed programs. In addition to regular objects, two kinds of distributed objects are supported by the proxy generator: dispersed objects and objects fragmented by remote inheritance. They contribute to distribution transparency both for distributed programs and for client/server systems.< >
As part of the ISO/IEC standardisation of the reference model for open distributed processing (RM-ODP), a trading function has been identified. It is an important infrastructure service that allows clients to locate s...
详细信息
As part of the ISO/IEC standardisation of the reference model for open distributed processing (RM-ODP), a trading function has been identified. It is an important infrastructure service that allows clients to locate services using service types and associated parameters as selection criteria. The paper introduces the concept of service trading and presents a coloured Petri net model of the trader's object interactions. The model is analysed using occurrence graphs with equivalence classes which results in a significantly reduced equivalent occurrence graph.
暂无评论