The architectures analyzing for computational workflows designing in a distributed environment

The paper presents a model of computational workflows based on end-user understanding and provides an overview of various computational architectures, such as computing cluster, Grid, Cloud Computing, and SOA, for building workflows in a distributed environment. A comparative analysis of the capabilities of the architectures for the implementation of computational workflows have been shown that the workflows should be implemented based on SOA, since it meets all the requirements for the basic infrastructure and provides a high degree of compute nodes distribution, as well as their migration and integration with other systems in a heterogeneous environment. The Cloud Computing architecture using may be efficient when building a basic information infrastructure for the organization of distributed high-performance computing, since it supports the general and coordinated usage of dynamically allocated distributed resources, allows in geographically dispersed data centers to create and virtualize high-performance computing systems that are able to independently support the necessary QoS level and, if necessary, to use the Software as a Service (SaaS) model for end-users. The advantages of the Cloud Computing architecture do not allow the end user to realize business processes design automatically, designing them "on the fly". At the same time, there is the obvious need to create semantically oriented computing workflows based on a service-oriented architecture using a microservices approach, ontologies and metadata structures, which will allow to create workflows “on the fly” in accordance with the current request requirements.