Keywords: component-oriented, object-oriented, parallel computing, distributed computing, server-oriented architecture, task parallel library.

This paper examines the use of .NET's Task Parallel Library (TPL) and Windows Communication Foundation (WCF) in finite element analysis. These are two relatively recent technologies that are designed to facilitate distributed and parallel computing. Any programming technology needs to be assessed for its impact in terms of:

• Speed of execution
• Program design (software engineering)
• Software capabilities

WCF and TPL are assessed in each of these categories within the context of finite element analysis, though the emphasis will be on the latter two aspects.

TPL is designed to make parallel programming easier, and adopts a task based approach. This enables the programmer to focus more on the tasks to be carried out, rather than the threads on which they are executed. The code produced is similar in execution speed to a direct thread based approach, but offers more features to facilitate the programming.

WCF is an example of Service-Oriented architecture (SOA). It is a natural progression on the object oriented, component oriented software engineering road. The key aspect of a service is that it is completely self-describing. As far as the client is concerned the service is described entirely in terms of operations and data contracts, thus minimising coupling.

WCF was used to implement equation solvers. The code was similar in terms of execution speed to the direct TPL approach. The advantage was that equation solvers could be offered as a service by a powerful computer cluster. On a single computer, a single equation solver service could be used by a variety of programs that needed equation solution as part of their operation. For very large scale problems a dedicated cluster of computers could be set up to offer the solving service.

WCF allows the logical and physical separation of the structural and finite element models in a finite element application. The client program has the physical model, i.e. it has the geometric description, applied loads, material properties and support conditions. In addition it contains information to guide mesh generation. This data is sent to the service which creates the finite element model. A distributed finite element object model is used, and this facilitates transfer of data between the client and the service.

This leads to advantages in program design, leading to reduced coupling between the structural and finite element models. Perhaps more importantly, it opens up the possibility of small applets being developed that make use of finite element analysis, the analysis being supplied by a possibly remote service.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £50 +P&P)