Software architecture of ubiquitous scientific computing environments for mobile platforms

  • Authors:
  • Tzvetan T. Drashansky;Sanjiva Weerawarana;Anupam Joshi;Ranjeewa A. Weerasinghe;Elias N. Houstis

  • Affiliations:
  • Purdue Univ., West Lafayette, IN;Purdue Univ., West Lafayette, IN;Univ. of Missouri, Columbia;Purdue Univ., West Lafayette, IN;Purdue Univ., West Lafayette, IN

  • Venue:
  • Mobile Networks and Applications - Special issue on mobile computing and system services
  • Year:
  • 1996

Quantified Score

Hi-index 0.00

Visualization

Abstract

Recent and anticipated technological advances in wireless computing will permit users to compute ubiquitously, “anywhere” and “any time”. However, mobile platforms are unlikely to have the computational resources to solve even moderately complex problems that users routinely solve on static workstations today. In the SciencePad project our aim is to develop “Ubiquitous” Problem Solving Environments (UPSEs) to support mobile aware applications. The objective of this paper is to address the architectural design of UPSEs on wireless notebook platforms supported by stationary compute servers over low bandwidth connections. There is extensive literature dealing with networking and data management issues in mobile computing. However, we address design issues at several layers of the mobile system. The resulting architecture uses a proxy-based methodology to partition the application components across the dual (wireless and stationary) network of computational units and build a distributed PSE architecture. Our architecture uses parameterized objects and templates, lazy evaluation, and also deals with fuzzy and unstructured I/O. We articulate the general principal of the architecture in the context of scientific computing, which provides us with a complex, real world application scenario. It also allows us to build on our previous work on various aspects of PSEs for scientific computing. Our architecture is currently being validated by building a ubiquitous version of PDELab (a scientific computing package) on a two mega-bit wireless Ethernet network of notebook platforms and heterogeneous parallel machines.