Research Associate: S. Markus
Sponsor: Intel
Enabled by advances in hardware, networking infrastructure and algorithms, highly compute intensive problems in many areas can now be successfully attacked using networked scientific computing. In the network-based paradigm, vital pieces of software and information used by a computing process are spread across the network, and are identified and linked together only at run time. This paradigm is in contrast to the current software usage model where one purchases a copy (or copies) of a general-purpose, monolithic software package for use on local hosts, possibly distributed on a collection of local hosts. With network-accessible software repositories and networked computing, the view of software changes from a product to a service. A network-accessible repository provides access to up-to-date copies of software components on an as-needed basis, so called "disposable software."
The Net //ELLPACK project is a Web-based network computing system which allows users to access and use the problem solving power of the //ELLPACK problem solving environment. Building on our earlier Web //ELLPACK system, Net //ELLPACK is a next generation Web computing environment which uses //ELLPACK's back-end but completely replaces the user interface environment with a Java applet based on a different usage metaphor. Net //ELLPACK eliminates most of the limitations of Web //ELLPACK and allows users to solve complex PDE problems using any Web browser that supports Java applets. It uses a Java applet that serves as the graphical user interface, a stateful text-based protocol (NetPP) and a Net //ELLPACK server (NetPPD) running on some machine or machines anywhere on the network. In addition to the its native protocol, the NetPPD also supports HTTP. This allows the use of existing public domain or proprietary applets to visualize or postprocess the computation results.
Using this prototype, we are addressing several research issues in network computing including user interface and protocol design, application security, software delivery in directly usable form, and network computational servers. In on-going work with the next generation of networked PSEs, we are working on solutions that use CORBA and IIOP as well as on enabling parallel computing on the network.
Information regarding this project is available at
http://pellpack.cs.purdue.edu/netpp/.