Elastohydrodynamic Lubrication

The Grid

What is the Grid?

At its most basic the Grid is simply computers networked together. The idea also has notions of transparent access to the resources. The common example used is that of the national power grid - when you plug a light into a mains socket you don't care which power station produced it. Siumilarly, when running a software application you usually just want the answer as quickly (and economically) as possible.

Pertaining to be more than just "a collection of computers on a network" the Grid is more a concept for how to manage these large remote resources seemlessly. Clearly if you are using resources scattered around the world then there must be a sensible way of deciding which one to use. There must then also be some permission granted to use that machine by the owner. There must be methods of getting the code to the machine and results back to the person launching the application, and these transfers must be secure against eavesdroppers. This is all termed Grid middleware.

Why do EHL on the Grid

In the section on parallel EHL it was outlined why using large supercomputers would be beneficial for individual EHL problems. If these large machines are put on the Grid (as is expected) then it is necessary to interface with them. Another reason for Grid use is that commercially companies, such as Shell, will want to run many hundreds of EHL calculations at once. Being able to spread these out over a network of processors, be they in a single computer or scattered around the internet, would give results much quicker improving productivity.

Grid based PSEs

Using IRIS Explorer as a PSE means that we can take advantage of the work in the COVISA-g project. This turned a simple finite volume code modelling a pollution plume into a interactively steerable Grid based application. This is described in the paper of Walkley, Wood, and Brodlie. The resulting EHL PSE is described fully in Goodyer, Wood and Berzins.

In brief, the system works by a module n IRIS Explorer using Globus to launch the job onto the Grid. This Grid application may be parallel, and one of the processes will be in contact with the launching IRIS Explorer process via sockets. Information can flow from the input modules in the PSE to the launching module where it will pass the current inputs onto the Grid process whenever the Grid process asks for them. The application will also use the same socket to pass pre-formatted output data back to the IRIS Explorer module to be formatted fr visualisation further down the map.

Since IRIS Explorer is being used it is possible to make this Grid session collaborative by sharing the input modules. As before, only one participant needs to launch the job, but the others can help steer the computation.

Introduction

Variable Timestepping

Mesh Adaptation

PSEs

Parallelism

The Grid

Optimisation