Web Services System (WSS)

Contents

Overview

The PRISM Web Services system enables users to perform numerical climate simulations. It allows for coupling model components such as atmosphere, ocean, biosphere and chemistry on remote computing hosts via the internet. The infrastructure allows the user to configure, submit and monitor experiments and subsequently postprocess, archive and diagnose the results of these coupled model experiments.

The figure below shows the outline of the Web Services System.

Diagram of Web Services System

In building this system a number of important sub-tasks have been involved:

  • Interfacing with climate model code.
  • Definition of the PMIOD metadata files.
  • Creation of a PMIOD configuration Graphical User Interface
  • Creation of a new experiment scheduling and monitoring tool.
  • Creation of a web security system.
  • Creation of a system installation package.

A lot of the effort has gone into the infrastructure of the system to enable the new functionality to be added and in adapting the system to interact with the climate models.

The applications served by the WSS and which make up the climate modelling infrastructure are

  • prepIFS - an interactive meteorological application to prepare research experiments using the Integrated Forecasting System (IFS) at ECMWF
  • OASIS-4 coupler GUI - for composition of the coupling configuration between climate models.
  • SMS and WebCDP - a scheduler application that enables users to run a large number of individual computing tasks which may have inter-dependencies or time dependencies. Comes with a remote monitoring and command client.
  • Web System security service - provides authentication and authorisation of users and ensures message confidentiality and integrity.

Why is the PRISM infrastructure important to you?

A model does not exist on its own, it operates in an environment of hardware and support software such as operating systems, compilers, build systems and batch queues. All these things can be called resources. In order to efficiently run a model you will then need to manage the resources, such as disc space, CPU allocation etc. This leads to modellers being resource managers instead of scientists. What we need is CONVENIENCE.

When you have convenience you can:

  • Focus on the science
  • Increase complexity
  • Stop fiddling with scripts and configurations

Convenience can be had with automation.
PrepIFS/SMS allows you to automate the configuration and running of models to a point and click state. It does this by letting you to define configurations that you know work combined with a modularised concept of running your model as groups of tasks such as getmodeldata,runmodel, postprocess and archive that are easy to monitor, control and rerun.

This organisation of the model environment leads to :
  • Better consistency and therefore reproducibility.
  • Increased flexibility. It is easy to replace/add a module or configure for another platform.
  • Less errors by removing the human factor.
  • Cloneability. It will be easy to run the same experiment again and again, with modifications such as for ensembles.

This comes at a price, standardisation. Standardisation is necessary to achieve high levels of automation. In PRISM this standardisation is defined by configurations, build and run environment, the use of SMS containers and agreed ways to exchange data between models. These standards allow modellers to add features and refinements to their software in a modular and coordinated way.

Infrastructure

So we have now moved from resource managers to just-do-it modellers, who can concentrate on the job of doing science and adding more levels of complexity and detail to their implementations. This also comes at a price, support. Someone has to look after the infrastructure. If you have 15 users (modellers) spending 1 hour every day tweaking configurations, sorting disc space, looking for log files, archiving etc you could hire 2 infrastructure support persons instead. Organisations and institutes should develop a strategy of using infrastructure support and standards to promote getting the job done.

Interactions

Once we have a standardised environment that is easy to use at home we turn our attention to the rest of the world, in other words interoperability and exchange.

PRISM supports the exchange of configurations and data between the models through its standards which can greatly ease the way scientists can share their experiments.

The next step would be to move the model instead of the configuration. Why do we want to do this?
There are scientific reasons such as running multi-model ensembles, coupling different models from different institutes but also because the IT world is changing towards commodisation of computing power.

Two factors will change the way models will run in the future:
  1. Increased bandwidth of network communication. It is now possible to achieve 400 Gbit/sec in fibre optics. This means data and models can be moved quickly around where there is computing power available at the right price.
  2. Grid technology. If you can provision your model to a virtual site and run it there you will have a lot more processing power available.

GRID technology

Grid technology separates the model from the environment of execution. The whole process of adopting a standard infrastructure such as PRISM is to be in control of the environment. So in a way, you are not only seeing the benefits of a organised, standardised environment today but also you will be preparing for the future. In this future the model in itself will not be enough, you will need the tools to manage it, the provisioning tools and a friendly environment to run it in.

Documentation

During the PRISM project requirements, design options and constraints (REDOC) have been evaluated, as detailed in the PRISM System Specification Handbook, available under the Publications section. The concepts and terminology used in this documentation have been explained in the document Design Options for the PRISM System Architecture

The Architecture Choices, Detailed Design and Implementation (ARCDI) of the WSS is further detailed in the document System architecture choice and relative role of modules.

The full documentation of the PRISM WSS (in pdf ) can be found here, under the Publications section of the site.

Presentations and Demos

The following presentations aid the understanding of the WSS and related concepts:

Title
Author
Format
Link
The PRISM Infrastructure Claes Larsson
pdf
The PRISM Security Solution Xavier Le Pasteur
pdf
An infrastructure for climate research in Europe. (from ESMF 3rd community meeting) Nils Wedi
pdf
Poster from EGU Nice, 2002 Patrice Constanza
pdf

 

Demonstration runs of prepIFS and SMS/XCDP are available below. These demos use the rfb playback applet.

Click on the link and press on the play button to start demonstration. Requires Java to be installed in your browser. The presentations take a couple of minutes and will work best if you have a T1 connection, lower speeds will take a bit longer...

  • PRISM CD (deliverable after the FP5 project which ended 1.1.2005)

 

A selection of screenshots from the Web Services System (click on image to see a larger version):
XML editing
Configuration changes window
Security system, authentication

 

[ Top ]

Site Hits: