Changes between Version 5 and Version 6 of WRF4G

Sep 7, 2011 9:53:57 AM (11 years ago)



  • WRF4G

    v5 v6  
    22== Introduction ==
     3WRF for GRID (WRF4G) is a framework for the execution and monitoring of the WRF Modelling System in distributed computer resources (see this presentation). It provides a flexible and easy way of designing complex experiments involving many simulations (multiple start/end dates, multiparametric simulations, long climate runs, ...). The monitor allows a precise control of the state of the experiment, where broken simulations are automatically detected and relaunched on the next submission.
    4 The Weather Research and Forecasting (WRF) modelling system is composed of
    5 several components which need to be executed sequentially. The manual execution
    6 of this workflow is a time-consuming and error-prone task. Thus, it is customary to automate the process to
    7 some degree. However, the automation depends on the experiment to be carried
    8 out since the workflow depends on the experiment. This leads to the development
    9 of specific WRF workflow automation scripts for each experiment. But this is a time consuming task.  When the experiment requires running more
    10 than a single model run, the complexity increases and the workflow of the
    11 different runs needs to be taken into account. At this point the problems multiply: the large
    12 number of simulations now require a monitoring system to check their successful
    13 completion; failed runs need to be re-run. If failures are common, the
    14 re-running process also needs to be automated. 
     5The computer resources the user can access are managed transparently and used to run the experiment assigned to them. Remote desktop computers, workstations, clusters, etc. can be merged and used at the same time to run different simulations of a WRF experiment.
    16 '''WRF4G''' is a flexible
    17 framework to manage the WRF workflow covering a wide range of
    18 simulation experiments composed of multiple runs with different degrees of
    19 dependence. The framework is layered to separate the experiment design from the
    20 execution environment. WRF4G includes a monitoring system and easily restarts
    21 broken simulations until the experiment is completed.
    238== Documentation ==