Modelling Regional Centers in the 'LARGE'
----------------------------------------

        The Regional Center profile defines the key features of a Regional 
Center:

        1) the scale of computing resources -- CPU, disk, and tape 

        2) the data resident at the center

        3) the services provided

        4) the connectivity to CERN and to institutions served by the
           center

With this information, it is possible to construct a reasonably detailed 
'initial' view of the actual operation of such a center and its interaction 
with its constituents and with CERN.

         The model will need to answer questions such as:

           Between CERN and the RC:
1) How much data will be moved from CERN to the RC per day to maintain the 
   supported datasets -- e.g. AODs -- at the RCs.
2) How much data will be moved from CERN (or otehr RCs) to support the
   ad hoc requests of users for data not permanently resident at teh RC?
3) How much data will be moved from the RC to CERN per day? What are the 
   sources and types of data moving in this direction? For example, are these
   post-reconstruction tasks, object recreation,...?
4) What caching and mirroring strategies will be used by the RCs for
   datasets which are frequently used but perhaps not enough so to
   warrant permanent storage on disk?
5) What is the correct balance between disk and tape storage (robotic) given
   various scenarios for technology evolution? Is shelf archiving likely
   to be required? under what circumatances is it acceptable?
6) How will the central code repository and key support databases such
   as the calibration database, the production database, and runtime conditions
   database be made avaiable to the RC and how will it be kept current? How 
   big are these databses and how often do they change?
  

          Between the user and the RC:
1) What kind of computing will the typical university user have on his desktop
   in 2005? What kind of 'workgroup' systems will be available at universities
   and what will their data storage capacities be? What kind of network
   throughput will be avaialbe between the university and the Regional Center?
   How many people will be using the system simultaneously?
2) How will the university users interact withthe Regional Center? What kinds
   of jobs will they run at the center? How many of the various kinds of jobs 
   will be running concurrently? How much and what kinds of data
   will be moved from the center the remote users? How much data will
   be 'published back' from remote users to the RC?
3) How will the central code repository and key support databases (see above)
   be made avaialble to the user? What response is required? How much data 
   is this? How can it be kept current?


        With answers to these and other questions, it should be possible to 
construct a high level model of the activities of the Regional Center. Various
strategies for carrying out the analysis task based on projections of network
capabilities, cost of disk and tape etc can be developed and modelled.