No. 130 (00307)

Family name : Legrand
Given name : Iosif 
Affiliation : California Institute of Technology 
Abbreviation : CALTECH
E-mail address : Iosif.Legrand@cern.ch 
Title : Using a Mobile Agent Architecture to Monitor, Control and
        Optimize the Operation of Distributed Systems 
Authors : I.C.Legrand, H.B. Newman,  P. Galvez,  C. Cirstoiu , R. Voicu
Abstract : 
The MonaLISA (Monitoring Agents in A Large Integrated Services
Architecture) system provides a distributed monitoring service.
MonALISA is based on a scalable Dynamic Distributed Services
Architecture implemented using JAVA / JINI and Web Services
technologies.   The scalability of the system derives from the use of
multithreaded Station Servers to host a variety of loosely coupled
self-describing dynamic services or agents and  the ability of each
service to register itself and then to be discovered and used by any
other services, or clients that require such information.

In this paper we described the development of specialized mobile agents
in the MonALISA framework to optimize data replication strategies for
data processing in GRID systems as well as to help and improve the
operation of the VRVS system, a complex videoconferencing distributed
application.   These agents are examples of distributed higher level
services which are using local monitoring information in performing a
global optimization task.  The agents are deployed to all the active
MonALISA services and perform supervision tasks for distributed
applications.  They act using a predefined heuristic to restart or to
reconfigure the applications when certain events are detected. 

Efficient data replication between distributed GRID sites requires to
dynamically creating a minimum path tree in the connectivity graph
defined by all local measurements of network quality and estimated
throughput.  The mechanism of dynamically reconnecting a large set of
distributed VRVS reflectors as a function of their availability and the
network conditions is also presented.  This is realized in nearly
real-time by many collaborating agents which dynamically create a global
minimum spanning tree and in this way optimize the interconnections
among the reflectors of the VRVS system.