MonALISA provides an easy mechanism to create pseudo clients able to use the discovery mechanism
in JINI and to find all the active services running. Such pseudo clients can subscribe to a set of
parameters or filter agents to receive selected information from all the services. This offers the
possibility to present global views from the dynamic set of services running in a distributed
environment to higher level services.
The received values are stored locally into a relational data base (mySQL or Postgres) and dedicated procedures
to compress older values may be used.
This information can also be used to create WEB repositories able to present a synthetic view of
how large distributed systems perform. A servelt engine is used to present historical and
real time values, statistics and graphical charts in a flexible way.
Basically, once started, the repository registers with
some predicates and stores the received values in the local database. A predicate has the following pattern:
Farm / Cluster / Node / start_time / end_time / function_list. These parameters (functions) are then plotted into
a large variety of graphical charts or statistics tables, following the configuration files describing the needed views, and
thus offering global or particular perspectives. The repository also uses an interactive map, for precise farm positioning
and monitoring registered sites, links, special links. Thus, it offers much of the
Interactive Client's features, adapted for
web and emphasizing on data analysis.
The same mechanism is used to offer access to this information from mobile phones using the
Wireless Access Protocol (WAP).
A schematic view of how these services are working:
Figure 1: MonALISA Repository Communication
Multiple Web/WAP repositories can easily be created using and configuring the available distributions
in order to globally describe
a group of distributed services and selected parameters/functions.