Grid computing becomes more and more popular in integrating distributed heterogeneous resources. Despite all that, efficient
monitoring to resources of different ownership in Grids remains a challenge. In this paper a novel resource-autonomy based
grid monitoring architecture named PIMISA (Plug-In Monitoring and Information Service Architecture) is proposed to address
this problem. GMA and SOA are subtly explored in the “Two-Level-Logic” philosophy of PIMISA: only measurement and control
metadata of resources are globally published, while monitoring data communication between elements of different ownerships
are handled by means of service invocation. The monitoring within the scope of an ownership follows a close-loop structure:
a dynamic configurable set of management logic is kept, and resource states are analyzed at real time so that corresponding
control indications could be made by decision-making. Moreover, above procedure can be achieved by a “Third-Party Mount” mode
to leverage caused load on resources proportion to their capabilities. A prototype is currently under development, and therefore,
the implementation scheme of PIMISA is presented in mathematical format.
This Work is supported by Natural Science Foundation of China under Grant 60573110, 60373005, 90412011, and 973 Project under
Grant 2003CB3169007.