Grid Middleware Basic Methodology

Grid middleware is built by layered interacting packages and may be taylored using different managers called by a common API so that the users are not concerned of the different syntax and access methods of specific packages. The most common example is given by the job scheduler that can be any of a more or less complex set of products like simple Unix fork, Condor, LSF, PBS, etc.

Different tasks are performed by existing software like MRTG for performance monitoring or MySQL for bookeeping. Summarizing the basic middleware methodology is a function of the following items:

• the grid is a net bound system that needs efficient and reliable communication and high performance browsers with high level security performance

• the grid is database bound at least for the following functions: grid topology management user access and certification dataset locations and replicas resource definition and dynamical management performance and user bookeeping

• the grid is bound to efficient matching algorithms for: user-certificate match job-resource match master data objects, subset and replicas identification.

• the grid depends on accurate clock performances to synchronize nodes and correctly handle task and job scheduling