DIET is built upon Server Daemons. The process of scheduling the requests is distributed amongst a hierarchy of Local Agents and Master Agents. The scheduler can use resource availability information collected from three different tools: from NWS [25] sensors which are placed on every node of the hierarchy, from the application-centric performance prediction tool FAST [21], which relies on NWS information, or from CoRI Easy, which is based on simple system calls and some basic performance tests (see Chapter 8). Figure 12.2 shows the hierarchical organization of DIET.
