Figure 7-13. Shown is the OSM script, spspinfo.sql, which shows basic information about the number and size of the shared pool subpools. In this example, three subpools exist.

Memory allocation is fairly straightforward. It follows a standard LRU algorithm combined with pinning and locking. When an Oracle process (server or background) requests memory, a portion of the Oracle's kernel called the heap manager is executed. While the details continue to change, the conceptual algorithm is pretty much the same.

Oracle processes ask for a specific amount of memory, which is transformed into multiple chunk-specific size requests. The heap manager searches for a single size chunk of memory that matches each request. Multiple chunks of memory, think noncontiguous, will not do. If the process requests 4KB of memory, the heap manager must return an address for a 4KB chunk of shared pool memory.