The first block of the table's first extent is known as the segment header block. As with all segment header blocks, they contain very special internal Oracle structures directly related to their type of segment. For table segments, part of the header block contents is information about the location of blocks that can accept additional inserted rows. These blocks are known as free blocks. When a process must insert a row into a table, it first retrieves the table's segment header block to find a free block, it retrieves that buffer, and then it inserts the row. If there are many processes concurrently inserting rows into the same table, a table segment head block induced buffer busy waits event can result.

Figure 6-30. A diagram of the order table segment head block (4,50), which contains free lists with links to free blocks

Fortunately, the solutions are very straightforward and they work very well. If you are using manual segment space management, then segment space management is controlled by what are called free lists. As shown in Figure 6-30, I picture a free list as a jellyfish-like structure, with each of the stinging strings a list containing nodes, which point to free blocks. Oracle's free list approach usually works fine, but in high-concurrency situations, the existing free lists simply cannot handle the workload. Fortunately, we can easily alter the table to create additional free lists in another one of the segment's blocks. The result is a header block popularity reduction, and hence a reduction in buffer busy wait contention. Simply keep adding free lists until the contention subsides. The number of free lists can be found in the free lists column in the dba_segments view. If the column is empty, then we know free lists are not being used and automatic segment space management (ASSM) is being used.