Demand Segmentation

Although demand paging is considered the most efficient virtual memory system, a significant amount of hardware is required to implement it. When this hardware is lacking, less efficient means are sometimes devised to provide virtual memory. A case in point is demand segmentation.
Operating system allocates memory in segments, rather than in pages. It keeps track of these segments through segment descriptors, which include information about the segment's size, protections, and location. A process does not need to have all its segments in memory to execute. Instead, the segment descriptor contains a valid bit for each segment to indicate whether the segment is currently in memory. If the segment is in memory, the access continues unhindered. If the segment is not in memory, a trap to the operating system occurs. Operating system then swaps out a segment to secondary storage, and brings in the entire requested segment. The interrupted instruction then continues.
To determine which segment to replace in case of segment fault, operating system uses another bit in the segment descriptor called an accessed bit. It is set whenever any byte in the segment is either read or written. A queue is kept containing an entry for each segment in memory. After every time slice, the operating system places at the head of the queue any segments with a set access bit.
It then clears all access bits. In this way, the queue stays ordered with the most recently used segments at the head.
Demand segmentation requires considerable overhead. Thus, demand segmentation is not an optimal means for making best use of the resources of a computer system.