Generic sets are good for tracking lots of tiny structures. 

They track slabs of X structures at a time. They allocate a slab of X structures with an array of X bits indicating whether a node is used or not. The structure overall has how many are used, so once full, a block can be quickly checked whether there is anything free. Then when checking a block that might have room, the availablility is checked 32 bits at a time, until a free spot is found. 

Sets of 1024 members of x,y coordinates for example are good for this sort of storage. the points are often static, once loaded they all exist until none of them do. This storage has gross deletion methods too, quickly evaporate all allocated chunks. Storing tiny chunks in a slab is more efficient because every allocation method has some sort of tracking associated with it - an overhead of having it. Plus, when operating on sets of data, a single solid slab of exatly the structures you are working with is more efficient to cache.

typedef struct genericset_tag {
  struct genericset_tag * next;
  struct genericset_tag ** me;
  _32 nUsed;
  _32 nBias;
  _32 bUsed[1];
} GENERICSET, * PGENERICSET;