Coteries are general schemes for distributed mutual exclusion solutions. Garcia-Molina and Barbara developed a transformation algorithm to transform one nondominated (ND) coterie to another [5]. The algorithm may properly be used to partially enumerate ND coteries. However, the transformation algorithm proposed is not appropriate for the case of large coterie design because properties of the new produced ND coterie are not easily predictable. We introduce a hierarchical transformation concept to provide a simple way to produce new larger ND coteries through transformation from smaller ND coteries. By systematically applying our transformation, coteries can scale well for a large number of nodes. Furthermore, the new generated coteries will inherit the properties of original coteries. This makes the transformation more predictable than trial and error. We also devise a heuristic algorithm to construct hierarchical structures aimed at maximizing availability in the context of heterogeneous systems.
Read full abstract