Maintaining coherency of dynamic data in cooperating repositories

In this paper, we consider techniques for disseminating dynamic data--such as stock prices and real-time weather information--from sources to a set of repositories. We focus on the problem of maintaining coherency of dynamic data items in a network of cooperating repositories. We show that cooperation among repositories-- where each repository pushes updates of data items to other repositories--helps reduce system-wide communication and computation overheads for coherency maintenance. However, contrary to intuition, we also show that increasing the degree of cooperation beyond a certain point can, in fact, be detrimental to the goal of maintaining coherency at low communication and computational overheads. We present techniques (i) to derive the "optimal" degree of cooperation among repositories, (ii) to construct an efficient dissemination tree for propagating changes from sources to cooperating repositories, and (iii) to determine when to push an update from one repository to another for coherency maintenance. We evaluate the efficacy of our techniques using real-world traces of dynamically changing data items (specifically, stock prices) and show that careful dissemination of updates through a network of cooperating repositories can substantially lower the cost of coherency maintenance.