A Superstabilizing log(n)-Approximation Algorithm for Dynamic Steiner Trees
SSS '09 Proceedings of the 11th International Symposium on Stabilization, Safety, and Security of Distributed Systems
Regular register: an implementation in a churn prone environment
SIROCCO'09 Proceedings of the 16th international conference on Structural Information and Communication Complexity
A super-stabilizing log(n)-approximation algorithm for dynamic Steiner trees
Theoretical Computer Science
Hi-index | 0.00 |
Emerging distributed systems have a dynamic structure that is self-defined at any time by entities that autonomously decide to locally run the same distributed application. As extreme, a distributed system might thus also cease its existence when no entity is currently active while at some later moment new entities arrive and connect to each other to form again the system. Therefore the set of entities that over time might form the system is potentially infinite. This infinite arrival model is the key distinguishing factor between dynamic systems and traditional distributed systems where, on the contrary, the set of system entities is fixed since the deployment of application components is controlled and managed. In this model not only the set of correct processes is not known in advance, but no finite set containing the set of correct processes is known in advance. This actually is a higher level of uncertainty to be mastered in dynamic systems. This makes the study of possible implementations of failure detectors, as 茂戮驴, of paramount importance and at the same time makes the problem of realizing such failure detector far from being trivial. The uncertainty posed by infinite arrival models brings to two different issues (1) discovering the finite set of processes currently running and (2) dealing with a possible infinite set of non-correct processes that may wake up at any time, covering with their up times the whole computation.