It is widely recognized that distributed systems would greatly benefit from the availability of a generic failure detection service. There are however several issues that must be addressed before such a service can actually be implemented. Traditionally, failure detectors or failure detection services provide a list of processes that are currently suspected by them. Mechanisms for propagating such information are implemented mostly for such traditional failure detectors. Recently, a family of failure detectors that provide the degree of confidence that a given process has actually crashed, called suspicion level. It is called the Φ failure detector which is an implementation of the notion of accrual failure detectors. In this paper, we highlight the issue on the propagation mechanism of information on crashed/suspected processes with the Φ failure detector. Since the suspicion level is represented as a continuous value, existing mechanisms are not appropriate for this type of failure detectors. Therefore, we propose a notification system that can efficiently propagate suspicion levels. It can provide such information to proper receivers and processes in distributed applications do not need to implement a function for failure detection by using the proposed system and the Φ failure detector.
展开▼
