Concurrency in Distributed Systems under Autonomous and Enforced Actions

Abstract

Recently, the formal approach of I-systems has been newly established through an abstract axiomatic system where events in system components are solely derived and defined from their bilateral interaction with other components as well as from the autonomy as found in decentralized systems. As a major difference to other concurrency models we do not assume that simultaneity or coincidence of events are observable in distributed systems. (This is a fundamental practical problem in autonomous decentralized systems.) We respect this by defining concurrency of events as a system-wide or global concept of causal independence which will be given and characterized by bilateral interaction properties about the components involved, detectable through local checks only. A novel concept of conflicting actions will be given as well. While not complementary to concurrency (as often understood in the area of Petri nets) concurrency and conflict will be complemented to constitute fundamental relationships between distributed events, through a new relation termed 'event funneling'. Practice-related issues related to this fine-grained event structure will be discussed