Transactions and Zero-Safe Nets

When employing Petri nets to model distributed systems, one must be aware that the basic activities of each component can vary in duration and can involve smaller internal activities, i.e., that transitions are conceptually refined into transactions. We present an approach to the modeling of transactions based on zero-safe nets. They extend ordinary pt nets with a simple mechanism for transition synchronization. We show that the net theory developed under the two most widely adopted semantic interpretations (collective token and individual token philosophies) can be uniformly adapted to zero-safe nets. In particular, we show that each zero-safe net has two associated pt nets that represent the abstract counterparts of the modeled system according to these two philosophies. We show several applications of the framework, a distributed interpreter for zs nets based on classical net unfolding (here extended with a commit rule) and discuss some extensions to other net flavours to show that the concept of zero place provides a unifying notion of transaction for several different kinds of Petri nets.