Siphon-based deadlock detection in Integrated Model of Distributed Systems (IMDS)

Abstract

Integrated Model of Distributed Systems (IMDS) is a formalism for specification and verification of distributed systems, especially following IoT (Internet of Things) paradigm. The formalism emphasizes such features as asynchrony of actions and communication, locality of decisions, and autonomy in executing actions. In conjunction with model checking, IMDS allows to analyze such features of distributed systems as deadlocks or distributed termination. However, the nature of model checking allows to find one deadlock in a single run of the verifier, which produces a counterexample.The conversion of IMDS specification to a Petri net is used to identify multiple deadlocks in one verification, using siphons. Model checking is used to verify if a siphon can become empty, which denotes a true deadlock in a purely cyclic system, like FMS (Flexible Manufacturing Systems). The extension of the verification by temporal checking allows to cover systems with any structure: cyclic, terminating, or with a more complex scheme. In addition, the proposed procedure allows to easily identify processes participating in partial deadlocks. Two types of deadlock can be identified: communication deadlocks and resource deadlocks. I. INTRODUCTION MDS (Integrated Model of Distributed Systems [1][2]) is a formalism for describing the behavior of distributed systems, especially for finding deadlocks. A system is modeled as a set of actions, having servers' states and agents' messages on input and on output. In IMDS, a communication dualism is exploited, since the modeled system is represented as server processes that communicate by messages, or alternatively by travelling processes (agents) that communicate by means of servers' states. A model of a distributed system is uniform (that is, it has a single form), but it can be decomposed ("cut") to a set of server processes or a set of agent processes. System actions are combined in sequences to form the processes. An action has a current server's state and an agent's current message on input, and it produces a similar pair (a new server's state and a new agent's message) on output.The two views of a system (server view and agent view) are obtained by the two possible groupings of a set of all actions into sequences. In the server view, actions of an indi-This work was not supported by any organization vidual server are grouped into a process (the definition of processes is included in Sect. IIIB). The server's states are the carrier of the server process, and the messages are the communication means between server processes. In the agent view, actions concerning an individual agent conform a process. Messages are internal to a process: they are the carrier of the process. The agent processes communicate via servers' states.

Modeling and Verification of Asynchronous Systems Using Timed Integrated Model of Distributed Systems

“…. -lines 21,22), and an initial configuration phrase (init → lines 24-29). A node type heading contains a set of formal parameters, such as the agents and nodes employed in the node type's activities.…”

“…Repeaters may precede actions in a node type definition for its compactness (for example, in an AVGS system Section 7.2, 2 repeaters are used for some actions). Vectors can be used to arrange node and agent instances (lines 21,22). The channels phrase (line 23) specifies channel delays, in this example, the channels have no delay.…”

“…While automata-based models, as well as IMDS, can be transformed to Petri nets [21], they are not so attractive in our opinion, because it is not easy to extract the processes from the specification. We use Petri nets for the structural analysis of a verified system and some kinds of model checking [22]. The temporal logic timed CTL (TCTL) is associated with TA [18,23].…”

Modelling Distributed Systems in Distributed Autonomous and Asynchronous Automata (DA3)

“…Also, some formal parameters have the form of vectors (l.2, 19, ...). Services and states may also be vectors (l. 4,23). For a compact definition, repeaters precede the actions in a server type (lines 10-17, ...).…”

Specification and Verification in Integrated Model of Distributed Systems (IMDS)

“…In the Dedan program, several new verification facilities were developed, including a converter to Petri nets for structural analysis and siphon-based deadlock detection. This procedure allows for finding multiple deadlocks in single verification run [57,[67][68][69][70]. Asynchronous and Autonomous Distributed Automata (DA 3 [57]), in two forms of server automata and agent automata, are designed for simulation of distributed systems over its elements, and to simulate the counterexamples.…”