Rainer Gawlick

LIVENESS IN TIMED AND UNTIMED SYSTEMS

We present a coordinated pair of general labeled transition system models for describing timed and untimed concurrent systems. Both of the models incorporate liveness properties as well as safety properties. The models are related via an embedding of the untimed model into the timed model, which preserves all the interesting attributes of the untimed model. Both models include notions of environment-freedom, which express the idea that the liveness properties can be guaranteed by the system, independently of the behavior of the environment in which it operates. These environment-freedom conditions are used to prove compositionality results for both models. This pair of models, which generalize several existing models, is intended to comprise a general formalism for the verification of timed and untimed concurrent systems.

Liveness in timed and untimed systems

When proving the correctness of algorithms in distributed systems, one generally considerssafetyconditions andlivenessconditions. The Input/Output (I/O) automaton model and its timed version have been used successfully, but have focused on safety conditions and on a restricted form of liveness called fairness. In this paper we develop a new I/O automaton model, and a new timed I/O automaton model, that permit the verification of general liveness properties on the basis of existing verification techniques. Our models include a notion ofreceptivenesswhich extends the idea ofreceptivenessof other existing formalisms, and enables the use of compositional verification techniques. The presentation includes anembeddingof the untimed model into the timed model which preserves all the interesting attributes of the untimed model. Thus, our models constitute acoordinated frameworkfor the description of concurrent and distributed systems satisfying general liveness properties.

On-line routing for permanent virtual circuits

The paper considers the problem of routing a set of permanent virtual circuit requests over a backbone network. Several factors make this routing problem complicated. Routing decisions must be made on-line without any knowledge of future request sets. Furthermore, frequent rerouting to correct inefficiencies that can result from the on-line routing decisions is not possible since rerouting creates a service disruption for the customer. Finally, the forward and reverse bandwidth of a virtual circuit must be routed over the same single path. Using an extensive set of simulations, the paper evaluates several different strategies for on-line permanent virtual circuit routing. The authors find that a strategy based on results in competitive analysis and ideas from combinatorial optimization consistently provides the best performance. The problem of admission control is closely related to the problem of routing. The paper also provides a theoretical lower bound that suggests that non-greedy admission control is a fundamental component of an efficient on-line permanent virtual circuit routing algorithm.

Designing algorithms for distributed systems with partially synchronized clocks

Much of modern systems programming involves designing algorithms for distributed systems in which the nodes have access to information about time. Time information can be used to estimate the time at which system or environment events occur, to detect process failures, to schedule the use of resources, and to synchronize activities of different system components. In this paper we propose a simple programming model that is baaed on the timed automaton model of Lynch and Vaandrager [9], which gives algorithms direct access to perfectly accurate time information. Unfortunately, this programming model is not realistic. In a realistic distributed system, clocks have skew and a finite granularity. Furthermore, other details neglected by the timed automaton model such as processor step times must also be considered. We provide two simulations that show how to transform an algorithm designed in the simple programming model to run in a more realktic distributed system. One of our simulations is an extension of previous results on the use of inaccurate clocks by Lamport [5], Neiger and Toueg [13], and Welch [17]. Our extensions suggest several powerful design techniques for algorithms that are to be run in distributed systems with clocks whose divergence from real time is bounded. We demonstrate these techniques by providing a new algorithm for distributed linearizable read-write objects. This algorithm significantly improves over previous results [1 O] in terms of time complexity and algorithmic simplicity.