Formal verification of replication on a distributed data space architecture

Abstract

We investigate the formal verification of safety-critical systems on top of the distributed data space architecture Splice. In Splice each component has its own local data space which can be kept small using keys, time stamps and selective overwriting. We use two complementary formal tools: first the µCRL tool set for a rapid investigation of alternatives by a limited verification with state space exploration techniques; next the most promising solutions are verified in general by means of the interactive theorem prover of PVS. These formal techniques are used to investigate transparent replication of certain components on top of Splice. We prove that a convenient solution can be obtained by means of a slight extension of the write primitive of Splice.

Semantic models of a timed distributed dataspace architecture

“…In more recent work on the verification of Splice-systems [16], we used a complex denotational semantics with environment actions.…”

“…Section 3 contains a formal syntax of a very simple Splice-like language and informally describes the meaning of this language. There is a slight difference with the semantics presented in [16,17]; the current paper contains a weak and realistic assumption about the synchronization of local clocks which simplifies the formalization significantly. The semantic difference is explained at the end of Section 3.…”

Verification of Distributed Dataspace Architectures

“…We use the new language to specify two small existing applications, studied in [21] and [18], respectively. The goal of these examples is to show that our language is very simple to use and to illustrate the typical kind of problems that space calculus is meant for: transparent distribution of data and transparent replication of applications.…”

“…This producer-transformer-consumer example illustrates a specific pattern in Splice systems. The transparent replication of the middle component was extensively studied in [18], using both μCRL and PVS. We show how to model the problem in space calculus ( Figure 6(b)), for the specific instance when two data items are produced, with values 0 and 1.…”

“…An answer to the last question would facilitate early architectural design decisions. Some of these questions have been answered for Splice already [11,18,20].…”