Safe compositional network sketches

Abstract

NetSketch is a tool for the specification of constrained-flow networks (CFNs) and the certification of desirable safety properties imposed thereon, conceived to assist system integrators in modeling and design. It provides compositional analysis capabilities based on a strongly-typed domain-specific language (DSL) for describing and reasoning about CFNs and relevant invariants. Users can model or design individual network components and perform manual or automated whole-system analysis of the properties thereof. Users can also assemble many instances of these components into larger networks, relying on NetSketch's less precise but more tractable compositional analysis capabilities. This ability to trade "precision of analysis" for "feasibility of analysis" according to available resources is among the novel features of NetSketch. In earlier work we illustrated how NetSketch is applied to actual domains [6], and provided a formal definition of its underlying formalism [7].While the NetSketch DSL provides automatic compositional analysis capabilities for modeling and designing entire networks, users may need to employ a wider variety of tools and techniques when modeling and designing individual network components. These can include common tools for reasoning about systems of constraints of various classes (such as linear constraints, quadratic constraints, and so on), as well as logical systems and ontologies that deal with concepts relevant to the application domain. We integrate the AARTIFACT [14] lightweight automated assistant for formal reasoning (which has also been applied in proving the soundness of the NetSketch formalism [15]) as a tool for modeling and designing individual network components. We present use cases within the context of an example application of the NetSketch DSL that demonstrate how the automated assistant provides NetSketch users with both an interface for reasoning formally about constraints, and a straightforward way to implicitly employ a rich domain-specific ontology of logical propositions. This allows users to verify common properties of constraints and constraint sets, and to reason about constraint relationships using automatically verifiable algebraic manipulations.

A Type-Theoretic Framework for Efficient and Safe Colocation of Periodic Real-Time Systems

“…The work presented in this paper underscores a number of thrusts of our broader research agenda, which includes: leveraging colocation for efficient management of virtual machine (VM) cloud infrastructures [28], using type-theoretic formulations in the modeling and analysis of network compositions [29], [30], and the development of automated and semi-automated tools based on light-weight formalisms [18]. With respect to the specific framework presented in this paper, our immediate future plans include enriching our SLA type hierarchy through the introduction of additional semantics (e.g., the consideration of other real-time schedulers), and the codification of prior results in the vast real-time scheduling literature into type transformations that could be used by our mapping service.…”

The syntax and semantics of a domain-specific language for flow-network design

“…The expansions of let-bindings in the proofs of Lemma 12 and Theorem 13 do not yet define a reduction (or rewrite) system for network specifications. Strictly speaking, these proofs only establish the existence of normal-form specifications, without explicitly specifying reduction rules.…”

“…The work reported hereinafter is a fraction of a collective effort involving several people, under the umbrella of the iBench Initiative at Boston University, co-directed by Azer Bestavros and the author. 2 An earlier version of the DSL in this paper, with its formal semantics and type system, was introduced in our work for NetSketch, an integrated environment for the modeling, design and analysis of large-scale safety-critical systems with interchangeable parts [13,14,15]. In addition to its DSL, NetSketch has two other components currently under development: an automated verifier (AV), and a user interface (UI) that combines the DSL and the AV and adds appropriate tools for convenient interactive operation.…”

User-friendly Support for Common Mathematical Concepts in a Lightweight Verifier

“…Figure 6 corresponds to a linear algebra homework assignment completed by students (as part of the deployment discussed further in Section 3.3 below). Figure 7 represents a very large formal argument: the proof of soundness of the NetSketch formalism [6,19]. The sizes of the components shrink at certain points because premises can fall out of scope during the validation process.…”

“…We have utilized [19] the aartifact system in defining and reason about a novel compositional formalism underlying a typed domain-specific language [6]. This formalism can be used to model × |Q| * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * Figure 6: R is a vector space ⇒ R 2 is a vector space and assemble networks, and to reason about and analyze constraints on flows through these networks.…”