Automating deployment planning with an aspect weaver

Deployment has emerged as a major challenge in distributed real-time and embedded (DRE) systems. Application deployment planners must integrate numerous functional and non-functional constraints, such as security and performance, to produce correct deployment plans. The numerous deployment constraints and their complex interactions make manually deducing correct/efficient deployments hard. This paper presents four contributions to the study of automated deployment processes. First, it shows that a deployment planner and an aspect weaver accomplish the same abstract problem—i.e., mapping items from a source set (advice or components) to items in a target set (joinpoints or nodes) according to a set of rules—and uses this abstract definition of deployment planning to automate it with an aspect weaver. Second, this paper describes how the ScatterML domain-specific aspect language incorporates complex global constraints for specifying deployment pointcuts. Third, we show how aspect weaving can be reduced to a constraint satisfaction problem and a constraint solver used to deduce a correct weaving. Fourth, we show that phrasing weaving as a constraint satisfaction problem and automating deployment through a constraint solver based weaver yields several key benefits, ranging from guaranteed deployment plan correctness to bounds on worst case solution quality.