A Domain Specific Modeling Language Supporting Specification, Simulation and Execution of Dynamic Adaptive Systems

Abstract

Abstract. Constructing and executing distributed systems that can automatically adapt to the dynamic changes of the environment are highly complex tasks. Non-trivial challenges include provisioning of efficient design time and run time representations, system validation to ensure safe adaptation of interdependent components, and scalable solutions to cope with the possible combinatorial explosions of adaptive system artifacts such as configurations, variant dependencies and adaptation rules. These are all challenges where current approaches offer only partial solutions. Furthermore, in current approaches the adaptation logic is typically specified at the code level, tightly coupled with the main system functionality, making it hard to control and maintain. This paper presents a domain specific modeling language (DSML) allowing specification of the adaptation logic at the model level, and separation of the adaptation logic from the main system functionality. It supports model-checking and designtime simulation for early validation of adaptation policies. The model level specifications are used to generate the adaptation logic. The DSML also provides indirection mechanisms to cope with combinatorial explosions of adaptive system artifacts. The proposed approach has been implemented and validated through case studies.

Concern-oriented language development (COLD): Fostering reuse in language engineering

“…Limited to the composition functionality, the language concern usage interface -in Neverlang-is realized by the provides clause. provides and requires clauses are also used by the AiDE [8] tool to automatically build a configuration tree (a CVL [55] ) out of the available slices making explicit the existing variability. In the COLD parlance, the configuration tree built by AiDE realizes the language concern variability interface .…”

Mechanisms for Leveraging Models at Runtime in Self-adaptive Software

“…Different techniques for reasoning and planning have been proposed and according to Fleurey and Solberg [38], they can be generally classified into two types of adaptation models.…”

“…An event triggers the adaptation process and conditions determine which reconfiguration action should be performed. According to Fleurey and Solberg [38], such approaches can be efficiently implemented with respect to runtime performance, and they can be simulated and verified early in the development process. However, if the number of rules grows, the approach suffers from scalability issues concerning the management and validation of the rules.…”

“…In general, search-based approaches solve the scalability problem of rule-based approaches, but they suffer from costly reasoning and planning processes, and weaker support for validation (cf. [38]). Since these processes have to be carried out at runtime, the runtime performance is crucial for any reasoning and planning mechanism.…”

“…Based on the different characteristics of rule-based and search-based approaches, Fleurey and Solberg [38] propose a mixture of them to balance their advantages and disadvantages. Their general idea is to use rules to reduce the variability space of the system and environment that subsequently has to be searched for suitable configurations.…”