[KIEKER-201] Tighter integration of EMF technologies

[rju]

JIRA Issue: KIEKER-201 Tighter integration of EMF technologies Original Reporter: Jan Waller

---

Brief explanation:

Kieker's TraceAnalysis tool allows to [discover and visualize architectural information|Kieker:architecture-discovery/] from trace information collected at runtime. Internally, a system (meta-)model is used to represent views on the discovered architectures in terms of component types, component assembly, execution environment (including servers), and deployment of components to the servers. The supported visualizations (sequence diagrams, call trees, and dependency graphs) are based on this meta-model. It is implemented directly in plain Java. Shortcomings here are that a) models cannot be stored/reloaded and that b) any further extension to the model are complicated — to mention just a few.

The goal of this project is to have a tighter integration of EMF technologies in Kieker in order to alleviate these shortcomings and to support the further development of Kieker analyses and visualizations. Ideally, the plain Java meta-model would be replaced with a suitable EMF-based meta-model. Several alternatives are possible:

1. EMF-based meta-model created for the existing Java-based meta-model

• Pros:
• Modeling is quite straight forward based on the existing Java-based meta-model
• We know it satisfies our (current) requirements
• Additional requirements can be implemented into the meta-model
• Cons:
• "Yet another (component) meta-model or ADL"
• Interoperability with other tools/meta-models requires M2M transformations
  1. UML 2 (see also KIEKER-1074 Open )
• Pros:
• Existing EMF model
• Established, widely known/used
• Allows modeling of static and dynamic views
• Possible interoperability with other tools (extractors, modeling, visualization)
• Cons:
• Large specification (but: don't need to use all concepts)
  1. Knowledge Discovery Metamodel (KDM) (see also KIEKER-281 Done )
• Pros:
• Existing EMF model
• Possible interoperability with other tools (reverse engineering, knowledge discovery)
• Cons:
• Not widely known
• Focuses on static views (as far as known to us)
• Some but not so much tool support
  1. further ideas?

EMF-based models are often large and memory intensive. To prevent problems with larger analyses and to support concurrent or distributed analyses, the integration of a model repository, e.g., CDO, should be considered. Model repositories provide DBMS-like services for models in the context of model-driven engineering (MDE), including persistence, versioning, querying, partial loading and collaboration (consistence, transactions, distributed working).

Related tickets:

• KIEK-648: Feature wish: TraceAnalysis meta-model should be implemented with Ecore
• KIEK-281: KDM Extractor Filter + Repository
• KIEK-753: Feature wish: Extraction of UML2 models
• KIEK-861: Feature wish: Integrate EMF-based model repository
• KIEK-1074: Extraction/export of UML models

Expected results:

After the successful completion of this project, Kieker uses an EMF-based meta-model for its internal system model.

Knowledge prerequisite:

The student working on this topic should

• be a practiced Java developer
• have modeling skills
• have basic knowledge with the chosen (if so) existing meta-model
• have some experience with the Eclipse Modeling Framework (EMF)
• existing knowledge is really helpful, but can also be gained during the project

Skill level:

• medium to high

Additional documentation:

• Eclipse technology for model-driven engineering:
• Eclipse Modeling Project (EMP)
• Eclipse Modeling Framework (EMF)
• Model Development Tools (MDT)
• Unified Modeling Language (UML)
• Knowledge Discovery Metamodel (KDM)

Candidate Mentor(s):

The primary contact for this project idea is marked in bold font.

• André van Hoorn ([Kieker developer|Kieker:author/avh/]; researcher with University of Stuttgart)
• Holger Knoche (senior software architect with b+m Informatik AG)
• Jan Waller ([Kieker developer|Kieker:author/jwa/]; researcher with Kiel University)
• Reiner Jung (researcher with Kiel University)

Checklist:

[rju]

author rju -- Sun, 18 Feb 2018 20:08:22 +0100

This should be restructured. We have a lot of iObserve/Palladio stuff, but it should not go into the Kieker analysis. It would make more sense to pack that separately. Also Jan is no longer developing for Kieker I suppose.

[rju]

author André van Hoorn -- Thu, 25 Apr 2019 09:48:06 +0200

We use EMF here and there.