CoSMIC Overview

Although rapid strides in QoS-enabled component middleware technology have helped to resolve a number of Distributed Real-time and Embedded (DRE) application development challenges, a number of challenges remain unresolved. It is in this context that Model Driven Development (MDD), such as the OMG's Model Driven Architecture (MDA) and Model Integrated Computing (MIC), and Generative Programming can be effectively combined with QoS-enabled component middleware technologies to resolve these challenges. The Model-driven Generative Programming tool-chain (shown in the Figure below) we are developing is called CoSMIC (Component Synthesis using Model Integrated Computing). CoSMIC comprises a collection of Domain Specific Modeling Languages (DSMLs) and model interpreters that apply MDD technologies to QoS-enabled component middleware technologies along along these seven points of integration described in this paper and presentation.

CoSMIC Tool Chain (clickable image)

• Specification and implementation, which enables application functionality specification, partitioning, and implementation as components. CoSMIC provides IDML (Interface Definition Modeling Language}, which is a DSML that can be used to specify component definitions. IDML also provides an importer called tao_picml that can transform preexisting IDL definitions into modeling elements. Implementation modeling is not supported yet in CoSMIC but eventually we will provide behavioral modeling and component interaction modeling. IDML has now been incorporated within PICML described next.

• Component assembly and packaging, which can bundle a suite of software binary modules and metadata representing application components. CoSMIC provides the PICML (Platform Independent Component Modeling Language}, which is a DSML that models the connections between various components to form assemblies. PICML enables these assemblies to be composed into packages that can be shipped to target nodes.
• Configuration, which allows packages to be customized with the appropriate parameters that satisfy the functional and systemic requirements of applications. CoSMIC provides the OCML (Options Configuration Modeling Language} (OCML) to model the middleware configuration rules, which are then synthesized into a rules engine and graphical environment that application developers use to configure the middleware.
  
  CoSMIC also provides the QoSML (Quality of Service Modeling Language), which is a DSML that supports modeling real-time CORBA configurations. Additionally CoSMIC provides the EQAL (Event Quality Aspect Language), which is a DSML that allows modeling publisher/subscriber communication paradigm, including federations.

• Planning, which makes appropriate deployment decisions including identifying the entities, such as CPUs, of the target environment where the packages will be deployed. The MIDCESS (Model Integrated Deployment and Configuration Environment for Composable Software Systems} (MIDCESS) DSML in CoSMIC can be used to model deployment plans for DRE system components. MIDCESS is now part of the PICML DSML.

• Deployment, which triggers the installed binaries and brings the application to a ready state. CoSMIC is integrated with a run-time framework called DAnCE (Deployment and Configuration Engine), which implements the OMG's D\&C specification and can be used to model the deployment of DRE system packages according to precisely specified plans.

• Analysis and benchmarking, which enables empirical benchmarking capabilities to validate the configuration and deployment of component-based applications. CoSMIC provides BGML (Benchmark Generation Modeling Language), which models DRE systems QoS requirements and synthesizes empirical benchmarking testsuites. Additional analysis capability is achieved via integration with external tools, such as Cadena from Kansas State University.

• Assurance, which enables run-time reconfiguration and resource management to maintain end-to-end QoS. CoSMIC provides J2EEML (J2EE Modeling Language), which is a DSML that is tailored to the autonomic computing needs of EJB applications. We are working on extending J2EEML's capabilities to cover other component middleware.