1. Data Protection Regulation (Synthetic Test Cases) 2. Trustworthy cloud for Industry 4.0 (Self-protection cloud applications) 3. Privacy awareness and security of cloud applications (privacy policies) 6. EU-Brazil common standards (OASIS TOSCA extension - models and policies - for Automated Deployment and Management of Self-* Applications on Hybrid Clouds) 7. Cloud sustained cyber-physical systems and systems of systems (Self-* applications deployed on hybrid clouds) 8. Cloud federation at cross-regional level (portable DSL for Automated Deployment and Management of Self-* Applications on Hybrid Clouds)
According to the usual distinction between cloud infrastructure providers, cloud application providers, and end users, the direct targets of this research initiative and its foreground technologies and standards are the developers and providers of cloud applications. These cloud infrastructure users face several pains related to difficulties and obstacles associated with: (i) the fulfillment of quality of service and security requirements, (ii) the lack of cloud interoperability, (iii) the lack of portability of cloud applications, (iv) the implementation of agile development cycles and continuous integration, delivery, and deployment processes, and (v) the efficient operation management of cloud applications.
DevOps is an emerging paradigm of integration of the development cycle with the production stage that boosts Agile approaches. To adopt DevOps effectively, cloud application developers have yet to find solutions to tough problems: i) how to design, deploy, and operate efficiently applications that fulfill both stringent scalability, availability and serviceability needs and severe security and data protection obligations; ii) how to cope with the lack of interoperability of cloud infrastructures and the lack of portability of cloud applications; iii) how to harmonize continuous integration, delivery, and deployment, and mission-critical quality assurance and test.
This position paper outlines an intercontinental research whose goal is twofold: (i) low code implementation of portable cloud applications with self-configuration, self-healing, self-recovery capabilities; (ii) low code automation of installation, configuration, and setup of self-* applications on hybrid clouds.
We plan to achieve this goal by researching, designing and implementing a solution based on three correlated elements:
- (i) a declarative, cloud agnostic, and extensible Domain Specific Language (DSL) for structural and behavioral modeling and policy definition;
- (ii) a generic, model-based, and policy-driven autonomic ops manager, enabling self-configuration, self-healing, and self-recovery of the managed application;
- (iii) a DSL workbench PaaS, equipped with editors, wizards, consoles, and dashboards for automated deployment and supervision of self-* applications on hybrid clouds.
The OASIS TOSCA declarative language let model the distributed application topology independently from the particular target cloud infrastructure. Its implementations allow the cloud-agnostic automation of the installation, configuration, and startup. We plan to design and implement our DSL as an extension of the TOSCA standard with traits that enable modeling: (i) application-level topology as a graph of logical components connected by service dependency wires; (ii) security and data protection provisions, such as standard authentication, confidentiality, and privacy; (iii) self-healing policies that drive runtime test automation - non-intrusive logging, non-intrusive passive testing, active testing (both non-intrusive and intrusive), all compliant with the data protection regulations; (iv) self-configuration policies - (a) circuit breaking, bulkheads, timeout management; (b) installation, configuration and setup of versioned builds (upgrading and backtracking) without service interruption; (v) self-recovery policies of failover and masking of transient network failures; Our autonomic architecture combines the managed application as-is with an ops manager composed by a supervisor and a collection of concurrent feedback loops (MAPE-K model). There are no interferences between the managed application and the Ops manager, except for those performed by the adaptation processes. Significant concerns for the supervisor are the stability, accuracy, short settling time, robustness, termination (no deadlock), consistency, scalability, and security of the adaptation processes. Advanced issues are coordination between autonomic ops managers and runtime policy change. The DSL Workbench as a service allows drafting, checking, testing, and validating deployment models and self-* policies, and enables users to install, configure, and setup self-* applications on hybrid clouds. We have identified real-world use cases for trials in the logistics, financial services, and high-tech industries.