Skip to Main content Skip to Navigation

Supporting management and orchestration of cloud resources in a multi-cloud environment

Hayet Brabra 1, 2 
2 ACMES-SAMOVAR - Algorithmes, Composants, Modèles Et Services pour l'informatique répartie
SAMOVAR - Services répartis, Architectures, MOdélisation, Validation, Administration des Réseaux
Abstract : With the increased adoption of cloud computing, multiple and heterogeneous configuration and management APIs/tools and platforms have been proposed to enable end-to-end management and orchestration tasks. However, this proliferation is one of the fundamental reasons that has intensified the heterogeneity issue in multiple respects, making the cloud interoperability very difficult to achieve. With the lack of interoperability, orchestrating and managing elastic cloud resources distributed across heterogeneous providers become very complex and costly missions for the cloud adopters.Towards fostering cloud interoperability, standardization is a definitive method according to many professionals and researches from both academic and industrial sectors. In this respect, we interested in two relevant standards, namely OCCI and TOSCA because of their broad adoption and holistic approaches in addressing cloud interoperability from management and orchestration perspectives. OCCI is essentially introduced to create remote management REST APIs for supporting the management of any kind of cloud resources while preserving a high level of interoperability. On the other hand, TOSCA is introduced to empowering cloud interoperability by modeling cloud applications in a technology-independent manner. With this specification, it ultimately aims at automating the whole application orchestration process, which includes the selection, deployment, monitoring, and runtime controlling of cloud resources.Despite the growing interest in TOSCA and OCCI, their adoption is still not prevalent in modern solutions. More specifically, there is a lack of holistic orchestration engines that supports TOSCA. While there is no assistance for developers to create interoperable management APIs according to OCCI. In this thesis, we believe that any innovative design and orchestration approaches adopting such standards would be beneficial in order to cope with the heterogeneity and interoperability issues. In addition to avoiding the above issues, preserving the desired quality of service (QoS) while optimizing the involved cost has paramount importance to both cloud users and providers. Elasticity is known as a key factor to ensure this cost-QoS trade-off. However, the rapid dynamicity of cloud and its inherent heterogeneity make supporting elasticity a very tedious task.In this thesis, we target to (i) provide guidance and assistance in the design of interoperable management APIs; to (ii) streamline and improve the orchestration of cloud resources and to (iii) support high-level management of multi-cloud elasticity. To achieve the first objective, we adopt patterns and anti-patterns as means to represent respectively the good and poor practices of both OCCI and REST best principles that API developers should be carefully taken on when designing their APIs. We propose then a semantic-based approach that allows automated detection of these (anti) patterns as well as provides recommendation support to guide cloud developers in revising their management APIs. To achieve the second objective, we support the idea of integrating TOSCA with DevOps solutions as an essential step toward alleviating the complexity related to the orchestration process while maintaining a desired level of interoperability. To support this integration, we propose a model-driven approach following MDE principles. To ensure the third objective, we realize that the elasticity features should be provided at the resource description level instead of relying on low-level and technology-dependent mechanisms. So, we propose a new Cloud resource elasticity description model based on state-machine formalism.Finally, to validate our approaches, we develop three proofs of concepts and conduct a set of extensive experiments, including ones with academics and professionals to demonstrate their effectiveness and feasibility.
Complete list of metadata

Cited literature [173 references]  Display  Hide  Download
Contributor : ABES STAR :  Contact
Submitted on : Friday, June 5, 2020 - 4:28:15 PM
Last modification on : Wednesday, January 27, 2021 - 12:46:02 PM


Version validated by the jury (STAR)


  • HAL Id : tel-02798770, version 1


Hayet Brabra. Supporting management and orchestration of cloud resources in a multi-cloud environment. Networking and Internet Architecture [cs.NI]. Institut Polytechnique de Paris; Université de Sfax (Tunisie). Faculté des Sciences économiques et de gestion, 2020. English. ⟨NNT : 2020IPPAS003⟩. ⟨tel-02798770⟩



Record views


Files downloads