Linh-Manh Pham - Roboconf: an Autonomic Platform Supporting Multi-level Fine-grained Elasticity of Complex Applications on the Cloud

Organisé par : 
Linh-Manh Pham
Intervenant : 
Linh-Manh Pham
Information détaillée : 

The Jury :

  • Prof. Françoise Baude, Université de Nice - Sophia Antipolis, Nice, rapporteur
  • Prof. Daniel Hagimont, INPT/ENSEEIHT, Toulouse, rapporteur
  • Asc. Prof. Vania Marangozova-Martin, Université Grenoble Alpes, Grenoble, examinateur
  • Asc. Prof. Alain Tchana, INPT/ENSEEIHT, Toulouse, examinateur
  • Prof. Noel de Palma, Université Grenoble Alpes, Grenoble, directeur de these
  • Prof. Didier Donsez, Université Grenoble Alpes, Grenoble, co-directeur de these
Résumé : 

Software applications are becoming more diverse and complex. With the stormy development of Cloud computing and its applications, software applications become even more complex than ever. The complex cloud applications may contain a lot of software components that require and consume a large amount of resources (hardware or other software components) distributed into multiple levels based on granularity of these resources. Moreover these software components might be located on different clouds. The software components and their required resources of a cloud application have complex relationships which some could be resolved at design time but some are required to tackle at runtime. Elasticity is one of benefits of Cloud computing, which is capability of a cloud system to adapt to workload changes by adjusting resource capacity in an autonomic manner. Hence, the available resources fit the current demand as closely as possible at each point in time. The complexity of software and heterogeneity of cloud environment become challenges that current elasticity solutions need to find appropriate answers to resolve. In this thesis, we propose a novel elasticity approach as an efficient solution which not only reflects the complexity of cloud applications but also deploy and manage them in an autonomic manner. It is called multi-level fine-grained elasticity which includes two aspects of application’s complexity: the multiple software components and the granularity of resources. The multi-level fine-grained elasticity concerns objects impacted by elasticity actions and granularity of these actions. In this thesis, we also introduce Roboconf platform, an autonomic Cloud computing system (ACCS), to install and reconfigure the complex applications as well

as support the multi-level fine-grained elasticity. To this end, Roboconf is also an autonomic elasticity manager. Thanks to this platform, we can abstract the complex cloud applications as well as automate their installation and reconfiguration that can consume up to several hundred hours of labour. We also use Roboconf to implement the algorithms of multi-level fine-grained elasticity on these applications. The conducted experiments not only indicate efficiency of the multi-level fine-grained elasticity but also validate features supporting this approach of Roboconf platform.