Optimal control techniques for automated resource reconfigurations in Cloud Computing Environments / Clouds
In my research, I focus on the problem of optimising the management of virtual resources in Cloud Computing environments under multiple resource constraints. More specifically, SLA violations in cloud environments will be autonomically detected and the system configuration will dynamically adapt accordingly, in two directions:
1. Manage the VM provisioning problem: This stage will be responsible for allocating resource capacity in the form of virtual machines to applications. It is driven by performance goals associated with the business-level SLAs of the hosted applications (e.g., response time, jobs completed per time unit). At the end of this phase information will be provided regarding which VMs shall be destroyed, created or resized.
2. Manage the VM placement problem: This stage will manage the mapping of virtual machines to physical nodes and is driven by data center policies to minimize resource management costs. At the end of this phase information will be provided regarding the placement of newly created VMs and possible VM migrations. The degrees of freedom of the problem include the number of VMs/Physical Nodes/services, the physical location of the VMs, network parameters as I/O performance, applications awareness and affinity, heterogeneity of applications and topologies. Towards handling the decision making process for selecting the optimal configuration plan, I investigate the possibility of using metaheuristics for the above multi-objective optimization problem, such as evolutionary algorithms and Constraint Satisfaction Problems. This problem contains many conflicting objectives such as balancing the load among the various virtual machines, minimising the overall operational cost, maximising the throughput, minimising the execution time. The aforementioned policy complies with the elasticity rules of clouds eliminating the need for dedicated resources as infrastructure will scale quickly, shrink dynamically and seamlessly handle violations.
Main research interests:
- Cloud Computing
- Optimal Control Theory in network and application layers
- Control Engineering
- Distributed Computing
- Service Composition
- Service Oriented Architectures (SOA), QoS in SOA
Chatziprimou K., Lano K., Zschaler S. : Runtime Infrastructure Optimisations in Cloud IaaS Structures. In Proc. of the IEEE CloudCom 2013, December 2-5, Bristol UK, 2013, p. 687-692, – Best Student Paper Award, LINK
Chatziprimou K., Lano K., Zschaler S. : Towards A Meta-Model of the Cloud Computing Resource Landscape. In Proceedings of 1st International Conference on Model-Driven Engineering and Software Development (MODELSWARD), Feb. 19-21, Barcelona, Spain, 2013, p. 111-116.
Chatziprimou K., Lano K., Zschaler S.: Resources Provisioning in the Cloud. In Proceedings of the First International Conference on Information and Communication Technologies for Sustainability (ICT4S), Feb. 14-16, ETH Zurich, Switzerland, 2013 (extended abstract).