Section: New Results

Cloud applications and infrastructures

Participants : Frederico Alvares, Gustavo Bervian Brand, Yousri Kouki, Adrien Lèbre, Thomas Ledoux, Guillaume Le Louët, Jean-Marc Menaud, Jonathan Pastor, Rémy Pottier, Flavien Quesnel, Mario Südholt.

We have contributed on SLA management for Cloud elasticity, fully distributed and autonomous virtual machine scheduling, and energy-efficient Cloud infrastructures.

SLA Management for Cloud elasticity

In [23] , we have introduced CSLA, the Cloud Service Level Agreement language. The CSLA language has been influenced by related work, in particular WSLA and SLA@SOI. It allows to describe the SLA between a cloud service provider and a cloud customer. One of the novelties of CSLA is that it integrates features dealing with QoS uncertainty and cloud fluctuations, such as confidence, penalty and fuzziness.

Cloud computing is a model for enabling on-demand access to a shared pool of configurable resources as services. However, the management of such elastic resources is a complex issue. In [24] , we have proposed a SLA-driven approach for optimizing the resources capacity planning for Cloud applications. We have modeled Cloud services using a closed queuing network model and proposed an extension of a Mean Value Analysis (MVA) algorithm to take into account the concept of SLA. Then, based on capacity planning method, our solution calculates the optimal configuration of a Cloud application.

Fully distributed and autonomous virtualized environments

Extending previous preliminary results of the DVMS prototype, we have consolidated this system to obtain a fully distributed virtual machine scheduler [13] . This system makes it possible to schedule VMs cooperatively and dynamically in large scale distributed systems. Simulations (up to 64K VMs) and real experiments both conducted on the Grid'5000 large-scale distributed system [34] showed that DVMS is scalable. This building block is a first element of a more complete cloud OS, entitled DISCOVERY (DIStributed and COoperative mechanisms to manage Virtual EnviRonments autonomicallY)  [66] . The ultimate goal of this system is to overcome the main limitations of the traditional server-centric solutions. The system, currently under investigation in the context of the Jonathan Pastor's PhD, relies on a peer-to-peer model where each agent can efficiently deploy, dynamically schedule and periodically checkpoint the virtual environments it manages.

Energy-efficient Cloud applications and infrastructures

As a direct consequence of the increasing popularity of Cloud Computing solutions, data centers are amazingly growing and hence have to urgently face with the energy consumption issue. Available solutions rely on Cloud Computing models and virtualization techniques to scale up/down application based on their performance metrics. Although those proposals can reduce the energy footprint of applications and by transitivity of cloud infrastructures, they do not consider the internal characteristics of applications to finely define a trade-off between applications Quality of Service and energy footprint. We have proposed a self-adaptation approach that considers both application internals and system to reduce the energy footprint in cloud infrastructure [31] , [11] . Each application and the infrastructure are equipped with their own control loop, which allows them to autonomously optimize their executions. In addition, these autonomic loops are coordinated to avoid inconsistent states. This coordination improves the synergy between applications and infrastructure in order to optimize the infrastructure energy consumption [16] .

We have extended our previous work on Entropy, a virtual machine placement manager, by the development of btrScript, a safe autonomic system for virtual machine management that includes actions and placement rules. Actions are imperative operations to reconfigure the data center and declarative rules specify the virtual machine placement. Administrators schedule both actions and rules, to manage their data center(s). They can also interact with the btrScript system in order to monitor the data center and compute the correct virtual machine placement  [25] . btrScript and Entropy have been packaged in a common software btrCloud.