Section: New Results

Cloud applications and infrastructures

Participants : Adrien Lebre, Thomas Ledoux, Yousri Kouki, Guillaume Le Louët, Jean-Marc Menaud, Jonathan Pastor, Flavien Quesnel, Mario Südholt.

In 2014, we have provided solutions for Cloud-based and distributed programming, virtual environments and data centers, in particular concerning energy-optimal Cloud applications.

Cloud and distributed programming

This year we have published results on a broker that provides better guarantees on service-level agreements in the Cloud. Furthermore, we have extended a class of formally-defined protocols, session types.

Service-level agreement for the Cloud

Elasticity is the intrinsic element that differentiates Cloud Computing from traditional computing paradigms, since it allows service providers to rapidly adjust their needs for resources to absorb the demand and hence guarantee a minimum level of Quality of Service (QoS) that respects the Service Level Agreements (SLAs) previously defined with their clients. However, due to non-negligible resource initiation time, network fluctuations or unpredictable workload, it becomes hard to guarantee QoS levels and SLA violations may occur.

We propose a language support for Cloud elasticity management that relies on CSLA (Cloud Service Level Agreement) [27] . CSLA offers new features such as QoS/functionality degradation and an advanced penalty model that allow providers to finely express contracts so that services self-adaptation capabilities are improved and SLA violations minimized. The approach was evaluated with a real infrastructure and application testbed. Experimental results show that the use of CSLA makes Cloud services capable of absorbing more peaks and oscillations by trading-off the QoS levels and costs due to penalties.

AO session types for distributed protocols

Multiparty session types allow the definition of distributed processes with strong communication safety properties. A global type is a choreographic specification of the interactions between peers, which is then projected locally in each peer. Well-typed processes behave accordingly to the global protocol specification. Multiparty session types are however monolithic entities that are not amenable to modular extensions. Also, session types impose conservative requirements to prevent any race condition, which prohibit the uniform application of extensions at different points in a protocol. We have proposed a means to support modular extensions with aspectual session types [32] , a static pointcut/advice mechanism at the session type level. To support the modular definition of crosscutting concerns, we have augmented the expressivity of session types to allow harmless race conditions. As a result, aspectual session types make multiparty session types more flexible, modular, and extensible.

Virtualization and data centers

In 2014, we have produced a variety of results on a new model for utility computing that addresses fundamental shortcomings of today's Cloud computing model. Furthermore, we have provided more powerful techniques for the virtualization of computations and the management of cluster-based environments, such as data centers.

Next generation utility computing

To accommodate the ever-increasing demand for Utility Computing (UC) resources while taking into account both energy and economical issues, the current trend consists in building larger and larger data centers in a few strategic locations. Although such an approach enables to cope with the actual demand while continuing to operate UC resources through centralized software system, it is far from delivering sustainable and efficient UC infrastructures. Throughout the Discovery initiative (http://beyondtheclouds.github.io ), we investigate how UC resources can be managed differently, considering locality as a primary concern. Concretely, we study how it can be possible to leverage any facilities available through the Internet in order to deliver widely distributed UC platforms that can better match the geographical dispersal of users as well as the unending resource demand. Critical to the emergence of such locality-based UC (LUC) platforms is the availability of appropriate operating mechanisms. We presented a prospective vision of a unified system driving the use of resources at an unprecedented scale by turning a complex and diverse infra structure into a collection of abstracted computing facilities that is both easy to operate and reliable [35] . By deploying and using such a LUC Operating System on backbones, our ultimate vision is to make possible to host/operate a large part of the Internet by its internal structure itself: A scalable and nearly infinite set of resources delivered by any computing facilities forming the Internet, starting from the larger hubs operated by ISPs, governments and academic institutions to any idle resources that may be provided by end-users. We highlight that this work is conducted through a collaboration between the ASAP, ASCOLA, AVALON and MYRIADS Inria Project-teams.

Adding locality capabilities to virtual machine schedulers

Through the DVMS proposal, we showed in 2013 the benefit of leveraging peer-to-peer algorithms to design and implement virtual machines (VMs) scheduling algorithms. Although P2P based proposals considerably improve the scalability, leading to the management of hundreds of thousands of VMs over thousands of physical machines (PMs), they do not consider the network overhead introduced by multi-site infrastructures. This over- head can have a dramatic impact on the performance if there is no mechanism favoring intra-site v.s. inter-site manipulations. This year, we extended our DVMS mechanism with a new building block designed on top of the Vivaldi coordinates mechanism. We showed its benefits by discussing several experiments performed on four distinct sites of the Grid’5000 testbed. With our proposal and without changing the scheduling decision algorithm, the number of inter-site operations has been reduced by 72% [29] . This result provides a glimpse of the promising future of using locality properties to improve the performance of massive distributed Cloud platforms. We highlight that this work has been performed in collaboration with the ASAP, ASCOLA, AVALON and MYRIADS Inria Project-teams.

WAN-wide elasticity capabilities for distributed file systems

Applications dealing with huge amounts of data suffer significant performance impacts when they are deployed on top of an hybrid platform (i.e the extension of a local infrastructure with external cloud resources). More precisely, through a set of preliminary experiments we shew that mechanisms which enable on demand extensions of current Distributed File Systems (DFSes) are required. These mechanisms should be able to leverage external storage resources while taking into account the performance constraints imposed by the physical network topology used to interconnect the different sites. To address such a challenge we presented the premises of the Group Based File System, a glue providing the elasticity capability for storage resources by federating on demand any POSIX file systems [28] .

Energy optimization

Demand for Green services is increasing considerably as people are getting more environmental conscious to build a sustainable society. Therefore, enterprise and clients want to shift their workloads towards green Cloud environment offered by the Infrastructure-as-a-Service (IaaS) provider. The main challenge for an IaaS provider is to determine the best trade-off between its profit while using renewable energy and customers satisfaction. In order to address this issue, we propose a Cloud energy broker [26] , which can adjust the availability and price combination to buy Green energy dynamically from the market to make datacenter green. Our energy broker tries to maximize of using renewable energy under strict budget constraint whereas it also tries to minimize the use of brown energy by capping the limit of overall energy consumption of datacenter. The energy broker was evaluated with a real workload traced by PlanetLab. Experimental results show that our energy broker successfully enables meeting the best trade-off.