You are here: Home > Future Internet Assembly > Budapest, May 2011 > Session III.1 The Network Lost in the Cloud > 
21.10.2016 : 8:59

Session III.1 The Network Lost in the Cloud



Nicolas Le Sauze, Alcatel-Lucent Bell Labs France, Marcus Brunner, NEC Europe


Cloud infrastructure typically consists of a set of servers, located in multiple data-centres and interconnected by one or several network domains, being public –e.g. the Internet, or private.  Since one big issue in this infrastructure is centred on resource sharing at various layers of the network and IT systems, this should be discussed from technical approach and from a business view. Cloud standardization is shaping up, but is still very scattered around various organizations, also since various aspects play into such paradigm.

SLA-driven approaches are good candidates to deliver guaranteed cloud services. As they are based on commercial contracts, they would ensure a clear and agreed revenue sharing between the cloud actors: telcos will provide dynamic connectivity/network services eventually composed with IT (computation/storage/transcoding/etc.) services or capabilities to build a global cloud service (which can either be sold by a telco or by a cloud SP, but including a clear revenue sharing with the collaborative partner(s)). In this model, the network part can already be a composed E2E network service resulting from the assembled network segments (or "goods") from several carriers.


Entitled "The Network lost in the Cloud?", the session will focus on issues related to infrastructures in support of Cloud services, raising discussions that will go along several dimensions and questions such as: what is the place of the network in cloud services? How to achieve global service assurance in the cloud age? What are the specific inter-cloud issues?

A key benefit from a cloud infrastructure is the dynamic allocation of infrastructure resources to customers according to immediate or scheduled needs. While dynamic server management is now quite mature, the tight coupling with the dynamic end-to-end network resource management remains challenging. In addition, other issues have to be clarified for a holistic approach:

  • About business models; a per-usage payment model is generally foreseen in the cloud area, but it is not yet set in stone.
  • What about other kinds of resources (more soft resources like knowledge and information sharing) and their roles in the cloud?
  • What about neutrality? The discussions are focusing on the network side only, but the cloud will need to address the same or similar issues if it comes to resource sharing;
  • How flexible is the cloud model to incorporate IT and Network services from various actors? It must rely on trusted actors with some guarantees to deliver what has been promised, but should remain open to any new third party.


The session will be in a panel format with several speakers addressing different views along the above mentioned lines and followed by a discussion of those in the plenum with the audience.


16.30-16.40: undefinedSession Introduction – Reminder on the objectives (Session Chairs)
16.40-17.55: Invited talks

  • Opportunities and requirements from over the top SMEs providing cloud services: the example of HPC (undefinedAlban Schmutz, Oxalya)
  • How are Service Providers helping their customers to deliver IT as a Service? (undefinedPeter Glock, Orange Business Services)
  • Cloud connectivity, interconnect and SLAs - From a network perspective (undefinedHåkon Lonsethagen, Telenor, FP7 ETICS)
  • Building a distributed cloud infrastructure from the bottom: the SAIL approach to cloud networking. (undefinedVictor Souza, Ericsson Research, FP7 SAIL)
  • The revolution of infrastructure provisioning (undefinedPascale Vicat-Blanc, Lyatiss, FP7 Geysers)

17.55-18.25: Panel Discussions
18.25-18.30: Closing remarks (Session chairs)


Opportunities and requirements from over the top SMEs providing cloud services: the example of HPC
Abstract: Over the top (OTT), means running cloud services over plain IP without any involvement of a network provider. Taking as an example the High Performance Computing (HPC) application, this talk will focus on expectations from such an OTT actor on the network. The HPC market users will grow through the cloud, just like other market segments. Through the example of Oxalya, the presentation will discuss the issues of SMEs needing end-to-end network SLA to ensure high levels of services for its customers, facing real big industry requirements in terms of result delivery and analysis in Numerical Simulation.

How are Service Providers helping their customers to deliver IT as a Service?
Abstract: In the world of Cloud Computing there is a tendency to describe everything ‘as a Service’. In future, perhaps even now, users of IT services will expect to be able to consume IT that is delivered from in-house IT in the same way that they consume IT at home and from third party Cloud vendors. IT organisations will have to undergo a radical transformation to become the service provider of choice for their user communities otherwise business units may attempt to source IT direct from the Cloud.
What can organisations learn from the experiences of long established service providers like Orange Business Services, both from our own IT transformation program and from the competitive programs we have in place to help our customers make best use of the Cloud.

Cloud connectivity, interconnect and SLAs - From a network perspective

Abstract: Cloud services are growing in importance and as businesses and consumers become dependant on these services, the need for extensive SLAs become apparent. However, the cloud SLAs are of no use without a network supporting them. This presentation will look at cloud services and SLAs in relation to the ETICS EU project that aims to provide inter-domain Assured Service Quality (ASQ), specified using end-to-end network SLAs. The ETICS architecture will thus be an enabler for end-to-end cloud service SLAs.
Cloud computing introduces some important requirements towards the provisioning of network SLAs due to its on-demand characteristics, leading to high dynamicity, mobility etc. that must be supported by the ETICS architecture. Additionally, focus should be on the interface needed between cloud providers and network providers in order to bundle the cloud application SLA and network SLA towards the user.

Building a distributed cloud infrastructure from the bottom: the SAIL approach to cloud networking.
Abstract: This talk will focus on a new approach for deployment of cloud infrastructure services that is fully distributed within a network. SAIL Cloud Networking complements existing datacenter centric approaches with distributed processing leveraging upon network locality. The benefits of such a deployment are manifold: lower latency to reach end-users, avoidance of congested links, high scalability, amongst others. Furthermore, cloud networking integrates distributed computing and storage with networking, allowing for a holistic optimization of the application execution.

The revolution of infrastructure provisioning

Abstract: Many emerging Internet and Cloud applications have increasing requirements not only in terms of computing power but also in terms of network flexibility, bandwidth, QoS and manageability. These applications are still unable to exploit the potential of current optical network technologies. This is because the network layer is not aware of applications’ dynamic requirements and the Network and IT resources are provisioned and controlled separately without any integration between the Network Control Plane and the application. This talk will present the extension of the virtualization paradigm to the whole infrastructure to better serve the requirements of Internet and Cloud customers as proposed by the GEYSERS project. The GEYSERS architecture is a bottom–up approach able to provide service providers with dynamic, cost-efficient, mission-specific virtual infrastructures that can be provisioned and managed by a Virtual Infrastructure composition layer and controlled by means of an enhanced IT- and energ  aware Network Control Plane. The Virtual Infrastructure concept and the associated representation and the provisioning issues will be discussed.