New Reading Group: “Orchestration and engineering of virtual networks”

Context:

A global tendency is the virtualization of equipment and infrastructures. This tendency concerns the horizontal dimension (from the user equipment to the core network and the data center) as well as the vertical dimension (from the physical layer to the application layer). This virtualization opens the way for the generalization of multi-player environments: multiple infrastructure vendors, multiple service operators, etc.

In this context, there is a clear need for orchestration of infrastructure resources to match service needs. Orchestration at different levels is to be considered.  There is first the need for allocating physical resources to the different virtual machines or containers. A second, higher level of orchestration is then needed that ensures that resources are properly allocated to the different services, knowing the needs of each service and the operator’s objectives. While the main focus of research and standardization works has been on the first level of orchestration, there is a clear need for researching the second level of orchestration by experts that have the proper knowledge on network engineering and stochastic traffic analysis. Finally, the vertical dimension of virtualization requires the definition of a modular approach for building customized L2-L7 stacks that implement scalable data plane services.

Use cases:

Although orchestration in general is a topic of interest to LINCS researchers, we’ll focus on the following use cases that are natural extensions of our current research works:

1-      Virtual Radio Access Networks (VRAN) where base band equipment of base stations is replaced by virtualized equipment, opening the way for interesting resource allocation problems, both on the radio resource level and on the hardware resource level. Virtual RAN is also viewed as an enabler for Mobile Edge Computing with many interesting applications, e.g. related to the Internet of Things.

2-      Information Centric Networking is a popular design for the evolution of the Internet infrastructure.  Despite its promising benefits, its deployment has been refrained by the severe modifications required to network devices.  On the one hand, a virtual approach will enable ICN deployment without any modification to the underlying devices. On the other hand, this open new challenges in the design of software data plane modules to support ICN as virtual network service.

Challenges:

The following challenges are linked to the need for orchestration at different levels:

1-      Identification of the different services and their modeling.

  • Competences on abstraction and semantic approaches are needed.

2-      Algorithmic aspects for resource allocation taking into account two new dimensions: the geographical dimension (aggregate resources at different positions when needed) and the modular behavior of the physical resources (allocation of servers to the different services)

  • Competences on traffic engineering/scheduling/resource allocation
  • Competences on networking (systems and protocols)

3-      Traffic and performance monitoring, with the need of defining new virtual interfaces between network nodes

  • Competences on traffic metrology

4-      Verification/validation of decisions taken by the self-organizing mechanisms, taking into account the dynamic behavior of services

  • Competences on self-organizing networks

5-      Business models, knowing the presence of multiple actors

  • Competences on security/trust
  • Competences on economic modeling

6- Virtual data plane services

  • Competences on networking (systems and protocols)

Technical animation:

The technical animation of this working group will be ensured by Daniel.