Resource Elasticity for 5G Network Architecture

Wednesday, 20 June 2018, 16:00-17:30, E3 hall
Session Chair:
  • David Gutierrez Estevez (Samsung Electronics R&D Institute UK, UK)

Vertical markets and industries are addressing a large diversity of heterogeneous services, use cases, and applications in 5G. It is understood that for networks to satisfy those needs, a flexible, adaptable, and programmable architecture based on network slicing is required. Moreover, a softwarisation and cloudification of communications networks are already happening, as network functions (NFs) are transformed from monolithic pieces of equipment to programs running over a shared pool of computational and communication resources. However, this novel architectural paradigm requires new solutions to exploit this flexibility, given that a careless multiplexing of NFs over the same resources could result in resource wastage if over-provisioning or service disruption if under-provisioning. Furthermore, introducing resource-awareness to the NFs would enable maximising the efficiency in the use of resources.

In the 5GPPP 5G-MoNArch project, we refer to the ability to flexibly adapt the operation of function (or even a service) as resource elasticity. This new family of solutions can be applied both to resources in the cloud (i.e., CPU, memory, storage) and communications resources (i.e., spectrum, fiber). Although elasticity in networks has already been traditionally exploited in the context of communications resources, in this special session we focus on the computational aspects of resource elasticity, as we identify the management of computational resources in networks a key challenge of future virtualized and cloudified 5G systems.

The special session program will be divided into five parts. The first presentation starts by describing the 5G-MoNArch overall architecture, defining elasticity, and describing its architectural implications. Then, we will address the different types of elasticity we envision in three different presentations: i) computational elasticity in the design and scaling of NFs, ii) orchestration-driven elasticity achieved by flexible placement of NFs, and iii) slice-aware elasticity via cross-slice resource provisioning mechanisms. Finally, to close the special session we propose a techno-economic analysis of elasticity.

Programme

Overall 5G-MoNArch Architecture and Implications for Resource Elasticity
David Gutierrez and Mehrdad Shariat, Samsung, United Kingdom

The fifth generation (5G) of mobile and wireless communications networks aims at addressing a diverse set of use cases, services, and applications with a particular focus on enabling new business cases via network slicing. The development of 5G has thus advanced quickly with research projects and standardization efforts resulting in the 5G baseline architecture. Nevertheless, for the realization of native end-toend (E2E) network slicing, further features and optimizations shall still be introduced. In this paper, essential building blocks and design principles of the 5G architecture will be discussed capitalizing on the innovations that are being developed in the 5G-MoNArch project. Furthermore, building on the concept of resource elasticity introduced by 5G-MoNArch and briefly resummarized in this paper, an elasticity functional architecture is presented where the architectural implications required for each of the three dimensions of elasticity are described, namely computational, orchestration-driven, and slice-aware elasticity.

On the Benefits of Bringing Cloud-Awareness to Network Virtual Functions
Pablo Serrano, Universidad Carlos III de Madrid, Spain

We are currently observing the softwarization of communication networks, where network functions are translated from monolithic pieces of equipment to programs running over a shared pool of computational, storage, and communication resources. As the amount of this resources might vary over time, in this paper we discuss the potential benefits of introducing resource awareness to softwarized network functions. More specifically, we focus on the case of computational elasticity, namely, the ability to endure shortages of computational resources while providing an adequate (although non-ideal) service. We discuss how to enable this ability by re-designing network functions, and illustrate the potential benefits of this approach with a numerical evaluation.

Dynamic Deployment of Virtual Network Functions in Heterogeneous Telco Clouds
Emilio Calvanese, CEA-LETI, France

The next generation of cellular networks will leverage on network slicing to deal in a cost efficient way with the service requirements of the vertical markets. With this approach the operator can logically split a single network infrastructure in multiple instances whose virtual network functions (VNFs) are designed for each specific service. This new paradigm requires new functionalities and interfaces to take fully advantage of the pool of computational and communication resources and VNFs shared across multiple network slices. In this paper, we introduce the concept of orchestration-driven elasticity that enables the 3GPP Network Slice Management Function (NSMF) to dynamically adapt the VNF placement in the cloud infrastructure according to the momentary service requirements, and to implement proactive measures that avoid computational outages.

Slice-Aware Elastic Resource Management
Sina Khatibi, NOMOR Research, Germany

Network softwarisation and network slicing are two important 5G technology enablers. Implementing mobile networks over the commercial data-centres has proven to have considerable benefits. Realising cloud-based mobile networks and serving multiple network slices with different requirements over the same virtualised physical infrastructure are challenging tasks. The IP tsunami and dramatic temporal and spatial variation of traffic demands made the situation worse. In this situation, slice-aware elastic resource management approaches are required to guarantee the quality of offered services to the slices while minimising the capital and operational expenditure of networks. This paper provides a brief overview on the design guidelines for slice-aware elastic resource management.

Techno-economic analysis of elasticity
Julie Bradford, Real Wireless, United Kingdom

This short paper considers some of the economic drivers and considerations for elasticity as proposed in virtualised 5G network architectures such as under the 5G-MoNArch project. The current commercial challenges faced in today’s mobile networks are first introduced, and how network slicing and improved network flexibility and elasticity promises to go some way to addressing these. An evolved mobile ecosystem enabled by network slicing and virtualisation is applied to an example use cases to understand how network elasticity may drive better cost efficiencies within this framework.