Network slicing in software datapaths for 5G multi-tenant networks

Abstract

5G is reshaping the current landscape in mobile networks, opening up new business opportunities. Mobile industry can now address use cases that are not feasible for 4G technology dueto the strict requirements demanded in terms of delay, reliability, bandwidth, scalability andhigh connection density. This will enable to address the requirements of challenging eMBB,URLLC and mMTC use cases in highly dynamic and evolving environments.To meet this ambitious QoS requirements, the evolution from 4G to 5G has imposed theshift to a new paradigm, leaving behind the static architectures based on the "one-size-fitsall" approach commonly used by 4G and its precursors, bringing a SBA with the ability toaddress the specific requirements of vertical industry to customer segments. ISPs are facingsignificant financial investments to adapt their networks to this new landscape, devising newbusiness models in a growing and dynamic global telecommunication market. 5G relies onvirtualization and softwarization, using technologies such as SDN and NFV as key enablers,leading to a reduction in both CAPEX and OPEX. Softwarization and virtualization facilitatethe deployment of multi-tenant networks in which several tenants share the same physicalinfrastructure. In this context, network slicing has emerged as a cornerstone technology,enabling the deployment of different virtualized logical networks running on top of thesame physical network while providing network isolation performance between tenants andguaranteeing QoS requirements.The notion of network slicing is an E2E concept that entails its implementation in allsegments throughout the 5G network. This represents a significant challenge due to thecomplexity of 5G networks. This research work aims to deliver network slicing capabilitiesin the Edge-Core segment, more precisely in the software datapath, which is responsible for connecting the virtual machines with the common shared physical network. To this end,this research work proposes a novel software datapath able to handle all types of networktraffic envisioned in 5G networks, including encapsulated traffic inherent to multi-tenant 5Gnetworks. This solution is based on SDN principles, providing the control and managementlayers with extended programmability beyond the current state of the art.The proposed solution has been designed, prototyped, validated, and evaluated in thecontext of the H2020 5G-PPP Phase II SliceNet project. Its suitability has been empiricallytested in the edge-core segment of a realistic 5G multi-tenant architecture where promisingresults in terms of bandwidth, delay, reliability, and scalability have been achieved. Theprototype has been successfully applied to cope with different scenarios such as securityin 5G NB-IoT networks, network slicing support for SFC in VNF and network slicing forheterogeneous traffic in 5G IoT networks.In the security use case, the solution operates as a firewall for NB-IoT traffic, beingable to simultaneously deal with traffic from 1 million devices while achieving a bandwidthof 4 Gbps and 8% of packet loss. In the SFC scenario, the datapath software adopts therole of SFF agent with network slicing capabilities, being able to cope with 16384 SFChops at 20 Gbps while delivering a maximum delay of 11 microseconds and no packet loss.Finally, in the most stressful scenario, the proposed solution simultaneously copes withheterogeneous traffic from 5 IoT use cases representing all 3 ITU categories or a combination.The gathered results demonstrate that this novel software datapath guarantees performanceisolation between slices while meeting the stringent 5G KPIs. It provides high levels ofscalability for mMTC traffic (up to 1 million devices), ultra-high reliability and ultra-lowlatency for delay-sensitive critical-mission traffic (0% packet loss ratios and delay in the orderof microseconds) while reaching peaks of 15 Gbps for eMBB use cases. These promisingoutcomes indicate the feasibility of the proposed solution for the software segment of the 5GDP with the aim of delivering E2E network slicing capabilities in 5G multi-tenant networksand, thus, satisfying the demanding QoS requirements foreseen in 5G use cases.