Schedule a free product or technology session with Grandmetric Engineer
schedule a video call

Blog

IP and Mobile Trends and Education

 

5G-mobile vs. IP-networks techniques: Slicing

Author:


05.07.2016

Within this series of posts, I want to review the 5G concepts from the IP-networks perspective and explore the available techniques on the market which allow translating the presented concepts into real infrastructures. This post is related to a concept known as slicing.

Recently, there are many documents describing 5G principles and architectures. One of them, “SDN for Mobile Networks: 5G-NORMA Perspective” by Bessem Sayadi, et al. (from 5G-NORMA project), concentrates on Software-Defined Networking (SDN) and Network Function Virtualization (NFV) in conjunction with the network slicing and multi-tenancy concepts.

Main characteristics of 5G architecture are flexibility and adaptability. Those two can be achieved by SDN and NFV. In turn, SDN and NFV are concepts of software controlled network infrastructure which can deliver many independent services to different groups of customers, from different groups of providers (i.e. content enablers) on different levels of QoE (Quality of Experience), and still functioning as one intelligent, secure organism.

Slicing

5G should bring the required flexibility to support many services in a multi-tenant environment with different requirements, throughput, and availability. That is why slicing is the key concept in terms of network virtualization. The concept of multi-tenancy or slicing is rather known for IP Networking world and can be achieved on different levels. Because 5G needs to support an end to end slicing, it is important to design on RAN, as well as on Backhaul, Internet Edge, and Data Centers. Each of segment should be able to slice the available network resources with given business SLA.

Slicing in 5G

Figure 1. Slicing in 5G High-Level view

Layer 2 slicing can be achieved by Virtual LAN (VLAN) instances or VxLAN (encapsulated L2). The first one is used widely in the market by Service providers as well as in Enterprise environments. Through the years VLANs serve the role of traffic separation and broadcast domain limitation and are used in almost every segment of the network. VxLANs have became popular nowadays as cloud environments are gaining popularity. Since VLANs have viewable limitations, VxLANs solve problems of scalability and service location stickiness (VLAN: 2^12 = 4096 Tag space, VxLAN: 2^24 = 16777216). VxLAN frames are encapsulated with UDP thus can be transported across L3 clouds.

Layer 3 slicing is performed mainly with Virtual Routing and Forwarding (VRF) instances. VRFs creates virtual domains and routing tables giving the isolation of IP packets thus allowing service providers to utilize common infrastructure for many customers even when customers use overlapping IP addressing spaces. In LAN environments VRFs are used for traffic separation hop by hop by isolating particular segments. In SP environments, VRF accommodates the Route Target and Route Distinguisher tags which travel with IP Packets and are imposed at Provider Edge router when coming into SP cloud and exposed at PE when coming out from SP cloud. Of course, behind this global customers separation few more SP techniques like L3 VPNs in MPLS network, MP-BGP routing, and underlying IS-IS routing come into play.

Slicing 5G L2 L3

Figure 2. Isolation through different segments

Slicing within Data Center was and is still often performed mainly with VxLANS and VLANs from the networking point of view. But since IT world became virtualized and there is also demand for VMs separation (virtual switch level) there are other aspects like Endpoint Group Isolation, vNIC isolation, Security Profiles, VSANs (storage communication). Also, Data Center technologies nowadays demand high percent of SLA which is achieved also through High Availability based on geographical distribution. This, in turn, requires all techniques described above and more to communicate with Data Centers distributed along long distances. Another aspect which determines isolation rules in Data Center is DC design itself. Because old, standard technologies like Spanning Tree Protocols (STP) were found to be inefficient, Spine and Leaf topologies were introduced which radically improve performance, eliminate bottlenecks and provide faster recalculations and convergence. Spine and Leaf topology (based on Trill protocol on top of the IS-IS) requires more sophisticated approaches to traffic isolation and network slicing.

Summary

As shown above, 5G-NORMA architecture can be translated precisely to available IP-networks technologies. In this post, I focused on Slicing paradigm of 5G and how it can be implemented in real networks. As we can see there are a lot of aspects how to achieve slicing in particular segment of the network, so those subjects will be also discussed in the future posts.

 

Author

Grandmetric

Grandmetric is an IT Next Generation Systems integration company helping clients with their IT transformation, infrastructure automation, LAN, WiFi, SD-WAN & SDN delivery. Fast growing Grandmetric team is becoming also a referal point in Cloud migrations and DC Stack management with their Storage, OS and virtualization experience. Grandmetric provides technical insights along with technical trainings in areas of expertise. Latest projects cover also IoT subjects R&D in the area of IoT backend development, big data analysis and monitoring. Based on above experience in production systems maintenance, new division – Grandmetric Managed Services (GMS) maintaining IT infrastructure of corporates & globally present customers is available for demanding IT environments.

1 Comment

Leave a Reply

Your email address will not be published. Required fields are marked *

*

code


 

Newsletter