Written by Aaron Williams, LF Edge Developer Advocate

As cloud computing migrates from large Centralized Data Centers to the Server Provider Edge’s Access Edge (i.e. Server-based devices at the Telco tower) to be closer to the end user and reduce latency, more applications and use cases are now available.  When you combine this with the widespread roll out of 5G, with its speed, the last mile instantly becomes insanely fast.  Yet, if this new edge stack is not constructed correctly, the promise of these technologies will not be possible.

LF Edge’s Akraino project has dug into these issues and created its second white paper called Cloud Interfacing at the Telco 5G Edge.  In this just released white paper, Akraino analyzes the issues, proposes detailed enabler layers between edge applications and 5G core networks, and makes recommendations for how Telcos can overcome these technical and business challenges as they bring the next generation of applications and services to life.

Click here to view the complete white paper: https://bit.ly/34yCjWW

To learn more about Akraino, blueprints or end user case stories, please visit the wiki: https://wiki.akraino.org.

Written by Oleg Berzin, Ph.D., a member of the Akraino Technical Steering Committee and Senior Director Technology Innovation at Equinix

Introduction

Why 5G

5G will provide significantly higher throughput than existing 4G networks. Currently, 4G LTE is limited to around 150 Mbps. LTE Advanced increases the data rate to 300 Mbps and LTE Advanced Pro to 600Mbps-1 Gbps. The 5G downlink speeds can be up to 20 Gbps. 5G can use multiple spectrum options, including low band (sub 1 GHz), mid-band (1-6 GHz) and mmWave (28, 39 GHz). The mmWave spectrum has the largest available contiguous bandwidth capacity (~1000 MHz) and promises dramatic increases in user data rates. 5G enables advanced air interface formats and transmission scheduling procedures that decrease access latency in the Radio Access Network by a factor of 10 compared to 4G LTE.

The Slicing Must Go On

Among advanced properties of the 5G architecture, Network Slicing enables the use of 5G network and services for a wide variety of use cases on the same infrastructure. Network Slicing (NS) refers to the ability to provision a common physical system to provide resources necessary for delivering service functionality under specific performance (e.g. latency, throughput, capacity, reliability) and functional (e.g. security, applications/services) constraints.

Network Slicing is particularly relevant to the subject matter of the Public Cloud Edge Interface (PCEI) Blueprint. As shown in the figure below, there is a reasonable expectation that applications enabled by the 5G performance characteristics will need access to diverse resources. This includes conventional traffic flows, such as access from mobile devices to the core clouds (public and/or private) as well as the general access to the Internet, edge traffic flows, such as low latency/high speed access to edge compute workloads placed in close physical proximity to the User Plane Functions (UPF), as well as the hybrid traffic flows that require a combination of the above for distributed applications (e.g. online gaming, AI at the edge, etc). One point that is very important is that the network slices provisioned in the mobile network must extend beyond the N6/SGi interface of the UPF all the way to the workloads running on the edge computing hardware and on the Public/Private Cloud infrastructure. In other words, “The Slicing Must Go On” in order to ensure continuity of intended performance for the applications.

The Mobile Edge

The technological capabilities defined by the standards organizations (e.g. 3GPP, IETF) are the necessary conditions for the development of 5G. However, the standards and protocols are not sufficient on their own. The realization of the promises of 5G depends directly on the availability of the supporting physical infrastructure as well as the ability to instantiate services in the right places within the infrastructure.

Latency can be used as a very good example to illustrate this point. One of the most intriguing possibilities with 5G is the ability to deliver very low end to end latency. A common example is the 5ms round-trip device to application latency target. If we look closely at this latency budget, it is not hard to see that to achieve this goal a new physical aggregation infrastructure is needed. This is because the 5ms budget includes all radio/mobile core, transport and processing delays on the path between the application running on User Equipment (UE) and the application running on the compute/server side. Given that at least 2ms will be required for the “air interface”, the remaining 3ms is all that’s left for the radio/packet core processing, network transport and the compute/application processing budget. The figure below illustrates an example of the end-to-end latency budget in a 5G network.

The Edge-in and Cloud-out Effect

Public Cloud Service Providers and 3rd-Party Edge Compute (EC) Providers are deploying Edge instances to better serve their end-users and applications, A multitude of these applications require close inter-working with the Mobile Edge deployments to provide predictable latency, throughput, reliability, and other requirements.

The need to interface and exchange information through open APIs will allow competitive offerings for Consumers, Enterprises, and Vertical Industry end-user segments. These APIs are not limited to providing basic connectivity services but will include the ability to deliver predictable data rates, predictable latency, reliability, service insertion, security, AI and RAN analytics, network slicing, and more.

These capabilities are needed to support a multitude of emerging applications such as AR/VR, Industrial IoT, autonomous vehicles, drones, Industry 4.0 initiatives, Smart Cities, Smart Ports. Other APIs will include exposure to edge orchestration and management, Edge monitoring (KPIs), and more. These open APIs will be the foundation for service and instrumentation capabilities when integrating with public cloud development environments.

Public Cloud Edge Interface (PCEI)

Overview

The purpose of Public Cloud Edge Interface (PCEI) Blueprint family is to specify a set of open APIs for enabling Multi-Domain Inter-working across functional domains that provide Edge capabilities/applications and require close cooperation between the Mobile Edge, the Public Cloud Core and Edge, the 3rd-Party Edge functions as well as the underlying infrastructure such as Data Centers, Compute hardware and Networks. The Compute hardware is optimized and power efficient for Edge such as the Arm64 architecture.

The high-level relationships between the functional domains are shown in the figure below:

The Data Center Facility (DCF) Domain. The DCF Domain includes Data Center physical facilities that provide the physical location and the power/space infrastructure for other domains and their respective functions.

The Interconnection of Core and Edge (ICE) Domain. The ICE Domain includes the physical and logical interconnection and networking capabilities that provide connectivity between other domains and their respective functions.

The Mobile Network Operator (MNO) Domain. The MNO Domain contains all Access and Core Network Functions necessary for signaling and user plane capabilities to allow for mobile device connectivity.

The Public Cloud Core (PCC) Domain. The PCC Domain includes all IaaS/PaaS functions that are provided by the Public Clouds to their customers.

The Public Cloud Edge (PCE) Domain. The PCE Domain includes the PCC Domain functions that are instantiated in the DCF Domain locations that are positioned closer (in terms of geographical proximity) to the functions of the MNO Domain.

The 3rd party Edge (3PE) Domain. The 3PE domain is in principle similar to the PCE Domain, with a distinction that the 3PE functions may be provided by 3rd parties (with respect to the MNOs and Public Clouds) as instances of Edge Computing resources/applications.

Architecture

The PCEI Reference Architecture and the Interface Reference Points (IRP) are shown in the figure below. For the full description of the PCEI Reference Architecture please refer to the PCEI Architecture Document.

Use Cases

The PCEI working group identified the following use cases and capabilities for Blueprint development:

1. Traffic Steering/UPF Distribution/Shunting capability — distributing User Plane Functions in the appropriate Data Center Facilities on qualified compute hardware for routing the traffic to desired applications and network/processing functions/applications.
2. Local Break-Out (LBO) – Examples: video traffic offload, low latency services, roaming optimization.
3. Location Services — location of a specific UE, or identification of UEs within a geographical area, facilitation of server-side application workload distribution based on UE and infrastructure resource location.
4. QoS acceleration/extension – provide low latency, high throughput for Edge applications. Example: provide continuity for QoS provisioned for subscribers in the MNO domain, across the interconnection/networking domain for end-to-end QoS functionality.
5. Network Slicing provisioning and management – providing continuity for network slices instantiated in the MNO domain, across the Public Cloud Core/Edge as well as the 3Rd-Party Edge domains, offering dedicated resources specifically tailored for application and functional needs (e.g. security) needs.
6. Mobile Hybrid/Multi-Cloud Access – provide multi-MNO, multi-Cloud, multi-MEC access for mobile devices (including IoT) and Edge services/applications
7. Enterprise Wireless WAN access – provide high-speed Fixed Wireless Access to enterprises with the ability to interconnect to Public Cloud and 3rd-Party Edge Functions, including the Network Functions such as SD-WAN.
8. Distributed Online/Cloud Gaming.
9. Authentication – provided as service enablement (e.g., two-factor authentication) used by most OTT service providers 
10. Security – provided as service enablement (e.g., firewall service insertion)

The initial focus of the PCEI Blueprint development will be on the following use cases:

• User Plane Function Distribution
• Local Break-Out of Mobile Traffic
• Location Services

User Plane Function Distribution and Local Break-Out

The UPF Distribution use case distinguishes between two scenarios:

• UPF Interconnection. The UPF/SPGW-U is located in the MNO network and needs to be interconnected on the N6/SGi interface to 3PE and/or PCE/PCC.
• UPF Placement. The MNO wants to instantiate a UPF/SPGW-U in a location that is different from their network (e.g. Customer Premises, 3rd Party Data Center)

UPF Interconnection Scenario

UPF Placement Scenario

UPF Placement, Interconnection and Local Break-Out Examples

Location Services (LS)

This use case targets obtaining geographic location of a specific UE provided by the 4G/5G network, identification of UEs within a geographical area as well as facilitation of server-side application workload distribution based on UE and infrastructure resource location.

 

Acknowledgements

Project Technical Lead: Oleg Berzin

Committers: Suzy GuTina Tsou Wei Chen, Changming Bai, Alibaba; Jian Li, Kandan Kathirvel, Dan Druta, Gao Chen, Deepak Kataria, David Plunkett, Cindy Xing

Contributors: Arif , Jane Shen, Jeff Brower, Suresh Krishnan, Kaloom, Frank Wang, Ampere

Written by Tapio Tallgren, Technical Leader at Nokia Mobile Networks, Community Sub-Committee Chair of Akraino TSC,Ferenc Szekely, Program Manager, SUSE, Committer of Micro MEC blueprint of Akraino TSC and Tina Tsou, Enterprise Architect, Arm, Akraino TSC Co-Chair

The MicroMEC platform started life as a platform to run applications at the very edge of the network, like in a light pole. We joined the LF Edge’s Akraino project from the very beginning.

To find out what the use cases would be first, we participated in the IoThon hackathon in 2019 where we built a miniature city with sensors, cameras and small servers — also known as Raspberry Pis. Our plan was that we will provide APIs to enable developers to access the sensors, cameras, or other independent hardware devices attached to our small servers, ie. the MicroMEC nodes. It was clear that we wanted to deploy all the APIs as well as the apps in containers. We needed a tool like Kubernetes to help us build and manage the MicroMEC cluster. As we targeted “small” devices, with max 4GB of RAM -at that time- and low power consumption we looked into alternatives to k8s. That is how we picked k3s. 

By the autumn of 2019 we had our lab running Raspberry Pi 3B+ and 4B nodes with k3s. We had a successful hackathon – Junction 2019 – in Finland where the teams presented solutions utilizing the MicroMEC cluster. We also added OpenFaaS Cloud (OFC) into the mix and a developer UI to the platform. This allowed developers to write serverless applications for the MicroMEC cluster and deploy them with ease. They could concentrate on their core business: developing apps while MicroMEC with OFC took away the burden of cluster management, deployment etc.

Right after Junction, we were at the Akraino 5G MEC Hackathon in the USA. For this event MicroMEC had to become more “MEC”. This implied the implementation of MEC-11 interfaces and the UI to manage those apps that our MEC-11 implementation made discoverable for customers near the MicroMEC cluster. The MEC cluster runs on Arm architecture based hardware.

With all this activity, we missed the first two Akraino releases, but now we are very happy to join the Akraino R3 release! For this, we had to figure out what is the easiest way to install the stack on the device with a MMC card. The easiest way is to not install anything on the fragile card, but boot the stack from a network server. Eventually we made all MicroMEC nodes to boot from a network server using PXE and the storage of each node was attached via iscsi. This requires a fast enough LAN, but thankfully cheap gigabit switches are widely available these days. 

Learn more about Akraino here.