Professional Documents
Culture Documents
com/
distributed-cloud
Edge computing
and 5G
Harnessing the distributed cloud for 5G success
June 2019
2 Ericsson | Edge computing and 5G
By 2023, 5G will make up around one-fifth How about the public clouds? distributed sites and integrated seamlessly
of all mobile data traffic, with subscription Public cloud service providers will have into the public cloud infrastructure and APIs.
uptake forecast to reach 1 billion. 4G LTE to make major investments to build their Others are rolling out highly distributed
has also exploded since its launch in 2010, own edge infrastructure. This presents an infrastructures; for example, Google has
accounting for some 50 percent of total opportunity for operators to play a role in deployed more than 1,000 Google Global
traffic in 2017 and still growing. Distributed delivering these new services by providing Caches. However, none of these solutions
cloud computing is paving the way for the the following assets: supports the vast number of potential mobile
future of network communications and, –– Geographical presence and distribution use cases where the computing capacity
rather than waiting for 5G, operators of sites at cell towers as well as sites would have to be deployed between the
should build tomorrow's networks today. across the network, creating sweet antennas and the current peering points.
spots for every deployment These use cases require two things:
What is the edge? –– Value-added connectivity with QoS –– Local breakout of the traffic close to
Edge computing provides execution capabilities – not just pipes but smart the user and to the compute – this is
resources (compute and storage) with connectivity with the ability to steer, not available today, but is something
sufficient connectivity (networking) at secure and break out traffic that only operators can provide
close proximity to the data sources, typically –– NFV-enabled network, providing the –– A compute infrastructure to run
within or at the boundary of access networks. ability to place initial workloads and the platforms – this can either
Additionally, deployments within the grow into the edge be operator-owned or public
customer premises managed or hosted cloud-owned infrastructure
by the operator are possible, for example Solutions complementing public clouds
in homes, enterprises and factory floors, are necessitated by use cases with high Additionally, integrating with hundreds
as well as vehicles, including trains, planes requirements on latency, bandwidth and of operators will be cumbersome for public
and private cars. The core benefits of edge local survivability/reliability. Some providers clouds, commercially and technically.
solutions are low latency, high bandwidth, use platforms like Greengrass (AWS) or
and trusted computing and storage. IoT Edge (Azure) that can be deployed at
3 Ericsson | Edge computing and 5G
Distributed cloud is suitable for use Which model is best depends on the strategy This is possible in several ways. Developers
cases with demanding requirements for and market position of the CSP, the type of could be approached either directly or
scalability, security, availability, low latency industry being addressed, and how far their through a facilitator such as Ericsson’s
and bandwidth. These use cases account for ecosystems have developed already. Edge Gravity. Another possibility is direct
25 percent of the 5G business potential, The assets held by CSPs make them collaboration with public cloud providers
and therefore revenue potential. uniquely placed to contribute with essential which have existing relationships with
capabilities based on distributed cloud developers, but no distribution capabilities.
Capturing edge business opportunities and exposure of cloud, network and Other benefits we see are transport and
When addressing edge opportunities, one orchestration logic. Since many developers processing offload, which will be significant
of the key aspects is how to work with the don’t know how telecom networks work, with the increased traffic of a 5G world, as
application ecosystem to provide solutions it is important to simplify API exposure as well as increased efficiency of connected
for industries. There are three ways for much as possible. CSPs should aim for a devices that are offloading to the edge.
communications service providers (CSPs) converged solution with cloud execution, The business benefits are already
to do this: service exposure, and dynamic workload apparent: we have systems in commercial
–– Work with industries to drive the placements through industry-aligned APIs production with several tier 1 and tier 2
ecosystem to provide solutions to best meet the needs of both industries customers. These customers experienced
–– Drive the ecosystem on their own and developers. dramatic improvements in the overall
–– Leverage the hyperscale cloud To succeed, operators also need to time-to-market of their network services,
providers’ ecosystems attract content and application developers. with cost reductions of up to 75 percent.
25%
With the app closer to the user and
Application latency 5G radio, the latency can be <1ms
and benefit new use cases.
Where to best place the applications? Data-heavy applications require broader of the distributed cloud and telecom
From a cost perspective, centralization is distribution to optimize bandwidth application program interfaces (APIs), or
generally preferred. However, the location consumption. Key requirements are the geographical location of edge locations
depends on the use case and the value it reliability and availability for applications in the network, in order to secure the SLAs.
provides, therefore the edge cannot be a that must always work, even if connectivity Discussions with customers and
fixed point. to central sites fails. application providers have confirmed that
In the example of AR, pictures from The function for location of the using computing at the edge must be as
a phone are sent to a datacenter. This applications should be integrated in the simple as using public cloud services.
activity is neither latency-sensitive nor overall service and network orchestration
bandwidth-intense, so a centralized to be used for NFV applications, such as Cloud native applications
datacenter is best. Then augmentation 5G Core and virtualized RAN functions 5G and IoT unleash a variety of workloads
data like videos and 3D models are sent to (vRAN). This is important to facilitate with different characteristics and needs.
the device, a bandwidth-demanding task automatic placement and network Platforms must therefore be able to cater
requiring content delivery network (CDN) slice orchestration. for the use of different deployment models
functionality, which is best placed at supporting both virtual machine and
distributed sites. Finally, videos or pictures Exposure and application development container-based workloads.
can be augmented in the network, which There must be a solution to enable
is very latency-sensitive and requires application development and onboarding.
close proximity. The developers need not know the details
Mobile
broadband
AI
PPF
Enterprise vRAN OSS ENM, EO, etc.
RCF
AR/VR UPF
Local Full Packet Core
Dual-Mode
IoT breakout UPF EPC/5GC Public
clouds
IMS
Critical IoT GW-U
AGW
Edge Communication
Media IMS
computing services
Private LTE Internet
3pp 3pp
Fixed wireless
Local PoP Edge computing
access CDN CDN
Dual-Mode
EPC/5GC
New
applications
Good business models and Management, orchestration and LCM It will be crucial for the success of edge
API exposure for global reach With potentially hundreds of edge locations deployments to maintain a competitive total
Historically, application developers like central offices, radio base stations or on cost of ownership (TCO) for edge sites. We
were required to learn telecom network enterprise premises, management will be need to support systems of different sizes
architectures and standards, instead of a challenge if these systems are managed with minimal hardware and/or storage
simplifying the representation of network locally and all the layers of the stack are units. We also need to be able to scale
assets. The challenges of bringing solutions managed separately. Also, as we move to efficiently depending on business needs and
without an adequate business model for NFV we are looking at having CI/CD loops reuse existing site and networking assets.
the exposure (such as operators taking for cost and efficiency reasons. The servers need to have a small
only individual approaches and a lack of The possibility to remotely monitor footprint, low power requirements and
business APIs) have caused service exposure performance and perform low-touch be NEBS3 compliant to cater for extreme
attempts to fail during the last 20 years. life-cycle management activities such as environmental conditions without cooling.
software installations, restarts and upgrades
Operational efficiency and support is crucial to avoid expensive site visits. Scaling
In NFV, the number of re-certification/ A software-defined infrastructure (SDI) Whilst most NFV-type deployments
validation cycles required dramatically capability can greatly improve the efficiency today consist of a few datacenters set up
increases due to the dynamic nature of and utilization of the compute and storage for redundancy and basic geographical
the overall software stack, including NFVI, resources.2 This should have the ability distribution, the amount of sites will
multi-vendor VNFs and support functions. to discover and manage all infrastructure dramatically increase with the introduction
Without automation, test organizations will elements such as configuration and control. of edge infrastructure. It is essential that the
become overwhelmed, causing reduced infrastructure management solution, the
agility and potentially lower quality. A related Data and regulatory compliance networking concept and the orchestration
challenge is how to locate faults and root With this new business need for flexible are capable of handling the scale of
causes in the entire stack of NFVI and VNFs. cloud solutions, data handling is just as managed sites.
The edge solution must have the important as computing, and regulatory
capability to locate and handle faults compliance is an urgent driver. Operators Solution integration can
quickly, in the entire stack and across a may have to navigate regulations restricting be lengthy and costly
multi-vendor and multi-tenant environment. the movement of data between geographical Investments in hardware and software are
In cloud-based systems, isolating an issue locations such as a country, a region or significant, the main costs likely being the
to a particular VNF is no longer sufficient. enterprise premises. integration and deployment of the solution.
The performance of each VNF is highly The early NFV adopters in the telco industry
dependent on the underlying infrastructure Infrastructure deployment cost and TCO embraced a best-of-breed approach that
layers, management layers, and the overlay Most current NFV infrastructures are built turned out to be far more time consuming
network which may comprise of multiple for larger datacenter deployments. The and costly than first anticipated.
vendors. It should be possible for the same reduced floor space and cooling capacity
operational model and team to manage in locations closer to the network edge
both the edge infrastructure and the place high demands on hardware
central NFV solution. efficiency and footprint.
2
Ericsson, SDI - software defined infrastructure (2018)
6 Ericsson | Edge computing and 5G
Overall solutions for edge computing Low TCO solution integration Edge Gravity
and distributed cloud As we have learned from several complex Ericsson Edge Gravity is a live global edge
The distributed cloud provides an execution and challenging NFVI integrations, a better computing network in partnership with
environment for cloud application approach than trying to integrate your close to 100 service providers. Gravity
optimization across multiple sites, managed own optimized components is to rely on started in 2016 with video streaming
as one solution and perceived as such by a pre-configured, pre-integrated and and content acceleration. The solution is
the applications. It is based on SDN, NFV tested, system-verified solution from designed to help service providers create a
and 3GPP edge computing with end-to-end an experienced partner. global network that is available to content
service orchestration and external exposure While pre-integrated, Ericsson’s Edge providers and application providers who
and management interfaces. This enables solution retains the flexibility to choose need more capacity and better performance
multi-access and multi-cloud capabilities hardware and is open for any type of VNF for their consumer applications.4
and unlocks networks as open platforms through our partner certification program.
for application innovations. A vast number of parameters, in the Open source
Ericsson NFVI includes Ericsson Cloud range of thousands, are pre-set to achieve Open source is starting to be seen by
SDN built on OpenDaylight SDI, including maximum performance. We use automated some to be the standard for composable
hardware for central, distributed and edge scripts to configure the solution to the cloud infrastructure.
sites (based on Open Compute Platform) customer’s requirements. This has resulted Open source communities like CNCF
and Cloud Execution Environment based in dramatic improvements in the and projects like Kubernetes provide key
on OpenStack and Ericsson Orchestrator time-to-market of network services, ingredients of an edge software stack.
(cloud orchestrator NFVO).3 with cost reductions up to 75 percent.
3
ricsson, Build cloud & NFV infrastructure (2019)
E
4
Ericsson, Edge gravity by Ericsson - Edge cloud platform (2019)
7 Ericsson | Edge computing and 5G
Ericsson Orchestrator
Open APIs and applications Industrialized End-to-end
architecture – 3PP and service assets FM/PM
Ericsson
SDN
Edge solution management tasks such as upgrades, restarts or addition End-to-end and multi-vendor
and orchestration /removal of devices will be developed with solution support
The Ericsson NFVI and Edge NFVI solutions a one-click approach in mind. As well as helping to remedy and restore
can be enhanced with the flexible Dynamic VNFs are managed by the Ericsson Cloud Ericsson products that may be experiencing
Orchestration solution to be able to manage Deployment Engine (ECDE) platform, which trouble, we can help to isolate where the
the distribution of network functions, such as is used for streamlining and automating NFV issue is occurring, whether it is an Ericsson
packet core or communication services, and delivery and operations. issue or not. Our AI-based tools, combined
the orchestration of network slicing in 5G. ECDE is designed to automate VNF with experienced engineers, will trace and
Orchestration is vital for distributed cloud, onboarding and validation workflows. It restore issues across multiple vendors.
providing end-to-end management provides capabilities to simplify setup as
of networking, cloud infrastructure, well as to track, operate and manage the
workloads and network services. VNF onboarding and validation processes
The solution is ONAP compatible. using self-serve web portals and
Our edge solutions enable management management dashboards.
of a large number of sites through an Ericsson also supports operators
end-to-end dashboard. Automated through their network and operational
workflows for common operational transformation to telco cloud.
Dynamic Orchestration
Mobile
of 1,000s of sites
broadband
APP
Ericsson NFVI VNF Single pane of glass
Enterprise
Service
Ericsson orchestration
Edge NFVI
Resource DC network
IoT Public
Ericsson orchestration automation
clouds
Edge NFVI UPF E-VNFM
UPF UPF
Media CCD CEE
Internet
CCD CCD CEE
SDN
Fixed wireless SDI SFF SDI – DC & CO
access /NEBS HW /NEBS HW SDI – DC HW
Unified management for:
–– Physical infrastructure
New –– Virtual machines
applications –– Containers