Edge computing from standard to actual infrastructure deployment and software development

Edge Computing:
from standard to actual infrastructure deployment
and software development
White Paper
October 2019
Authors: Dario Sabella, Andrew Alleman, Ellen Liao, Miltiadis Filippou, Zongrui Ding,
Leonardo Gomes Baltar, Srikathyayani Srikanteswara, Krishna Bhuyan, Ozgur Oyman,
Gershon Schatzberg, Neal Oliver, Ned Smith, Sharad D. Mishra, Purvi Thakkar.
Intel Confidential

Rev. 1.0, October 2019 2/41
Intel White Paper
1 Introduction
Edge Computing is widely recognized as a key technology, supporting innovative services for a wide
ecosystem of companies, ranging from operators, infrastructure owners, technology leaders, application
and content providers, innovators, startups, etc.
The advent of 5G systems has also increased the attention to Edge Computing [1][2], as this technology
supports many of the most challenging 5G use cases, especially those for which high end-to-end (E2E)
performances are required. In fact, due to the presence of processing platforms and application
endpoints in close proximity to end-users, Edge Computing offers a low latency environment and
significant gains in terms of network bandwidth, which are key benefits for the deployment of 5G
networks.
Edge computing can span a variety of network locations, form factors, and functions, as depicted in
Figure 1 below. Centralized computing is performed deeper into the network/cloud, with applications
addressing a large number of users, and edge platforms hosting multiple applications simultaneously. In
contrast, distributed computing [3] takes place in proximity to end users, with applications being more
attuned to specific endpoints and functions. Depending on the various deployment options considered,
edge applications can be characterized by spatial and temporal proximity to clients, real-time
responsiveness, interactivity, and mobility, and are well-suited for use cases such as industrial control,
video analytics, interactive (XR) media and healthcare use cases.
Figure 1 - Edge Computing spanning a variety of network locations, form factors, and functions
The area of Edge Computing (often termed after as “MEC” - Multi-access Edge Computing in this paper,
to emphasize its access-agnostic nature) falls broadly within the scope of various initiatives, ranging from
standardization bodies and industry groups and associations to open source communities and projects.

Rev. 1.0, October 2019 3/41
Intel White Paper
The aim of this white paper is to provide an overview of these initiatives, together with some Intel
guidelines, to help the reader to better understand the various challenges for a real deployment of Edge
Computing. The focus is to mainly address the needs of two stakeholder categories in the MEC
ecosystem: infrastructure owners (e.g., operators and cloud providers) and software developers (e.g.,
applications / content providers, innovators and startups). In fact, both stakeholder categories are key
to the success of MEC, and their engagement depends also on the way the various challenges are
addressed.
In particular, this paper aims to:
• provide an overview of standardization efforts, including initiatives from industry groups,
associations, open source communities and projects (e.g., Open Network Edge Services
Software (OpenNESS)), to highlight the key deployment findings from an infrastructure point of
view (especially for the MEC deployment in 5G systems), as well as to draw a coherent mapping
from a SW development perspective (e.g. including all components needed to build an edge
cloud, according to the current initiatives in the space);
• derive some deployment considerations related to Edge Computing (for the first category of
MEC stakeholders), with special emphasis on the recurrent question "Where is the edge?",
complemented by a comparative analysis of the different deployment options, together with a
suitable performance evaluation in few use cases of interest;
• describe an exemplary case study customized to use cases relevant to the automotive vertical
segment, thus, offering to the reader an E2E example involving topics such as orchestration,
security and onboarding. Such an example aims to constitute a useful reference for the software
developer (i.e., the other main category of MEC stakeholders), when it comes to answering to
the other recurrent question "How to use MEC from application development point of view?").
The overall market success of Edge Computing will depend, in fact, on how infrastructure owners will
be able to address all deployment aspects and also on how software communities will create a wide set
of applications and innovative services. Intel is committed to engaging the entire ecosystem, considering
standard solutions and open source projects, towards ensuring interoperability, while opening the
market to proprietary implementations and added-value propositions.

Rev. 1.0, October 2019 4/41
Intel White Paper
2 Edge Computing: standardization efforts,
industry groups and open source initiatives
Edge Computing incorporates the benefits of virtualization and cloud computing to enable high-
powered computing capabilities as close as possible to subscribers. The set of use cases supported by
this technology is very wide, and spans many vertical market segments, e.g., transportation, connected
cars, industrial automation, retail, healthcare and social assistance, media & entertainment, use cases
for smart cities and internet of things, and others. The market potential for Edge Computing is huge1
,
nevertheless, actual deployment of this technology depends on both its maturity and its specified
definition by relevant standardization bodies, industry groups and open source projects. For this reason,
the aim of this section is to present the main initiatives in this area from different angles and
perspectives, to analyze complementarities and synergies, and also potential overlaps and gaps. Finally,
a functional mapping depicts these activities by projecting them onto the various elements of the edge
architecture, as a possible guide for all stakeholders, i.e., both application developers and infrastructure
owners. In fact, to avoid market fragmentation, a coherent approach should be adopted by
stakeholders, to align implementations and enable interoperability, also by reducing costs.
2.1 The standardization landscape
Standardization in the area of Edge Computing started in ETSI at the end of 2014 with the creation of
the Industry Specification Group (ISG) on MEC [4]. Currently, ETSI ISG MEC is still the only international
standard available in the technology field, however, new emerging initiatives have started in 3GPP
aiming at the integration of MEC in 5G systems. Other Standards Developing Organizations (SDOs) have
also started working around Edge Computing from different perspectives/angles (e.g., the Internet
Engineering Task Force - IETF, for security aspects). The following overview is limited to current
standardization efforts, and will be complemented by subsequent sections on other initiatives, such as
industry groups and open source initiatives.
2.1.1 ETSI ISG MEC
According to the ETSI ISG MEC definition, MEC offers to application developers and content providers
cloud-computing capabilities and an Information Technology (IT) service environment at the edge of the
network. This definition is quite compact and summarizes the main technologies introduced by ETSI for
Edge Computing: cloud-computing capabilities and an IT service environment. In fact, MEC does not only
introduce cloud computing at the edge of the network, but also offers the capability to expose edge
services to application developers.
Following the developments of the ETSI MEC standard, MEC provides an environment in proximity to
the end user, ultra-low latency, high bandwidth, and access to added value information through MEC
1
https://www.ericsson.com/en/blog/2018/9/edge-computing-successa-distributed-cloud-approach

Rev. 1.0, October 2019 5/41
Intel White Paper
Application Programming Interfaces (APIs), such as real-time access to access network, context
information, location awareness and others. MEC will permit to open the cloud infrastructure to
operators, service providers and third parties, i.e., application developers and content providers, helping
to meet the demanding Quality-of-Service (QoS) requirements of new 5G systems. Due to its access-
agnostic nature, MEC guarantees a smoother deployment, independent of the underlying Radio Access
Network (RAN) (e.g., Long Term Evolution (LTE), 5G New Radio (NR), or other non-3GPP radio access).
Figure 2 - ETSI MEC Architecture [5]
The MEC system reference architecture is defined in ETSI ISG MEC Group Specification (GS) MEC003 [5]
and it is visualized in Figure 2. In particular, a MEC system incorporates two levels: the MEC host level
and MEC system Level. The former consists of the MEC host, the MEC platform and the Virtualization
Infrastructure Manager (VIM), while, the latter is composed by the MEC Orchestrator, the Operations
Support System (OSS) and the user application (or, “User App”) Lifecycle Management (LCM) proxy,
which acts as entry point for the requests coming from a device application (“Device App”) toward the
MEC system. The MEC host facilitates MEC applications (“MEC Apps”), offering a virtualization
infrastructure that provides computation, storage and network resources, as well as a set of
fundamental functionalities (“MEC services”) required to execute MEC Apps, known as the MEC
platform. The MEC platform is, thus, a key component in the MEC system, and is characterized by the
following main functionalities: enabling applications to discover, advertise and consume MEC services,
and providing the virtualization infrastructure with a set of rules for the user traffic forwarding plane
(such rules are based on the policies associated to the MEC applications). The MEC platform configures
also local Domain Name System (DNS) handling, which assists the user traffic in reaching the desired
MEC application, and communicates with other peer MEC platforms via the Mp3 interface, which allows
peer platform clustering.
MECSystemLevelMECHostLevel
MEC platform
MEC Host
Virtualization Infrastructure
Data plane
Device
app
VIM
Other
MEC
platform
MEC platform manager
Multi-access edge
orchestrator
Mm3
Mm1
Mx2
Mm8
Mm9
Mp2
Mp1
MEC
app
MEC
app
Service
Mm2
Mm5
Trwaffic
rules
control
MEC
service
Service registry
DNS
handling
MEC
platform
element
mgmt
CFS
portal
Mx1
Mp3
Other
MEC Host
MEC
app
rules &
reqts
mgmt
MEC
app
lifecycle
mgmt
Mm7
User
app
LCM
proxy
Operations Support System
Mm4
Mm6
Reference Point names:
Mp: MEC platform functionality
Mm: management
Mx: connections to external entities

Rev. 1.0, October 2019 6/41
Intel White Paper
ETSI ISG MEC specifies also a set of RESTful APIs [6] as standardized interfaces, which can be used by
application developers to access Radio Network Information (RNI API), location information (Location
API), and other types of data pre-processed either by the MEC platform, or, by instantiated MEC
applications. In particular, thanks to the RNI API, context information from the RAN can be provided to
user level applications or other services for network performance and QoS improvements. As a further
example, with regards to Vehicular-to-Everything (V2X) use cases, ETSI ISG MEC is introducing a “MEC
V2X API” to assist the MEC system in exposing a set of information to applications, which permit
developers, car Original Equipment Manufacturers (OEMs) and their suppliers to implement Intelligent
Transportation System (ITS) services in an interoperable way, across different access networks, owned
and managed by different Mobile Network Operators (MNOs) and vendors.
2.1.2 3GPP support for Edge Computing
When it comes to standardization activities in 3GPP, Working Group (WG) SA1 is the working group to
standardize 3GPP service requirements which are regarded as "stage 1" (i.e., high level requirements)
and would trigger the related studies in downstream working groups, including SA2 for system
architecture, SA5 for network management, SA3 for security aspects, and SA6 for application
architecture.
More specifically, with regards to 3GPP support for Edge Computing, various services characterized by
extreme Key Performance Indicator (KPI) values needed to support Ultra Reliable, Low Latency
Communication (URLLC) have given Edge computing an important role in a 5G system. In particular, in
3GPP Rel-15 TS 22.261, two service requirements set the tone for Edge Computing in support of 5G
services:
• Resource efficiency: To meet the various KPIs defined for 5G, the 5G network shall support
mechanisms for minimizing user plane resource utilization by including in-network caching and
application in a Service Hosting Environment closer to the end user.
• Efficient user plane: Based on operator policy, the 5G network shall be able to maintain user
experience (e.g., QoS, QoE) and support routing of data traffic between a UE attached to the
network and an application in a Service Hosting Environment for specific services, and modifying
the path as needed when the UE moves or application changes location during an active
communication.
As a note, a “Hosted Service” is defined as a service containing the operator's own application(s) and/or
trusted 3rd
party application(s) in the Service Hosting Environment, which can be accessed by the user.
A “Service Hosting Environment” is defined as the environment, located inside the 5G network and fully
controlled by the operator, where Hosted Services are offered. In other words, the presence of this
Service Hosting Environment is a first sign of the requirements for Edge Computing in 5G systems, where
SA1 WG defined the assumptions providing the landing requirements that are currently going on in the
other 3GPP WGs, for more specific standardization (e.g. SA2, SA5 and SA6).

Rev. 1.0, October 2019 7/41
Intel White Paper
2.1.2.1 Release 15 and 16
In SA2, the following definition for Edge Computing has been provided in 3GPP TS 23.501 [7] since Rel-
15:
"A concept that enables operator and 3rd
party services to be hosted close to the UE's access
point of attachment, to achieve an efficient service delivery through the reduced end-to-end
latency and load on the transport network."
To enable the Edge Computing feature in 5G system architecture, the design of the 5G system
architecture requires several key enablers2
:
1. Applicable to both non-roaming and Local-breakout roaming cases;
2. Flexible placement of User Plane Function (UPF): concurrent access to two (e.g., local and
central) data networks;
3. Support of Multi-homed Protocol Data Unit (PDU) Sessions: using either “Uplink Classifier” (UL
CL) or multi-homed IPv6 (Branching point).
4. The Session and Service Continuity (SSC) mode 3: a connection through a new PDU Session
Anchor point is established before the previous connection is terminated to allow for better
service continuity and the IP address (IPv4 or IPv6) is not preserved in this mode when the PDU
Session Anchor changes.
5. Support of Local Area Data Network (LADN)
6. Application Function (AF) influence on traffic routing: the AF can make a request to influence
traffic routing for a target UE by indicating traffic description (Data Network Name - DNN, Single
Network Slice Selection Assistance Information - S-NSSAI, Application IDs, traffic filters, etc.),
traffic routing requirement per Data Network Access Identifier (DNAI), etc., or request to
subscribe the notification of user plane management events.
According to the definitions introduced by SA2 in TS 23.501 [7], we could also identify a sort of mapping
between ETSI MEC entities and 3GPP entities, in order to understand how MEC system can be deployed
in a 5G network:
• User Plane Function(s) (UPFs) “handle the user plane path of PDU sessions. […]. A UPF that
provides the interface to a Data Network supports the functionality of a PDU session anchor.”
o According to this definition, the logical UPF in the 3GPP architecture may correspond to
some functionalities defined in ETSI for the MEC Data Plane (ETSI GS MEC 003);
• Application Function (AF) in 3GPP contains the following high-level functionalities: application
influence on traffic routing, access network capability exposure, interact with the policy
framework for policy control;
2
However, until Rel-16, the following features are not supported and are out of scope of SA2: Discovery of target application
server, Application Function (AF) triggers mechanism for traffic redirection and application context transfer from a Source
Application Server (S-AS) to a Target Application Server (T-AS), and AF changes.

Rev. 1.0, October 2019 8/41
Intel White Paper
o According to this definition, the logical AF in the 3GPP architecture may correspond to
some functionalities defined in ETSI for the MEC Platform (ETSI GS MEC 003);
• Finally, the 5G system architecture introduces the Data Network (DN) receiving user plane traffic
from the UPF. Local DN deployments can be the perfect examples of environments hosting MEC
applications, in contrast to a remote DN (or, a central DN).
In summary, Figure 3 shows an example of possible MEC mapping to the 5G system architecture and the
related correspondence of logical entities introduced by the two standard bodies.
Figure 3 - Example of MEC mapping to the 5G system architecture
A more detailed example, depicted below in Figure 4, could further help to understand how to deploy
MEC in 5G systems under the assumption of considering a Network Function Virtualization (NFV)-based
network.
Based on the ETSI NFV framework, to consider the coordination between UPF and MEC application
deployment/management and to support 5G edge computing, the UPF Virtualized Network Function
(VNF) and the Edge Computing VNF (i.e., the MEC App) should be deployed at locations where the end-
to-end QoS requirements, including 5G QoS (i.e. on UE – UPF connection) and QoS for the N6 interface,
would be met (see Local DN below). When MEC application(s) are available/installed at an Edge
Computing VNF, a trusted AF (i.e., the MEC Platform) can request the Session Management Function
(SMF) via the Policy Control Function (PCF) to influence UPF (re)selection and traffic routing. All MEC
entities can be implemented as VNFs, and run in the same NFV Infrastructure Point-of-Presence (NFVI-
PoP), as in the example below, or in different locations.
UE (R)AN UPF
AF
AMF SMF
PCF UDM
DN
NRFNEF
N3
N2 N4
AUSF
Nausf Namf Nsmf
NpcfNnrfNnef Nudm Naf
N6
MEC Host
MEC platform
MEC Data Plane
MEC apps

Rev. 1.0, October 2019 9/41
Intel White Paper
Figure 4 - Example of MEC deployment in an NFV-based 5G network
SA5 is responsible for network deployment and management aspects in 3GPP network (management
plane). Currently in this WG, a Rel-16 study item [8] is studying the coordination between 3GPP
management system and Non-3GPP management system including the ETSI NFV management &
orchestration, and Edge Computing management systems.
2.1.2.2 Release 17
Standardization efforts are currently ongoing for the Rel.17 definition of the specifications. Based on
existing solutions defined in Rel-15 and Rel-16, a new study [9] in Rel-17 has been agreed in SA2 with
several objectives, including the following:
1. Enable support for seamless changes of the Application Server serving the UE within an Edge
Computing Environment (ECE) or between the ECE and the cloud, including discovery of the IP
address of the Application Server in the ECE;
2. Provide deployment guidelines for typical Edge Computing use cases including e.g., URLLC, V2X,
Augmented Reality (AR)/Virtual Reality (VR)/ Extended Reality (XR), Unmanned Aircraft System
(UAS), integration of satellite access in 5G (5GSAT), and Content Delivery Network (CDN).
Moreover, SA6 WG is currently working on a Rel-17 study of the application architecture for enabling
edge applications [10]. Due to many concurrent activities in 3GPP, and the existence of the ETSI MEC
standard, some coordination within 3GPP and between ETSI and 3GPP is envisaged, in order to
guarantee coherence among the various specifications, and thus avoid market fragmentation that could,
instead, increase the costs related to the adoption of Edge Computing. Here below are possible areas of
coordination, in order to ensure an alignment on the various activities in 3GPP and ETSI:
• SA2 and SA6: coordinate respectively system architecture and application architecture (e.g.,
definition of different entities, edge network providers, edge service subscription, relationship
between AF and Edge Enabler Server, APIs via network exposure functions, e.g., location
information of the UE, UE identifier, etc.);
• SA6: Coordinate between MEC-related definition and ETSI MEC (in progress);
NFVI
MEC Data Plane (3GPP UPF)
MEC platform VNF
(3GPP AF)
MEC
API
MEC
API
APIPrinciples
(MEC009)
MEC
App1
(VNF)
MEC
App2
(VNF)
Mp1
MEC011
App Enablement
UE 3GPP UPF
N9
Central
DN
Local DN
Application
Server
3GPP
SMF
3GPP
PCF
N4
N7
N5
Mp2
N6
3GPP
NEF
3GPP
vRAN
(VNF)
N3
N33

Rev. 1.0, October 2019 10/41
Intel White Paper
• SA5: Coordinate with ETSI MEC / ETSI NFV and support the management of 3GPP Edge
Computing with alignment to the outcome of SA2, and SA6 in Rel-17;
• SA3: address the security aspect of mobile edge computing, e.g., in EDGE-1, EDGE-2, EDGE-3,
and EDGE-4;
• All groups: Need more inputs on practical deployment use cases in areas of URLLC, V2X,
AR/VR/XR, UAS, 5GSAT, and CDN.
2.1.3 Other SDOs: focus on security aspects
Security decisions are, in principle, very complex, in part due to the rich stakeholder ecosystem of Edge
Computing. Different stakeholders bring differing expertise, capabilities and resources. A user workload
may execute in an environment involving multiple stakeholders, each supplying various aspects of a
deployment solution. Security decisions are linked to an understanding of which stakeholders are
authorized to supply which expertise, capability and resource. Edge orchestration and Service Level
Agreements (SLA) provide important context for security decisions by identifying stakeholders and
contracted (expected) support.
The American Council for Technology-Industry Advisory Council (ACT-IAC) published a report3
that
outlines a new approach aimed at cloud and edge security, known as “Zero Trust (ZT)”, which has the
potential to substantially change and improve an organization’s ability to protect their systems and data.
ZT is a security concept anchored on the principle that organizations need to proactively control all
interactions between people, data, and information systems to reduce security risks to acceptable
levels. The ACT-IAC report defines five stages of ZT, the first three of which focus on this concept of
identifying and assessing assets – data, users and devices. The latter two focus on continuous security
risk management:
1. Establish User Trust
2. Gain Visibility into Devices & Activity
3. Ensure Device Trustworthiness
4. Enforce Adaptive Policies
5. Zero Trust (end goal)
Edge standards will evolve to embrace these security imperatives. Trusted computing technology
addresses the device aspects of “Zero Trust” stages and includes capabilities such as hardware “Root of
Trust” (RoT) designs that establish attestable cryptographic identities for RoT components. Even simple
Internet-of-Things (IoT) platforms may consist of multiple components (e.g. CPU, IO controllers,
memory, network interface, sensor/actuator), one or more of which could contain a hardware root of
trust. More sophisticated platforms may have various forms of acceleration components (e.g., Graphic
Processing Unit (GPU), machine learning processors, Field Programmable Gate Array (FPGA)).
Datacenters construct racks containing reconfigurable compute clusters, storage pools and network
equipment. Every configurable component could contain a RoT element that helps realize “zero trust”.
The IETF security area defines standards for interoperable attestation and Trusted Execution
Environments (TEEs). A TEE is a hardened environment for running very security sensitive workloads.
3 https://www.actiac.org/system/files/ACT-IAC%20Zero%20Trust%20Project%20Report%2004182019.pdf

Rev. 1.0, October 2019 11/41
Intel White Paper
Often, they are a subset of a traditional workload consisting of the most security-relevant operations
such as cryptography, user authentication, biometrics, privacy, access control and rights management.
Most TEE technology is built using hardware RoT building blocks.
ZT security requires assessment of the various computing environments that host edge frameworks and
services known as ‘attestation’. Before an edge orchestration entity schedules workload execution on a
container cluster or a mesh of serverless nodes, the security properties of the intended computing
resources are assessed to determine if the sensitivity of workload data and processing matches the
protection capability of assigned resources.
2.1.3.1 IETF Trusted Computing Standards
The IETF Remote Attestation Procedures (RATS) working group defines interoperable attestation
architecture, information models, trustworthiness properties and protocol bindings. Existing and
emerging standards aimed at Internet connected Cloud, Edge and IoT systems will be able to support
attestation capabilities. Protocols, thus enabled, will become part of a pervasive infrastructure for
assessing trust.
The IETF Trusted Execution Environment Provisioning (TEEP) working group defines protocols and
formats for interoperable TEE provisioning using Open Trust Protocol (OTrP). OTrP requires support for
attestation to ensure provisioning steps are performed between trustworthy endpoints.
The IETF Software Update for Internet of Things (SUIT) working group defines architecture, information
model and data structures for the secure update of firmware and software of IoT and similar endpoints.
Secure software update depends on update procedures that are protected by a trustworthy
environment. In a ZT scenario, the software update server should verify the device receiving the update
is the correct device and that the update was applied correctly. Attestation allows the update server to
make this assessment even when the device being updated may be using compromised software.
2.1.3.2 DMTF
DMTF (formerly known as the Distributed Management Task Force) creates open manageability
standards spanning diverse emerging and traditional IT infrastructures including cloud, virtualization,
network, servers and storage. DMTF’s Redfish® is a standard designed to deliver simple and secure
management for converged, hybrid IT and the Software Defined Data Center (SDDC). Both human
readable and machine capable, Redfish® leverages common Internet and web services standards to
expose information directly to the modern tool chain.
The Security Protocol and Data Model (SPDM) Specification (DSP0274) provides message exchange,
sequence diagrams, message formats, and other relevant semantics for authentication, firmware
measurement, and certificate management. This specification for additional security defined by SPDM
has a goal of aligning component authentication and integrity objects across the industry and is being
designed to be referenced by other standards organizations.
Peripherals and devices implementing SPDM have the ability to report firmware measurements securely
using embedded device identities. When combined in a comprehensive network-wide attestation
capability, an orchestrator, user, scheduler, regional operations center, gateway, concentrator, cluster
manager, compliance auditor or other parties may be able to effectively apply “zero trust” objectives to
edge deployments.

Rev. 1.0, October 2019 12/41
Intel White Paper
2.1.3.3 Trusted Computing Group
The primary focus of the Trusted Computing Group (TCG, https://trustedcomputinggroup.org), as the
name implies, is trusted computing. It standardizes RoT mechanisms such as the Trusted Platform
Module (TPM) and Device Identity Composition Engine (DICE). The TCG also defines APIs for interacting
with trusted computing modules and trusted computing infrastructure.
The TCG Embedded System Workgroup (ESWG) and DICE work group define RoT capabilities and profiles
that meet the special needs of constrained environments. Interestingly, the RoT building block
technologies are relevant in the datacenter and edge computing environments since all are cost
conscious but require strong security.
The Infrastructure Workgroup (IWG) defines trusted computing infrastructure and architecture for
managing the lifecycle of a trusted device. Trustworthy device identity often involves the device
manufacturer and related supply chain entities that create the RoT, embed attestable identities and
certify them in a way that supports automated appraisal.
Trusted computing lifecycle also involves assembling databases of components, firmware and system
software that is designed to be secure. Before these components begin to function as a cohesive
solution, a security decision is required that determines whether each component is authorized to
participate in the solution. Attestation (described more in detail in the following section) is a security
capability that addresses this need, attestation verifiers use this data to compare with actual reported
values that may indicate the presence of untrustworthy software or an invalid configuration.
2.1.3.4 Attestation and trust in edge environments
Attestation is a security capability that allows stakeholders to evaluate security conditions that
determine whether each component of a solution is authorized to participate in that solution. Different
stakeholders may request attestation of a device in order to be assured that an agreed upon
configuration is indeed in place. Once stakeholder appraisals are satisfied, workload executions can
commence.
Security enforcement may need to be utilized by many stakeholders which requires use of enforcement
points not under the direct control of a particular stakeholder. Consequently, Edge Computing platforms
must have hardened and trusted execution capabilities that are commonly acceptable to the other
stakeholders supplying a portion of the solution.
Attestation is a simple conversation (See Figure 5) between edge nodes at a point before exchanging
data or performing important functions. It might make sense for an orchestrator to verify the Function-
as-a-Service (FaaS) nodes designated for workload execution or for a MEC host to verify the 5G
equipment it depends on for connectivity before blindly using it.

Rev. 1.0, October 2019 13/41
Intel White Paper
Figure 5 - A simple attestation conversation
Attestation and other security functionality rely on a RoT environment (See Figure 6) that is hardened
to resist a variety of software and hardware attacks. It is also a constrained environment because
complexity itself is a security concern. Vulnerabilities may lurk undetected simply because it is too
difficult to exercise all possible device states and determine if it has undesirable security properties. A
RoT environment may be reduced to only the essential functions as a way to improve its trustworthiness.
In some cases, there may be multiple RoT environments in the same device that cooperate to supply a
more complete set of trusted functionalities. For example, the TPM defined by TCG is a root-of-trust for
storage and reporting (attestation), but does not provide a RoT for securely booting the device.
Figure 6 - A RoT environment in an edge device
Attestation reveals the security and trustworthiness attributes of the edge device to a verifier or other
relying party so that the security risk can be assessed according to security best practices and good
manufacturing processes. Trust in the attestation functionality is another layer of functionality that
needs to be hardened. The TCG Device Identity Composition Engine (DICE) is a root-of-trust that
anticipates layering of trustworthy software so that complexity of trusted layers can be managed.

Rev. 1.0, October 2019 14/41
Intel White Paper
2.2 Industry groups and open source initiatives
Industrial associations, as non-standard bodies triggered by vertical market segments, are key players
able to influence the ecosystem of Edge Computing adopters. In addition, open source projects and
related initiatives offer complementary tools for the acceleration of implementation of edge-based
solutions. Some key representatives of these groups of stakeholders are described in the following.
2.2.1 Vertical associations on automotive
The automotive market is one of the key vertical segments driving the introduction of Edge Computing.
Figure 7 below depicts a typical automotive scenario, where, multiple vehicles, potentially belonging to
different car OEMs and other devices (e.g., smartphones and other Vulnerable Road Users - VRUs) are
connected to infrastructure (Road Side Units - RSUs) and a cellular network (RAN). The client application
instances are generically able to communicate with server applications, i.e., at edge clouds, remote
clouds, and/or OEM/private clouds).
Figure 7 - Edge Computing support for automotive scenarios
Two main industry groups are relevant in this segment: 5GAA (5G Automotive Association,
www.5gaa.org) and AECC (Automotive Edge Computing Consortium, https://aecc.org).
2.2.1.1 5GAA
Created in September 2016, 5GAA brings together automotive, technology and telecommunications
companies to work closely together to develop end-to-end connectivity solutions for future mobility,
smart cities and intelligent transportation services. The automotive industry comes with the vehicle
platforms, hardware and software solutions and the telecommunications industry with connectivity and
network systems, devices and technologies.
Several Edge Computing related activities have taken place in 5GAA. In December 2017, 5GAA published
the white paper4
“Toward fully connected vehicles: Edge computing for advanced automotive
4
http://5gaa.org/wp-content/uploads/2017/12/5GAA_T-170219-whitepaper-EdgeComputing_5GAA.pdf
RAN +
Edge Cloud
in-car Cloud
UE
www
Remote
App server
OEM cloud /
private Cloud
MEC
App Remote
App
Client
App
VRUs
Car #2
MEC
App
Car #1
Other cars
MEC/Remote
App
RSU
and/or
RAN

Rev. 1.0, October 2019 15/41
Intel White Paper
communications” under the leadership of Intel. In February 2018 it organized the 5GAA Open
Workshop5
on “Edge Computing and V2X”, where Intel both coordinated the effort and participated to
the event. At the Mobile World Congress 2018, 5GAA announced Edge Computing as one of the key
supporting technologies for many V2X services for connected vehicles and for autonomous driving. In
May 2018 a 5GAA internal paper titled “Towards a Future 5G Vision 5GAA Strategy for Edge Computing”
was finalized under the leadership of Intel. Several work items in 5GAA have relationship to Edge
Computing and in May 2019 a work item dedicated to Edge Computing was kicked-off, as well. The
project has targets to demonstrate the potential and added value of Edge Computing for automotive
services in a multi-MNO, multi-vendor and multi-OEM environment, whereby Intel has the overall lead
and also sub-tasks responsibilities.
2.2.1.2 AECC
Automobile Edge Computing Consortium (AECC) was formed in August 2018 initially by Toyota, Intel and
Ericsson. Currently, the members include companies across different industries and still growing. The
mission of AECC is to help automotive manufacturers, OEMs and the supply chain to accommodate
growing requirements by evolving current network architectures and computing infrastructures. Work
by the Consortium helps industry stakeholders to set the new route for connected cars by increasing
network and computing capacity. The shared strategic goal accommodates automotive big data smartly
between vehicles and the cloud by using Edge Computing and more efficient system design.
Edge Computing is believed to be a crucial technology to enable service scenarios such as intelligent
driving, high-definition maps, V2Cloud cruise assistance, extended services including Mobility as a
Service (MaaS), along with finance and insurance. A concept of “distributed computing on localized
networks” was introduced in an AECC white paper6
to address the problem of big data for automobile
and optimize the current mobile communication network architectures and cloud computing systems.
This concept is characterized by three key aspects: localized network, distributed computing and local
data integration platform. The technical details including the key issues and solutions identified for AECC
system was published in AECC TR in Sep, 20197
.
2.2.2 Vertical associations on industrial automation
The industrial automation market is another key vertical segment driving the introduction of Edge
Computing. As illustrated in Figure 8, focusing on manufacturing scenarios (also, known as “Factory of
the Future” – FoF), Edge Computing is expected to bring significant gains when it comes to performance
metrics such as production line efficiency via reliable, data analytics-based process monitoring, as well
as process control through the design of controllers and actuation policies based on low-latency sensory
information. The proximity of Edge Computing hosts to the system’s end points is a decisive feature
towards (i) enhancing system (“plant”) stability by improving Communication Service Reliability (CSR)
5
http://5gaa.org/news/c-v2x-edge-computing-the-winning-technologies-for-connected-vehicles-and-autonomous-driving/
6
https://aecc.org/wp-content/uploads/2018/02/AECC_White_Paper.pdf
7
https://aecc.org/wp-content/uploads/2019/09/AECC_WG2_TR_v1.0.2.pdf

Rev. 1.0, October 2019 16/41
Intel White Paper
and (ii) addressing the problem of accumulating and processing an explosive amount of data at a single
remote (cloud) server.
Figure 8 - Edge Computing support for industrial automation scenarios
Two main industry groups are relevant in this segment: 5G-ACIA (5G Alliance for Connected Industries
and Automation, https://www.5g-acia.org) and IIC (Industrial Internet Consortium,
https://www.iiconsortium.org).
2.2.2.1 5G-ACIA
The performance improvements delivered by 5G communications, as compared to previous generations
of wireless communications, are for the benefit of not only the (well established) enhanced Mobile
Broadband (eMBB) use cases, but also a newly introduced set of use cases grouped into two additional
service types: URLLC, as well as massive Machine Type Communications (mMTC – also termed after as
“massive IoT”). Driven by the latest technical developments and regarding the two latter service types,
the industrial automation vertical segment has shown increased interest in improving the efficiency of
manufacturing processes by replacing complex and costly wired communications infrastructure with 5G
wireless links. To turn the goal of “smart factories” into reality, the Information & Communication
Technology (ICT) and the Operational Technology (OT) stakeholder groups need to closely collaborate
in order for the OT service particularities and performance requirements to be fully understood by the
ICT ecosystem, in order for the latter to steer efforts towards fulfilling the needs of the automation
industry. To accomplish that, the 5G-ACIA was established as a global forum of discussing and evaluating
relevant technical, business and regulatory proposals concentrating on the vertical segment of industrial
automation8
.
8
https://www.5g-
acia.org/fileadmin/5GACIA/Publikationen/Whitepaper_5G_for_Connected_Industries_and_Automation/WP_5G_for_Connec
ted_Industries_and_Automation_Download_19.03.19.pdf

Rev. 1.0, October 2019 17/41
Intel White Paper
2.2.2.2 IIC
The Industrial Internet Consortium (IIC) was founded in March 2014 to bring together the organizations
and technologies necessary to accelerate the growth of the industrial internet by identifying,
assembling, testing and promoting best practices. More recently, OpenFog Consortium activities and
members have been acquired into the IIC organization so that IIC can continue the progress OpenFog
has made toward accelerating the adoption of fog computing.
The relevance of Edge Computing to IIC is evident from the recently published IIC architecture [11],
which is divided in three tiers: edge tier, platform tier and enterprise tier. The edge tier collects data
from the edge nodes, using the proximity network. The platform tier receives, processes and forwards
control commands from the enterprise tier to the edge tier. It consolidates processes and analyzes data
flows from the edge tier and other tiers. It provides management functions for devices and assets. It
also offers non-domain specific services such as data query and analytics. The enterprise tier
implements domain-specific applications, decision support systems and provides interfaces to end-users
including operation specialists. The enterprise tier receives data flows from the edge and platform tier.
It also issues control commands to the platform tier and edge tier.
Figure 9 - Three-tier industrial IoT system architecture (Source: IIC [11])
Many benefits are also identified by IIC on Edge Computing:
• Latency: The edge can provide latency in milliseconds, while multiple hops and long transmission
distances to the platform tier is in the 50-150 ms range. Latency to centralized data centers and
the public cloud is even greater.
• High throughput: The throughput available to the user from the edge, served via cached or
locally generated content, can be orders of magnitude greater than from a core data center.
• Data reduction: By running applications (such as Machine Learning (ML)/ Deep Learning (DL)
ones) at the edge, operators and application vendors can substantially cut down the amount of

Rev. 1.0, October 2019 18/41
Intel White Paper
data that has to be sent upstream. This cuts costs and allows for other applications to transfer
data.
• Isolation: A number of environments are not always connected to the Internet over high speed
links. The edge is able to provide services during periods of degraded or lost connections.
• Compliance: Edge applications can help with privacy or data location laws.
• Accuracy: Combine local results at the edge with global results at the platform to get a more
comprehensive view of the data in real-time.
Edge Computing has been implemented in a variety of Industrial IoT (IIoT) deployments; however, the
need to modernize edge architectures became apparent with the emergence of cloud computing. The
rapid decline in processor and memory cost enables more advanced decision-logic closer to where the
data is created, at the edge. The industry has learned that a “one-size-fits-all” approach has never been
adequate for IIoT. It is also true that the IIoT system designers always know where the edge boundary
is, and what devices in the system can be categorized as edge devices. System designers are challenged
to implement an architecture that is managed, orchestrated, trustworthy and secure.
2.2.3 Vertical associations on AR / VR
The ability to leverage cloud computing and Edge Computing can be instrumental in reducing
computational cost and complexity for the client devices when it comes to handling the high quality and
interactivity performance demands and corresponding processing requirements associated with
immersive media, including workloads such as decoding, rendering, graphics, stitching, encoding,
transcoding, caching, etc., which may all be performed over the cloud and/or edge. Low latency, high
throughput networks such as 5G could allow instantaneous access to remotely stored data, while
offering a local computing experience similar to a datacenter-based system. Such a high capacity
network with low latency characteristics also enables responsive interactive feedback, real-time cloud-
based perception, rendering, and real-time delivery of the display content.
With such an edge-based approach, it is sufficient to use low-cost thin client devices with minimal built-
in functions. For VR and AR, these include the display screen, speakers for output, vision positioning and
hand-controller sensors for input. The thin client simply uses the network to pass inputs to processing
functions at the cloud or edge, and receives the necessary images to display.
Cloud VR/AR/XR brings together the significant advances in cloud computing and networks with a high
degree of interactivity to provide high quality experiences to those who were previously priced out of
immersive technologies. This approach may, thus, help significantly increase the number of VR client
devices sold, and create major revenue opportunities from increased consumption VR/AR services for
content providers and operators.
A currently ongoing GSMA Cloud VR/AR initiative [12] aims to address this opportunity from the
perspective of mobile operators, who aim at leveraging their infrastructure to monetize VR/AR/XR
services, cooperating with content providers. MPEG’s Network-Based Media Processing (NBMP)
specification ISO/IEC 23090-8 [13], aims to specify metadata formats and APIs for intelligent edge media
processing as an enabler to offload compute-intensive media processing to the edge. For instance, such
a capability can be relevant to 3GPP’s Framework for Live Uplink Streaming (FLUS) service in TS 26.238,
in which videos captured by one or multiple omnidirectional cameras (without built-in stitching) may be
sent separately to the cloud or edge via an uplink connection, where they may be stitched to create 360o

Rev. 1.0, October 2019 19/41
Intel White Paper
videos and then encoded and encapsulated for live distribution. Edge enhancements enabled by 5G also
help in improving viewport-dependent 360o
video delivery, where high quality viewport-specific video
data (e.g., tiles) corresponding to portions of the content for different Fields of View (FoVs) at various
quality levels may be cached at the edge and delivered to the client device with very low latency based
on the user’s FoV information, as described in 3GPP TS 26.118 [14], MPEG’s ISO/IEC 23090-2 [15] as well
as in VR Industry Forum (VRIF) Guidelines [16]. Finally, VRIF is currently working on delivering broader
interoperability guidelines and implementation best practices around cloud VR/AR/XR with Edge
Computing, also considering the distribution of volumetric video content.
2.2.4 LF edge Akraino
Akraino Edge Stack is a project created in February 2018 by the Linux Foundation, with the aim to create
an open source software stack to improve the state of edge cloud infrastructure for carrier, provider,
and IoT networks. This project will offer users new levels of flexibility to scale edge cloud services quickly,
to maximize the applications or subscribers supported on each server, and to help ensure the reliability
of systems that must be up at all times.
This open source software stack intends to provide critical infrastructure to: enable line speed
processing, enable high throughput, reduce latency, improve availability, lower operational overhead,
provide scalability, address security needs and improve fault management.
The main motivation of Akraino Edge Stack is that there are many open source projects that provide
component capabilities required for Edge Computing. However, there is no holistic solution to address
the need for fully integrated edge infrastructure. Akraino Edge Stack, a Linux Foundation project
initiated by AT&T and Intel, intends to develop a fully integrated edge infrastructure solution, and the
project is completely focused towards Edge Computing.
• AT&T's seed code will enable carrier-scale Edge Computing applications to run in virtual
machines and containers. AT&T’s contributions, which will include support for 5G, IoT, and
other networking edge services will enhance reliability and enable high performance.
• Intel upstreamed its Wind River Titanium Cloud portfolio of technologies to open source in
support of additional blueprints in Akraino. Intel is also a premium member of LF Edge and
Akraino community, and is actively participating in various Akraino Blueprints.
Akraino Edge Stack community is focused on edge APIs, middleware, Software Development Kits (SDKs)
and will allow for cross-platform interoperability with 3rd
party clouds. The Edge Stack will also
enable the development of edge applications and create an application with the VNF ecosystem. The
Akraino Edge Stack is intended to support any type of access methodologies such as Wireless (4G/LTE,
5G), wireline, Wi-Fi, etc.
Akraino is also a complementary open source project, and interfaces with existing projects namely
Acumos AI, Airship, Ceph, DANOS, EdgeX Foundry, Kubernetes, LF Networking, ONAP, OpenStack, and
StarlingX. The Akraino Edge Stack is a collection of multiple blueprints, which are defined as the
declarative configuration of entire stack i.e., Cloud platform, API, and Applications. The intent of Akraino
Edge Stack is to support VM, container and bare metal workloads. Section 4 contains an example of
Akraino blueprint, customized for the automotive case study.

Rev. 1.0, October 2019 20/41
Intel White Paper
2.2.5 OpenNESS
Open Network Edge Services Software (OpenNESS) is an open source software toolkit that enables the
deployment of edge services on diverse platform and access technologies. It is foundational software,
because, “out of the box”, it supports the deployment of edge services in a network, but can also be
extended in functionality into a commercial platform, or re-used to add capability to existing edge
platforms. The primary goal of OpenNESS is to reduce the “deployment impedance” experienced by
network operators, Independent Hardware Vendors (IHVs), and Independent Software Vendors (ISVs)
in deploying edge services. OpenNESS provides a variety of features to achieve this goal:
• it exposes platform and hardware diversity to edge applications and orchestrators;
• it addresses the requirements of both On-Premise deployments (with access latency in the 1-10
ms range) and Network Edge deployments (with access latency in the 10-40 ms range);
• it supports the major access network technologies in use in edge networks, including wireline,
WiFi, and LTE and 5G mobile networks;
• it supports the extension of public cloud services into the edge network;
• it supports a variety of Artificial Intelligence (AI) and media computing application frameworks,
allowing edge services to take advantage of leading technologies in these areas;
• it incorporates a controller that provides for secure on-boarding of service instances, with APIs
that support integration with orchestrators;
• it provides infrastructure to allow applications to advertise themselves as services, and for
applications to subscribe to those services;
• it uses, and integrates with, industry frameworks such as Kubernetes, Openstack, and ONAP;
• and it is implemented in a microservices style to support extension and software re-use.
Figure 10 - Architectural diagram of OpenNESS
Figure 10 shows an architectural diagram of OpenNESS, with all the main components and interfaces
with access network, core network, public cloud and other infrastructure orchestration entities. The
Edge Virtualization Agent
Edge Node
OpenNESS Edge Platform
Edge Native App Public Cloud App Cloud Connector API
OS, Container/VM Runtime
Commercial
Infrastructure
Orchestrators
OpenNESS Edge
Controller
Commercial Service
Orchestrators
Edge Lifecycle Agent
Edge Virtualization Agent
Resource
Provisioning,
EPA Config
Core Network
Configuration Agent
Controller
Agent
Hardware: CPU, FPGA,
Accelerators, NIC
WiFi
AP
Cellular
Base
Station
Application Server
(eg. hybrid cloud)
Local Breakout
S1_U
IP
SGI
Core Network
Edge App Agent, Edge Auth Agent
OpenNESS APIs
Other APIs
Public
CloudLegenda

Rev. 1.0, October 2019 21/41
Intel White Paper
OpenNESS software framework is available as open source at www.openness.org. An example of
commercial offer of MEC platform is represented by Smart Edge (https://smart-edge.com/), which is
based on Intel architecture and designed to support enterprise MEC solutions. Smart Edge enables the
computation of traffic and services on edge of the network, closer to the customer.
2.3 Functional mapping of edge computing activities
The MEC architecture, as defined by ETSI ISG MEC, does not specify at the same level of detail all the
components within the framework. In particular, the implementation level ("stage 3") is only defined
for some components really needed to ensure interoperability among different stakeholders, while
other parts may need to be proprietary and not fully specified by the standard (thus, just at functional
level). Further, some components are being defined by other organizations like SDOs, industry groups,
or open source projects and communities (key examples of industry groups are given by those
associations focusing on specific verticals like automotive (5GAA), or AR/VR as described in Section 2.2).
Figure 11 shows the different functional entities in the MEC architecture, and their respective
SDO/project where they are specified. The MEC platform is a key component in the MEC system that
enables applications to discover, advertise and consume MEC services, and provide the virtualization
infrastructure with a set of rules for the user traffic forwarding plane as described in Section 2.1.1.
OpenNESS is an open source reference platform for MEC described in Section 2.2.5. 3GPP defines the
interfaces needed to implement MEC within the cellular edge and core network as described in Section
2.1.2.
Figure 11 - Functional mapping of Edge Computing activities
MEC
platform
Mp1
MEC
app
Service
Service registry
MEC
APIs
Service
APIs
User App
LCM proxy OSS
MEC
app
Management
and
Orchestration<FQDN>
<address>
Device app
Client app
Client app
Client app
Client app
NFVI
VIM
Compute, Network, storage, acceleration
Devices / Terminals RAN / Edge Cloud Data Center / Cloud

Rev. 1.0, October 2019 22/41
Intel White Paper
As we can see in the above Figure 11, there are multiple solutions for the Management and
Orchestration (MANO) part, at the right-hand side of this figure, and these components need also to be
better discussed, especially from an Edge Computing perspective. The next section analyzes in more
detail the main aspects related to edge orchestration, starting from some definitions, describing the
various frameworks available, and finally elaborating on the specificities related to Edge Computing.
2.3.1 Management and orchestration
2.3.1.1 Orchestration definitions and taxonomy
Orchestration: The way to automatically arrange, coordinate and manage complex hardware and
software services and resources. Different types of orchestration are possible:
Service orchestration (life cycle management): Orchestration of software and hardware services
delivered across one or more deployment infrastructures (see Figure 12 below). Note: the ETSI NFV
standard for MANO does not currently address service orchestration.
Infrastructure/resource orchestration (life cycle management): Orchestration of deployment
infrastructures, composed of physical or virtual compute, storage and network resources, on which one
or more services can run.
Virtual Infrastructure Manager (VIM): This term is interchangeable with Resource Orchestrator.
Figure 12 - Different levels of orchestration (lifecycle management)
2.3.1.2 OpenStack
OpenStack is a software platform for cloud computing, mostly deployed as Infrastructure-as-a-Service
(IaaS), orchestrating Virtual Machines (VMs) and other resources to be made available to applications
(VNFs) running on the infrastructure. The software platform consists of interrelated components that
control diverse, multi-vendor hardware pools of processing, storage, and networking resources

Rev. 1.0, October 2019 23/41
Intel White Paper
throughout a data center. A few vendors (Red Hat particularly) offer their own branded OpenStack
distributions. OpenStack is the most widely deployed resource orchestrator
2.3.1.3 Kubernetes
Kubernetes (commonly stylized as k8s) is an open-source container-orchestration system for
automating application deployment, scaling, and management. Kubernetes, therefore, spans the
application orchestration and resource orchestration layers. Many cloud services offer a Kubernetes-
based platform or IaaS (Platform-as-a-Service (PaaS) or IaaS) on which Kubernetes can be deployed as a
platform-providing service. Many vendors also provide their own branded Kubernetes distributions.
2.3.1.4 ONAP
The Open Network Automation Platform (ONAP) is an open source initiative created by combining
AT&T’s Enhanced Control, Orchestration, Management & Policy (ECOMP) and the Linux Foundation’s
Open Orchestrator (Open-O) projects for orchestrating the lifecycle of services and the component parts
that constitute them.
2.3.1.5 Open Source MANO (OSM)
Open Source MANO is an ETSI-hosted project to develop an orchestration stack that includes service
orchestration and that is closely aligned with the ETSI NFV standards. OSM was started by Telefonica
and is based on Telefonica’s OpenMANO initiative. OSM represents an alternative approach to ONAP
for service orchestration and is generally considered smaller and simpler than ONAP, but it is not clear
how broad its influence is on the industry as a whole.
2.3.1.6 Edge orchestration aspects
Edge Computing imposes a few unique challenges to orchestration. To name a few:
• Mobility: (1) Edge clouds may be small enough to reside on moving platforms, such as cars,
ships or airplanes. Characteristics of the communication links to the edge, such as
throughput and latency may be changing dynamically. (2) Additionally, services need to
operate seamlessly as the mobile edge traverses between adjacent service providers.
• Resource constraints: resources, such as power, CPU, storage may be severely constrained
at the edge.
• Scale: the number of edge clouds may be large.
• Autonomy: communication from the network core to the edge may be lost, necessitating
autonomous operation for extended periods of time.
A few commercial edge orchestration solutions are available, primarily targeting the enterprise market.
Typical architectures deploy a full orchestration stack in the core of the network, with lightweight, semi-
autonomous orchestration functionality in each edge location. For Communication Service Providers
(CoSPs), challenges are compounded by the requirement (‘Mobility’ (2) above) for interoperability
between operators. Finally, Edge Computing is an opportunity for operators, but risks to become a
strategic disadvantage especially to tier-1 CoSPs (since their leadership in cloud technology may not
extend to edge orchestration).

Rev. 1.0, October 2019 24/41
Intel White Paper
Work towards edge orchestration solutions for CoSPs is ongoing in all relevant open source communities
and in ETSI:
• The leading orchestration platforms (OpenStack, Kubernetes, ONAP) are currently focusing
on solving the scale challenge for the orchestration elements in the core of the network.
There is currently no real activity towards lightweight edge orchestrators.
• Akraino’s scope includes a ‘thin local’ control plane.
• StarlingX is another edge computing stack. Focus is on edge infrastructure orchestration.
For resource and service orchestration, StarlingX intends to use OpenStack and Kubernetes.

Rev. 1.0, October 2019 25/41
Intel White Paper
3 Edge Computing deployment aspects
A key aspect for the successful market adoption of Edge Computing is represented by the infrastructure
owners' engagement. These important stakeholders are described in the following, together with the
analysis of different deployment options offered by Edge Computing technology, and the related
performance tradeoffs. Future systems, including distributed computing and information-centric
networking, will be also described at the end of this section, as they can represent a natural evolution
path of edge computing deployments.
3.1 Decision makers for the edge
Kicking off the Edge Computing market is a complex process, which needs to involve necessarily all
stakeholders, from infrastructure owners to application developers. In fact, for a real market adoption
all of them should have a role and should actively engage, enabling the deployment of MEC servers and
also the creation of added-value applications exploiting edge services. In particular, if on one hand, the
infrastructure owners (at different levels, e.g., at telco edge, enterprise, cloud providers) are key to the
instantiation of MEC platforms and proper low-latency environments hosting MEC applications, on the
other hand, it is also equally important that a wide ecosystem of application developers (hence, not only
operators and service providers) will engage and help to expand and enrich the set of innovative services
enabled by Edge Computing technologies.
While Section 4 is mainly targeted to software developers and their needs, the present Section 3 is
specifically conceived to deployment issues, which are of highest interest from an infrastructure owner
point of view. In fact, in order to be convinced and adopt Edge Computing, the owner of an edge hosting
environment needs first to better understand many aspects, from benefits and constraints, to gains and
cost tradeoffs, related to all deployment options offered by the edge.
3.2 Deployment options for MEC
A suitable mean to conduct this comparative analysis is end-to-end (E2E) performance evaluation
through proper dimensioning and system-level simulations, offering also the advantage to understand
the system behavior in proper "what-if" scenarios of interest, in the view of a real deployment.
The E2E performance evaluation for different MEC deployment options is not a trivial task. The diversity
of available mapping options from logical system architecture to physical servers is further complicated
by varying MEC service consumption options. The standard "trial-and-error" approach often leads to
extreme long planning and deployment process. Thus, Intel is conducting this analysis by using a highly
configurable E2E simulation framework to facilitate efficient MEC deployment planning and
performance evaluations prior to system provisioning. This comprehensive framework, including both
radio and edge/core network components, also exploits the Intel CoFluent System StudioTM
Technology,
where multiple simulation models are integrated to provide necessary inputs. MEC servers at different
deployment locations are simulated using the CoFluent model directly. The whole software execution
flow is replayed according to different scenarios. The CoFluent model provides the flexibility to configure
the flows easily. As shown in Figure 13, the MEC servers can be deployed at different levels in the

Rev. 1.0, October 2019 26/41
Intel White Paper
CoFluent behavioral model. Typical examples for MEC deployment options include: universal Customer
Premises Equipment (uCPE), RAN-edge, Smart Central Office (CO) and Edge Data Centre (DC).
Figure 13 - Example of system modeling with different MEC deployment levels
To simulate meaningful KPIs (e.g., E2E latency and throughput), the CoFluent model leverages statistics
from the MEC application model, MEC services model and execution of LTE/5G network (the radio and
core network simulator) to simulate computing and communication cost with high accuracy and high
speed. The interface between the CoFluent model and LTE/5G network simulation follows the packet
flow for both uplink and downlink. Simulation results can be produced by comparing deployment
options (e.g. different levels of the edge), in order to assess different MEC platform and MEC App
workload models at various network load situations, for specific use cases of interest.
3.2.1 MEC deployment tradeoffs
When it comes to deploying MEC servers, and also instantiating MEC Apps, multiple tradeoffs can be
identified, for an effective decision on physical and cloud resource usage. In particular, the processing
power demands of customer devices, namely AR/VR, drones, and autonomous vehicles can be very
heterogeneous, and, depending on the use case, they can require very low latency, typically measured
in milliseconds. The place where processing takes place plays also a major role with respect to quality of
user experience and Total Cost of Ownership (TCO). If, on one hand, the centralized cloud decreases the
TCO, but fails to address the low latency requirement, placement at customer premises is nearly
impossible with respect to cost and infrastructure. Considering the cost, low latency, and high processing
power requirements, the best available option is to utilize the existing infrastructures (e.g. Telco’s tower,
central offices, and other Telco real estates): these will be the optimal zones for the edge placement, at
least in a first place. Figure 14 depicts an example of the various performance tradeoffs related to the
different edge deployment options.

Rev. 1.0, October 2019 27/41
Intel White Paper
Figure 14 - Example of performance tradeoffs for various edge deployment options
(source: Akraino website).
3.2.2 Case study: How MEC service consumption influences the E2E communication
latency
Several options are available to decision makers when it comes to MEC deployments, and essentially
they depend on the actual location of instantiating MEC applications, which are server-side endpoints
of the communication between UEs and the edge. These locations range from uCPEs (mainly at the
customer premises), to RAN-edge (thus, with MEC servers co-located with base stations), or to Smart
COs, where the platform could be e.g., hosted as co-located with Cloud-RAN (CRAN) aggregation points,
or again, to edge DCs at local/regional level, but also potentially also to remote DCs (even if, of course,
this would be not a "real" edge deployment).
On the other hand, a specific MEC deployment option can be convenient or not, depending on the
specific requirements, thus, a meaningful evaluation of these options should necessarily be tailored to
a specific use case of interest. In this perspective, having the MEC system based on a virtualized
infrastructure provides the required flexibility to adapt the MEC deployment and topology based on
varying needs and convenient performance-cost tradeoffs.
In addition to that, MEC applications are often consumers of edge services (e.g., through MEC APIs),
consequently, the actual performance tradeoffs depend also on the consumption of these services and
the actual instance locations of both MEC Apps and a MEC platform (which can be co-located or even
remotely accessible by MEC Apps).
As a first case study, we have evaluated in a simplified scenario the E2E latency (Round Trip Time – RTT)
considering three different workloads (“low”, “middle” and “high”) of an exemplary CDN service,
relevant to, e.g. In-Vehicle Entertainment (IVE) use cases in automotive environments.
We have also compared two MEC deployment options (depicted in the left-side of Figure 15):

Rev. 1.0, October 2019 28/41
Intel White Paper
• MEC at the RAN edge (consisting in co-location of edge hosts with Radio Access Points (RAPs)/
Base Stations (BSs)), and
• MEC at a Smart CO, e.g., at CRAN aggregation point, where multiple Radio Remote Heads (RRHs)
are connected to a Baseband Unit (BBU) pool via e.g., fiber optical fronthaul links,
where these options are also combined with three different cases of MEC service consumption:
1. MEC service consumption at the same locality (i.e., the same MEC host),
2. MEC service consumption at different localities, where the MEC App is consuming a service
running in the MEC platform (“remote app-to-service”), or
3. MEC service consumption again at different localities, however, the difference being in that the
consumer MEC App is in need of a service-producing MEC App (“remote app-to-app”).
Figure 15 - MEC infrastructure deployment options
In this simple case study, we evaluated as initial KPI the RTT performance for each of the two focused
MEC system deployments (illustrated in Figure 16), by taking into account all three workloads and all
the above mentioned cases of MEC App service consumption. It is observed that, (i) lower E2E delay is
caused when the MEC App is consuming local MEC Service APIs, (ii) higher E2E delay is caused when the
MEC App is consuming remote MEC Service APIs (i.e., not running at the same MEC host as the MEC
App), (iii) the considered MEC App computation load highly impacts the RTT, while, (iv) low gain is
observed when moving from “smart” CO to RAN edge deployments. The last observation is coherent to
a consolidated trend among the operators, initially oriented to early MEC deployments at the Smart CO
(CRAN aggregation point), which is often a first convenient location of deploying MEC servers (of course,
some other operators are instead preferring to consider edge RAN deployments, since better
performing and more suitable for more challenging use cases).
Web
Access Level
RRH
RRH
RRH
macro
BBU
Local/Regional Level
National Level
Remote Cloud
app
app
app
app
app
app
Global Level
app
uCPE

Rev. 1.0, October 2019 29/41
Intel White Paper
Figure 16 - Impact of different MEC system deployment options,
MEC service consumption variants and workload sizes on two-way delay (RTT).
3.3 Evolution path of Edge Computing deployments
Edge Computing brings cloud computing closer to the users at the infrastructure edge. However, in the
future, Edge Computing is likely to creep deeper into the edge and client devices in a more distributed
manner. These approaches aim to leverage the increasing processing and storage capabilities at client
devices towards reducing latency beyond what current Edge Computing can provide.
3.3.1 Distributed computing systems
Edge Computing can span a variety of network locations, form factors, and functions, as depicted in
Figure 1 at the beginning of this white paper. Further into the network and cloud is more centralized
computing, with applications addressing large numbers of users, and edge platforms hosting multiple
applications simultaneously. Distributed computing is closer to end users with applications being more
attuned to specific endpoints and functions. This “distributed edge” is characterized by spatial and
temporal proximity to clients, real-time responsiveness, interactivity, and mobility, and is well suited for
use cases such as industrial control, video analytics, together with interactive (XR) media and healthcare.
Figure 17 shows an oversimplified Venn diagram with some of the common and different attributes of
edge processing in different locations. It should be noted that this is not meant to be exhaustive or
prescriptive but rather illustrative.

Rev. 1.0, October 2019 30/41
Intel White Paper
Figure 17 - Edge deployment variations
There are a number of possible ways to deploy and interconnect distributed edge processing nodes,
such as the topology scenarios depicted in the following diagram, assuming a 5G System (5GS)
architecture.
Figure 18 - Distributed edge deployment scenarios
A distributed Edge Computing site may stand alone and be self-contained with no upstream control or
data connection, providing wireless connectivity and local processing, for example, in a small-scale
industrial control system. However, in most cases it is necessary to connect to upstream processing for
common control and aggregation across multiple access points. Two scenarios are shown in Figure 18;
in the first (A) case, the distributed site provides applications running locally with access to data behind
an intermediate UPF, while, also feeding into an upstream aggregation UPF in a common 3GPP 5G core
network. This setup allows the common control plane to steer data flows toward either local or
Distributed On-Prem DC
CO/Regional DC
Network Failover
UPF
5GC
<100W
300+W
Small Cell gNB
C-RAN
<5ms <5-10ms
<20ms
Private
Public
Client-oriented
Service-oriented
Site-oriented
Mass
Storage
Limited Storage
<50 clients
1000s+ of clients
100s+ of clients
Multi-tenant
Focused
function
Diverse
Environment
Server/R
ack
Multi-function
Single-
tenant
AMF SMF
PCF AF
(I-)UPF
5GC CP
N5
(REST API)
DU CU
gNB
Applications
UPFCU
gNB
Applications
DU
DU
DU
DU
DU
Distributed Edge Site A
Distributed Edge Site B
N2 / N4
N3 N6
N3 N6
CO/Edge Site
Hierarchical
(Aggregate)
Edge Processing
Datacenter Servers Cloud
N9
N9
UPF
N6
Applications
ISP/CSP Edge Site
Applications
Local
(In-line)
Edge Processing
Alternative
upstream
data plane
connectivity
models

Rev. 1.0, October 2019 31/41
Intel White Paper
upstream compute resources in a flexible manner by configuring appropriate forwarding rules in the
intermediate UPF. The behavior of the 5G network connecting the distributed sites can be influenced by
AFs interacting with the 5G core network through well-defined interfaces. In the second (B) case, the
distributed site provides applications running locally with access to data behind a local UPF which
terminates the 3GPP flow. The local distributed applications can connect upstream as needed via normal
IP networking. In fact, this is not precluded in the first (A) case either – once data is broken out locally,
it can be handled and distributed as needed. This allows data partitioning according to site-specific
performance and security requirements.
The drivers behind this distributed edge architecture are to provide targeted, unconstrained low-
overhead processing very close to clients for high performance, responsiveness, and reliability without
complexity. Leveraging the 3GPP management/control structure and user plane with QoS, even in a
private networking domain, provides a proven consistent framework for secure reliable wireless
networking with mobility management.
Some key aspects of this approach include:
• flexible local native 5G network deployments: small footprint, highly distributable;
• local aggregation points with in-line processing (e.g., media/AI) close to clients;
• fundamentally optimized for high reliability and low latency;
• part of a hierarchical network solution where desired.
There is also a need for distributed access control software for workload balancing so as to simplify
provisioning and reconfiguration across the local edge nodes. This is complementary to service-provider
network distributed service offerings, providing an alternative client-centric perspective:
• workloads specific to clients and applications installed on them, not pre-defined;
• workload images and lifecycles are dynamic, depend on use case / app running at the client;
• each workload has specific performance and QoS requirements.
3.3.2 Information-centric networking
Internet Protocol (IP) is based on host centric connectivity. The first step to obtaining a piece of
information is an IP address of the host. Next, a secure session needs to be established with the host
after which the host starts to send the data packets. The network layer’s only function through this
process is to route the packets from the source host to destination client. Instead, in Information-Centric
networking (ICN), you can ask the network for that content directly by its name. This could be over any
connection the device has to another device or network (wired or wireless). The network then responds
with that piece of information; in a sense the network is a database. A node sends an “interest packet”
to the network, where the packet contains the name of the data it wants [18]. There is no reference to
a physical location or an address. When a node receives the interest packet for named data, it first
checks the contents of its own cache. If the data is not present there, the query gets forwarded based
on a local routing table, and this process continues until the query arrives at a node that has the data.
As the data traverses back to the requesting node, any intermediate node can cache the data, and then
serve it if requested by another node. It is possible for any node to cache the data and serve it back
because security is now contained in the data. Currently with IP, security is based on connecting to a
trusted host. Since there are no end to end connections with ICN, the security is embedded with the

Rev. 1.0, October 2019 32/41
Intel White Paper
data itself. This structure enables the user to always get the data from the closest location, improving
network efficiency and scaling.
ICN can also be used to orchestrate computation. Instead of sending an interest packet for a piece of
information, the user (device) can request the execution of a function. The network then routes the
information to the closest resource that can compute the desired functions and returns the processed
data back. This is a powerful paradigm where the entire edge can be a computation server. If the user is
in a new environment, without any knowledge of the closest edge server, the device can still request
the network to orchestrate the computation. Named Function Networking (NFN), Named Function as a
Service (NFaaS) and Remote Method Invocation over ICN (RICE) [17] are examples of implementing
dynamic and distributed computation within the network. In NFN [19], the network’s role is to resolve
names to computations by reducing λ-expressions. NFaaS builds on very lightweight VMs and allows for
dynamic execution of custom code. RICE presents a unified approach to remote function invocation in
ICN that exploits the attractive ICN properties of name-based routing, receiver-driven flow and
congestion control, flow balance, and object-oriented security.

Rev. 1.0, October 2019 33/41
Intel White Paper
4 Edge applications and ecosystem
engagement: an automotive case study
Edge Computing represents cloud-computing capabilities and an IT service environment at the edge of
the mobile network. It brings virtualized applications much closer to mobile users ensuring network
flexibility, economy and scalability for improved user experience. It facilitates a service environment that
allows seamless access experience and responsiveness for content, services, and applications. In this
section we will explore platform capability needed to support a thriving application ecosystem using
“Real-time Situation Awareness and High-definition Maps” as the example edge application.
4.1 Real-time situation awareness and high-definition maps
An automated, semi-automated or manual-driven vehicle is moving on a road (route), heading towards
a specific road segment, which presents unsafe and unknown conditions ahead. This host vehicle is made
aware in a timely manner of situations detected and shared by remote vehicles and/or road
infrastructure nodes such as Road Side Units (RSUs). Situations may include such things as accidents,
adverse weather, road conditions, traffic, and construction, among others. The shared situations are
relevant along the host vehicle’s navigation route or current road of travel.
The situation is shared via data already embedded or to be embedded in a high-definition map. A
centralized Edge Computing node collects the situations detected and by remote vehicles and/or road
infrastructure nodes such as RSU. The Edge Computing node then consolidates the information collected
and intelligently distributes/reports the relevant information to the corresponding road users. This Edge
Computing node may be part of or connected to a traffic management authority.
4.1.1 High-definition map
An HD map is necessary as a reliable off-board sensor containing carefully processed a-priori information
to “detect” features that are not easily detectable by on-board sensors or to provide a redundant source
of information for on-board sensors, including location-based operational design domain determination,
environment modeling in adverse conditions and precise semantic understandings in complex driving
situations. In situations where on-board sensors cannot reliably detect features, the HD map can be
utilized as a more reliable redundant source of information.
The HD map consolidates static and dynamic information (e.g., vehicle position, pedestrians and
obstacles) and is mandated for autonomous driving. Creating and distributing the map require many
data transactions with high capacity as well as efficient processing to keep the information up to date.
This HD map must be able to accurately localize dynamic objects including vehicles, which is required
for automated driving beyond the traditional route guidance. A large amount of data transfer is
especially required to update the map. Data is collected from on-board cameras, radar sensors, and laser
scanners (LIDAR), transferred and processed in the cloud or in the edge. Typically, what might get sent
to the cloud are deviations (Map says X, but Camera says Y). These deviations are sent to the cloud or
to the edge to update the HD map.

Rev. 1.0, October 2019 34/41
Intel White Paper
The completed map information is stored in the center server or the edge server and needs to be
distributed to relevant vehicles in a timely manner.
4.2 Example of automotive blueprint for Edge Computing
Blueprints are defined in Akraino as the declarative configuration of entire stack i.e., Cloud platform,
API, and Applications. Intend of Akraino Edge Stack is to support VM, container and bare metal
workloads. Section 4.2.1 below contains an example of blueprint, customized for the automotive case
study, based on the current developments in the Akraino project.
4.2.1 Connected Vehicle Blueprint
The Connected Vehicle Blueprint project in Akraino9
focuses on establishing an open source MEC
platform, which is the backbone for V2X Application. As per Akraino criteria and requirements, blueprint
code that will be developed and used with Akraino repository are using only open source software
components either from upstream or Akraino projects.
The goal is to establish a MEC platform for connected vehicle use cases, since an edge platform for
deploying connected vehicle application does not exist in Akraino so far. In the table below, the main
parameters and characteristics of this blueprint are listed:
Case Attributes Description
Blueprint Name Connected Vehicle Blueprint
Type New Blueprint for the Edge
Blueprint Family It is a independent blueprint, NOT a blueprint family yet.
Use Case MEC platform used for Connected Vehicle.
Initial POD Cost
(capex)
The Minimum Configuration: 4 Servers in total
Scale & Type MEC Platform (1 Server) + 1 App Server (1 Server) + 2 Simulators (2 Server)
Applications
The MEC platform which can be used to connect vehicles, the general data
flows are itemized below:
1. Grab the traffic/vehicle information
2. Dispatch the traffic/vehicle information to the corresponding edge
process unit. Note well: The dispatch policy can be configurable.
3. Process the data in the Edge or Cloud and figure out the suggested action
item for the vehicle driver
Send the suggested action items to the vehicle driver
Power Restrictions
Less than 6KW. The Maximum Power consumption for each server is around
1500W, thus in total 1500 * 4 = 6000W
9
https://wiki.akraino.org/display/AK/Connected+Vehicle+Blueprint

Rev. 1.0, October 2019 35/41
Intel White Paper
Infrastructure
orchestration
• Docker + K8s
• VM and OpenStack/StarlingX
PaaS Tars and OpenNESS
Network OVS, DPDK, VPP
Workload Type Bare metal, VM, Container
The following figure clarifies also how OpenNESS fits into the Akraino Connected Vehicle Blueprint.
Figure 19 - Akraino Connected Vehicle Blueprint.
4.3 Life cycle management of edge applications
We see edge applications as an evolution of cloud applications. So, there is an inherent expectation for
edge application life cycle management to leverage the best practices of cloud native applications e.g.,
DevOps and CI/CD (Continuous Integration / Continuous Delivery). With this paradigm the main
requirements coming from two categories of stakeholders (Developers and IT Ops) are combined
together (see figure below).
Figure 20 - System requirements in a DevOps paradigm.

Rev. 1.0, October 2019 36/41
Intel White Paper
Edge cloud infrastructure must provide the following sets of service for Edge Application Life Cycle
Management:
1. Application Provisioning Service
a. Compute, Storage and Network Provisioning.
b. Application Package and Artifact Provisioning.
c. Security and QoS Provisioning.
2. Application Monitoring and Diagnostic Service
3. Application Auditing and Accounting Service.
As the application will be executing in a multi-tenant environment, it is expected that the edge cloud
infrastructure provide strong hardware/software enforced application isolation to protect applications
privacy, confidentiality and resource allocations. We also expect edge cloud infrastructure to provide
various level of resource abstractions to the application
Abstraction Compute Storage Network
IaaS Virtual Machine Virt BLK Virt Net
CaaS Container Storage Plugin CNM/CNI
Serverless Functions SQL/NoSQL/Object Storage API HTTP API EndPoints
Edge applications have some unique infrastructure requirement that is not necessarily applicable to
cloud native application e.g.:
• Secure, Scalable, Manageable and Multi-tenant;
• Edge/Cloud Asset (Pi to Rack) lifecycle management. (Config, Security and Ownership etc.);
• Intelligent Automatic fluid/seamless function delivery and execution (cloud to edge);
• Intelligent/Transparent Automatic State and Data mirror/shadow;
• QoS/SLA driven and real-time (for mission critical and time sensitive applications);
• Zero-touch (plug-n-play) edge;
• Serverless (No Ops, Transactional billing, auto application reliability/scalability);
• Uniform programing model.

Rev. 1.0, October 2019 37/41
Intel White Paper
5 Conclusions
Edge Computing (often named also as MEC) is recognized as a key technology, supporting innovative
services for a wide ecosystem of companies. The focus is to address the needs of two stakeholder
categories in this ecosystem: infrastructure owners (e.g., operators and cloud providers) and software
developers (e.g., applications / content providers, innovators and startups). In fact, both stakeholder
categories are key to the success of MEC, and their engagement depends also on the way the various
challenges are addressed.
In particular, in this paper we provided an overview of standardization efforts, including initiatives from
industry groups, associations, open source communities and projects (e.g., OpenNESS). We also derived
some edge deployment considerations (of high importance for infrastructure owners), trying to answer
the recurrent question "Where is the edge?". Since many use cases are in principle supported by MEC
(and indeed, MEC performance depends on the specific KPIs derived from these use cases), in order to
provide a more specific analysis we describe an exemplary case study customized to the automotive
domain. Such an example aims to constitute a useful reference for the software developer (i.e., the
other main category of MEC stakeholders), when it comes to answering to the other recurrent question
"How to use MEC from application development point of view?").
In the end, the overall market success of Edge Computing will depend, on how infrastructure owners
will be able to address all deployment aspects and also on how software communities will create a wide
set of applications and innovative services. Intel is committed to engaging the entire ecosystem,
considering standard solutions and open source projects, towards ensuring interoperability, while
opening the market to proprietary implementations and added-value propositions.

Rev. 1.0, October 2019 38/41
Intel White Paper
Appendix - OpenNESS Applications and Services
The Open Network Edge Services Software (OpenNESS) Toolkit is an open source software toolkit that
enables orchestration and management of edge services on diverse platforms and network
technologies. OpenNESS was introduced in Section 4 of this White Paper. In this Annex, we present the
behavior of OpenNESS Applications and Services.
The OpenNESS concept of a service is derived from, and similar to, a service in the ETSI MEC standard.
In the ETSI MEC standard, an application may provide a service to other applications. The reader is
directed to Reference [6] for additional details, while a simplified description is provided here.
• An ETSI MEC service registers itself by sending a message to the MEC platform. Other applications (“relevant”
applications, as documented in [6]), which have subscribed to the service, receive notifications from the MEC
platform that the new service is available. Thereafter, until the service deregisters itself, the service, according to its
operation, sends notifications to the subscribed applications. The registration, subscription, and notification
operations are carried out via RESTful APIs defined by the ETSI MEC standard. Subscribing applications arrange to
receive notifications by providing the MEC platform with a callback reference URI, which is shared with the service.
• In addition to this functionality, the ETSI MEC standard defines several designated services, including Radio Network
Information (RNI) and Location services. Their functional behavior, with respect to subscription and notification, is
analogous to that of application services.
OpenNESS applications and services behave similarly to ETSI MEC Apps and services, but include
additional functionality for secure communication between an application or service and the edge
platform.
Figure 21 - OpenNESS applications and services.

Rev. 1.0, October 2019 39/41
Intel White Paper
The architecture of an edge platform, including applications, is shown in Section 2. The primary changes
in the OpenNESS toolkit are:
• Communication between an application/service and the edge platform occurs over a secure
socket, which is established by the application prior to registration or subscription operations.
• Services and applications find each other via an agreed identifier that is in a namespace, to
support multiple independent applications.
• The edge platform provides the transport for the notification from service to all subscribed
applications.
• The OpenNESS distribution does not include the standard ETSI MEC Services, such as RNI or
Location. It is expected that system integrators will provide implementations of these services,
which could be done via the OpenNESS APIs.
Figure 22 - OpenNESS Edge Application Agent (EAA) APIs
OpenNESS APIs are defined as RESTful APIs, documented according to the OpenAPI standard. They may
be found at https://openness.org/api-documentation/?apieaa#/ .
The OpenNESS project is committed to working with SDOs and other special interest groups (SIGs) to
enable the rapid deployment of edge services in the marketplace.

Rev. 1.0, October 2019 40/41
Intel White Paper
References
[1] iGR White paper: "The Business Case for MEC in Retail: a TCO analysis and its Implications in the 5G era",
https://www.intel.com/content/www/us/en/communications/multi-access-edge-computing-brief.html
[2] Chetan Sharma Consulting: "5G Mobile Edge Computing: Redefining the Sports Experience", Link:
https://builders.intel.com/docs/networkbuilders/5g-mobile-edge-computing-redefining-the-sports-experience.pdf
[3] Ericsson blog: "Edge computing success—a distributed cloud approach", Link:
https://www.ericsson.com/en/blog/2018/9/edge-computing-successa-distributed-cloud-approach
[4] ETSI ISG MEC website, Link: https://www.etsi.org/technologies/multi-access-edge-computing
[5] ETSI GS MEC 003 V2.1.1 (2019-01), "Multi-access Edge Computing (MEC); Framework and Reference Architecture ",
https://www.etsi.org/deliver/etsi_gs/MEC/001_099/003/02.01.01_60/gs_MEC003v020101p.pdf
[6] ETSI MEC APIs, Link: https://www.etsi.org/committee/1425-mec
[7] 3GPP TS 23.501 System architecture for the 5G System (5GS),
http://www.3gpp.org/ftp//Specs/archive/23_series/23.501/23501-g20.zip
[8] 3GPP SA5 Study on management aspects of edge computing, FS_MAN_EC,
https://portal.3gpp.org/desktopmodules/WorkItem/WorkItemDetails.aspx?workitemId=800039
[9] 3GPP SA2 Study on enhancement of support for Edge Computing in 5GC, FS_enh_EC,
[10] 3GPP SA6 Study on Application Architecture for enabling Edge Applications (FS_EDGEAPP),
[11] Industrial Internet Consortium: "The Industrial Internet of Things - Volume G1: Reference Architecture", version 1.9,
June 19, 2019, Link: https://www.iiconsortium.org/pdf/IIRA-v1.9.pdf
[12] GSMA Cloud AR/VR Whitepaper, available at: https://www.gsma.com/futurenetworks/resources-2/gsma-online-
document-cloud-ar-vr-whitepaper/
[13] ISO/IEC 23090-8: "Information technology -- Coded representation of immersive media -- Part 8: Network-based media
processing".
[14] 3GPP TS 26.118: "3GPP Virtual reality profiles for streaming applications".
[15] ISO/IEC 23090-2: "Information technology -- Coded representation of immersive media -- Part 2: Omnidirectional media
format".
[16] VR Industry Forum Guidelines, available at: https://www.vr-if.org/guidelines/
[17] M. Król et al.: “RICE: Remote Method Invocation in ICN”, ACM ICN Conference, 2018
[18] Lixia Zhang, KC Claffy, Patrick Crowley, Christos Papadopoulos, Lan Wang, Beichuan Zhang, “Named Data Networking”,
ACM SIGCOMM Computer Communication Review, Vol 44, No. 3, July 2014
[19] Christian Tschudin, Manolis Sifalakis, “Named Functions and Cached Computations”, IEEE 11th Consumer
Communications and Networking Conference (CCNC), 2014

Rev. 1.0, October 2019 41/41
Intel White Paper
Intel Notices and Disclaimers
Intel provides these materials as-is, with no express or implied warranties.
All products, dates, and figures specified are preliminary, based on current expectations, and are subject to change without notice.
Intel, processors, chipsets, and desktop boards may contain design defects or errors known as errata, which may cause the product to
deviate from published specifications. Current characterized errata are available on request.
Intel technologies' features and benefits depend on system configuration and may require enabled hardware, software or service activation.
Performance varies depending on system configuration. No product or component can be absolutely secure.
Check with your system manufacturer or retailer or learn more at http://intel.com.
Some results have been estimated or simulated using internal Intel analysis or architecture simulation or modeling, and provided to you for
informational purposes. Any differences in your system hardware, software or configuration may affect your actual performance.
Intel and the Intel logo are trademarks of Intel Corporation in the United States and other countries.
*Other names and brands may be claimed as the property of others.
© Intel Corporation

Edge computing from standard to actual infrastructure deployment and software development

More Related Content

What's hot (20)

Similar to Edge computing from standard to actual infrastructure deployment and software development (20)

More from DESMOND YUEN (20)

Recently uploaded (20)

Edge computing from standard to actual infrastructure deployment and software development