O-RAN Alliance and the ONF 5G SD-RAN
There are several organizations involved in O-RAN specification and development. These
include the 3GPP, the O-RAN Alliance, the Telecom Infra Project (TIP), the Small Cell Forum,
the Open Networking Foundation (ONF), and the OpenAirInterface Software Alliance (OSA).
Each of these groups is making a useful contribution to O-RAN development
The 3GPP is the preeminent RAN standards organization, and this will remain the case.
However, where the RIC is concerned, the O-RAN Alliance is the critical organization. The
RIC is specified by the alliance as an integral part of the O-RAN architecture and is not (yet)
part of the 3GPP RAN standards. The O-RAN architecture dovetails with the 3GPP RAN
architecture and, generally speaking, should be seen as additive to existing standards.
There are two working groups focused on the RIC in the O-RAN Alliance:
- WG2: The Non-real-time RAN Intelligent Controller and A1 Interface Workgroup
- WG3: The Near-real-time RIC and E2 Interface Workgroup
The O-RAN Alliance supports the O-RAN Software Community in a cooperation with the
Linux Foundation. In the summer of 2020, the alliance also announced it would partner with
the ONF to launch the 5G Software-Defined Radio Access Network (5G SD-RAN) project to
encourage and facilitate the creation of open source software for mobile 4G and 5G RAN
deployments. Initially, the project will “focus on building an open source near-RT RAN
Intelligent Controller compatible with the O-RAN architecture.” It is backed by major
operators and tech companies, including hyperscale public cloud companies.
The existence of the SD-RAN group does not mean that all RICs will be open source—or
even that they will be based on the open source code developed by the project group.
However, the software community that is expected to grow up around this open source
project will influence RIC development—particularly the wider xApps/rApps ecosystem. An
example of this is the involvement of the TIP RAN Intelligence and Automation (RIA)
subgroup with the 5G SD-RAN program. The RIA aims to develop and deploy AI/ML-based
applications (as xApps) for a variety of RAN use cases, including SON, radio resource
management (RRM), and massive multiple input and multiple output (MIMO).
What are the RIC use cases?
The RIC customizes RAN functionality by optimizing radio resources according to operator
policies. To use an SDN analogy, the RIC is a controller/orchestrator and the base station a
type of specialized forwarder. The RIC has an impact on RAN performance in three main
- Network intelligence: For example, measuring and reporting how the RAN is
performing. This generates data, in standardized formats, that can be analyzed (e.g.,
using AI/ML techniques) to create new algorithms and polices.
- Resource assurance: To ensure that devices/users and services receive the
required performance (e.g., via radio link control optimization, handover
optimization, or prioritization).
- Resource control: To ensure the RAN system is operating efficiently when multiple
user groups may be competing for adequate resources.
xApps/rApps and RIC use cases
The services delivered by a RIC are composed of xApps or rApps, or a combination of both.
There is no hard limit on the types of xApps or rApps that can be built, and it is anticipated
that more than one xApp or rApp will be executed in the RAN at a time. It is also expected
that some RIC services will use a combination of xApps and rApps.
Sample xApps/rApps suggested to date include the following:
- Context-based dynamic handover management for vehicle-to-everything (V2X)
- Dynamic radio resource allocation for unmanned aerial vehicles
- Traffic steering
- Quality-of-service/quality-of-experience (QoS/QoE) optimization
- Massive MIMO beamforming optimization
- RAN sharing
- QoS-based resource optimization
- RAN slice service assurance
- Multi-vendor slice performance management
- Dynamic spectrum sharing
- Dynamic spectrum sharing
- Network slice subnet instance (NSSI) resource allocation optimization
- Local indoor positioning in RAN
Further detail on these can be found in the O-RAN Alliance’s O-RAN Use Cases and
Deployment Scenarios White Paper (February 2020), which can be found here: www.oran.org/s/O-RAN-Use-Cases-and-Deployment-Scenarios-Whitepaper-February-2020.pdf.
Initial xApps will be focused on health check functions. For example, are the RAN nodes
operational and functioning as normal? In a second phase, xApps will extend deeper into
observability by gathering more granular data from RAN nodes for analysis. xApps that
make close to real-time changes (i.e., less than 1 second time cycles) will come in later
phases. In terms of control plane decisions, an initial approach will be to complement the
current RRM functions implemented in the CU and DU, wherein the policies from the RIC
could change or override the local RRM logic. A more aggressive approach of offloading the
RRM function completely to the RIC will be a longer-term event.
Initial rApps, hosted in the non-RT RIC, will initially resemble today’s centralized SON
(C-SON) applications. They have the potential to evolve quickly as RAN data collection is
paired with ML techniques to create algorithms that bring new forms of optimization to the
THE RIC IN THE MOBILE NETWORK ARCHITECTURE
The RIC is specified in the O-RAN architecture, which is overlaid on the 3GPP RAN
architecture. Figure 1 below shows a simplified view of the combined architecture that
highlights the role of the RIC. The 3GPP-defined RAN functions that interface with the RIC—
the DU and CU—are shown in dark red. The near-RT RIC, shown in bright red, connects
over the O-RAN Alliance’s E2 interface to the DU and CU. The non-RT RIC, shown in light
red, connects directly to the radio nodes and to the near-RT RIC over the O-RAN Alliance’s
O1 and A1 interfaces.
The DU is the baseband unit in a 5G RAN. It performs Layer 1 and 2 processing and
executes critical functions such as encoding/decoding, scheduling, MIMO processing, and
beamforming. It is the heart of a RAN system and makes sub-millisecond decisions on how
to allocate radio resources within a cell based on factors such as user equipment type and
distribution, interference conditions, and higher level policies. The near-RT RIC does not
take over these real-time RRM functions, which must remain on the DU due to stringent
latency requirements. However, what the near-RT RIC can do (in principle, in the future) is
programmatically configure the DU to improve how these functions work. For example, the
near-RT RIC could be used to change scheduler behavior on the DU.
The CU is a new node in 5G that does not exist as a discrete function in 4G. In the 3GPP
RAN architecture, the CU provides Layer 3 functions such as session management and
mobility management. It is split into the CU-UP for user plane processing and the CU-CP for
control plane processing. The near-RT RIC again interfaces with the CU-UP and CU-CP over
the E2 interface to deploy a policy or apply dynamic control. For instance, in a scenario
where devices have high speed mobility requirements (e.g., motor vehicles), a different
handover management algorithm may be superior.
In this architecture, the RIC is therefore largely additive and complementary to the existing
3GPP-defined architecture. By bringing new capabilities, it can enhance how these RAN
nodes perform and, as an open platform, it can attract software innovators to the radio
domain. There may be instances, particularly with incumbent 5G vendors, where the new
control loops introduced by the RIC conflict with their own procedures for updating policies
on the DU and CU functions. This should be resolvable, and it is hoped that incumbent
vendors will also support the E1 and O1 interfaces.
A near-RT and non-RT RIC model
The non-RT RIC is used for tasks with longer time cycles, including functions traditionally
associated with C-SON, as well as emerging ML approaches to RAN optimization. It connects
to the RAN over the O1 interface to collect data (in standardized formats) and to deploy
policies (in the form of rApps) where longer time cycles are appropriate. The non-RT RIC
can also deploy policies over the A1 interface to inform the behavior of xApps in the near-RT
RIC, which then are deployed into the RAN over the E2 interface.
With its placement higher up in the architecture, the non-RT RIC can access larger RAN
datasets, generated over longer time periods, to gain deeper insight into performance and
identify potential optimizations that are not apparent in sub-second processing. The non-RT
RIC can interface with other network data sources (e.g., to co-optimize radio, IP
networking, and edge cloud performance). It can also access datasets external to the
networks itself, such as those related to traffic, emergency services, weather, mass public
events, etc., so the network can prepare or respond to real world events. Traditional C-SON
can do some of this, but the process is typically reactive, and it can be slower to implement
changes than the new RIC model.
The non-RT RIC is particularly interesting in the area of ML for RAN optimization. RANs
generate vast amounts of data that can be processed to provide insight into network
behavior. This data is classically analyzed offline using RAN analysis tools, with the results
used to inform how a network could be better optimized. The use of ML to analyze this data
and identify improvements is showing great promise. The non-RT RIC could become the
platform on which new policies, developed using ML insight, can be created and hosted prior
to being pushed down into the network. This may result in faster training of models that are
specific to the network in question and faster implementation in the network than is
currently achieved with C-SON platforms.
The importance of the E2 interface
The E2 interface has a lot of work to do in the O-RAN architecture. It is the critical interface
between the near-RT RIC and radio nodes (radio nodes are sometimes called E2 nodes by
RIC developers), and it is imperative that this interface is unambiguously open. A key test
of O-RAN will be the how well E2 is supported by RAN equipment companies.
The “E2 manager” is an integrated part of the near-RT RIC platform and is sometimes
referred to as “xApp zero.” It is used to initiate E2 connections with the RAN nodes and then
store RAN configuration information learned at connection setup (and kept updated over
time). The first release of the RIC specifications is now available, and O-RAN Work Group 3
will continue to be developed to add functionality; for example, E2SM (E2 Service Model)
specifications are now in development.
Two key work items currently underway are to standardize E2SMs that enable traffic
steering and QoS/QoE optimization through the RIC. Traffic steering targets idle mode
mobility load balancing (MLB), inter-intra frequency MLB, carrier aggregation, and dual
connectivity. QoS/QoE optimization allows the RIC to control network functions related to
QoS control, radio resource allocation, radio access control, connection management, and
It is anticipated that operators will deploy RICs on a one-to-many basis, with each RIC
connecting to multiple O-DUs and O-CUs. The E2 must therefore support failure handling for
improved resiliency. However, according to the specification, RAN nodes must also be able
to function normally independently of the RIC in case the E2 interface and/or the RIC fails.
RIC DEPLOYMENT ARCHITECTURE
The 5G RAN is hierarchical. Multiple RUs connect to one (or more) DUs, multiple DUs
connect to a CU, and multiple CUs connect to the core network. At each stage, traffic is
aggregated and control is centralized. The RIC will be deployed at some point between the
cell site and the core network. The O-RAN Alliance Working Group 6 (WG6), which is
focused on cloud architecture and deployment scenarios for virtualized O-RAN, distinguishes
between the core cloud, the regional cloud, and the edge cloud at the cell site. The RIC will
be deployed in one of these locations.
If the nature of the use case is not latency sensitive, it is possible to deploy the RIC in the
central cloud at same locations as core network functions are deployed today. A deployment
of this type would be similar to a C-SON deployment and would be capable of making
changes on longer time cycles, such as performed by the non-RT RIC. The core cloud would
not, however, be suitable for xApps that target near-real-time tasks.
In most cases, it would also not make sense to deploy the near-RT RIC at the cell site
because the value of this function is that it coordinates across multiple radio nodes to
improve system performance. An exception to this would be private network deployments,
or venue networks, where a RIC could be deployed onsite to monitor and control the local
Regional and edge cloud deployment
The right place for near-RT RIC deployment, in most cases, will be the regional cloud (to
use WG6 terminology). WG6 identifies several O-RAN deployment scenarios. From a RIC
perspective, the following three appear most likely:
- Scenario A: The near-RT RIC, O-CU, and O-DU functions are all virtualized on the
same edge cloud platform, and the interfaces between those functions are within the
- Scenario B: The near-RT RIC is deployed on a regional cloud platform, while the
O-CU and O-DU functions are virtualized on an edge cloud platform.
- Scenario C: The near-RT RIC and O-CU are deployed on a regional cloud platform,
while the O-DU is virtualized on an edge cloud platform.
Scenario C is particularly interesting because the near-RT RIC covers a geographical area
that is comparable to the number of sites that would be aggregated on a CU node. It also
has the same kind of latency requirements as the CU in relation to the DU. The CU-CP, for
example, carries out tasks such radio link control and handover messaging that can tolerate
milliseconds of delay, while the CU-UP processes user plane traffic and forwards it into the
IP layer. It both cases, it makes sense to deploy these nodes at locations where it is
possible to aggregate traffic from relatively large areas (e.g., aggregating 10s–100s of cell
SUMMARY AND NEXT STEPS
O-RAN is a new way to build mobile RANs based on open interfaces, modularization, off-theshelf hardware, and innovation in software. Operator trials are underway in all major
regions, with the first commercial networks to be deployed and put into service from 2021
onwards. Heavy Reading believes there is potential for O-RAN deployments to scale as
performance is shown to be competitive and as integration of O-RAN into operator planning
and management tools improves.
Software control of the RAN is a key breakthrough of O-RAN. Programmatic control of the
network enables operators to adapt their networks for different service types and user
groups. This, in turn, introduces software-driven operating models to a part of the network
that is classically hardware-oriented and where change occurs over long time cycles.
Programmatic control is vital to automation and therefore significantly affects the cost and
efficiency of the network.
The key messages of this paper are the following:
- The RIC is a core component of the O-RAN architecture; it is what makes O-RAN
dynamic and software-centric, which ultimately enables higher performance 5G
- The RIC is more than a RAN controller. It is also an open platform that can host
xApps and rApps developed by specialist software suppliers. This has the potential to
stimulate investment RAN control applications and make it faster to deploy
innovation into the live network
- It is early days, however. The baseline specifications are now available and there is a
software community growing up around the RIC, including open source development.
The first RIC use cases are emerging focused on reporting and observability.
- Heavy Reading expects the RIC and xApps/rApps to become progressively more
sophisticated and, in later phases, able to make active changes to live networks. This
will require careful testing and monitoring to ensure the stability and reliability of the
RAN nodes when under the control of the RIC. This is a big change in how RANs
operate, and it will take time for the industry to be comfortable with this.
- In terms of deployment, a regional cloud model is probably optimal in most cases.
The near-RT RIC covers a geographical area similar to that served by a regional
cloud point-of-presence. This deployment offers the potential to co-optimize the
RAN, core, transport, and cloud to deliver superior end-user services than would
otherwise be possible.
ABOUT STL – STERLITE TECHNOLOGIES LIMITED
STL is an industry-leading integrator of digital networks.
We design and integrate these digital networks for our customers. With core capabilities in
Optical Interconnect, Virtualised Access Solutions, Network Software, and System
Integration, we are the industry’s leading end-to-end solutions provider for global digital
networks. We partner with global telecom companies, cloud companies, citizen networks,
and large enterprises to deliver solutions for their fixed and wireless networks for current
and future needs.
We believe in harnessing technology to create a world with next-generation connected
experiences that transform everyday living. With an intense focus on end-to-end network
solutions development, we conduct fundamental research in next-generation network
applications at our Centre of Excellence. STL has a strong global presence with next-gen
optical preform, fibre, and cable manufacturing facilities in India, Italy, China, and Brazil,
along with two software development centres across India and one data centre design
facility in the UK.
To realize the gigabit connectivity for all, STL has been focusing more on the access part of
the network infrastructure. We are offering futuristic 5G and FTTx solutions that are based
on the concept of openness, disaggregation, and virtualization. In the wired connectivity
realm, we are offering a platform that is entirely virtualized OLT and ONT software stack
integrated with white box solutions to support fiber to the home or enterprise. On the
wireless side, we have strategically invested in and partnered with leading companies in the
virtualization domain to offer a COTS-based vRAN solution that can host both 4G and 5G
concurrently, as well as radio units for different deployment configurations. This includes
small cells and macro RUs to meet various deployment archetypes that operators have for
both 4G and 5G. To complement these, we are also offering software platforms such as the
RAN intelligent controller (RIC) and orchestrator—essential platforms for a cloud centric
With expertise in all these capabilities, STL is well positioned to address the needs of telcos
and private enterprises through our Access Solutions portfolio offerings, which are
fundamentally open, disaggregated, and virtualized platforms. To learn more about them in
detail, you may visit https://www.stl.tech/virtualised-access-products.