White paper

Cisco public

Upping the Game with SON

Why today’s networks need machine-learning SON

Why self-organizing networks?

Self-organizing network (SON) technology is widely deployed and has shown it can help
mobile network operators (MNOs) reduce costs and improve performance. Not all SON
solutions are created equal. Some are relatively simple, only managing power for cell sites,
self-healing, and load balancing of neighbor cells. Advancements in SON solutions are
necessary because of increases in network complexity and user service and bandwidth
demands.
An advanced SON solution can simplify and automate your network operation practices.
It improves efficiency because you can automate routine tasks and run 24/7 closed-loop
optimization algorithms. With a SON solution automating operations, you have more time to
do research, innovate, and roll out new technology. In the same number of working hours,
you can get more done with fewer errors.
As the industry moves into the 5G era, advances in RAN automation need to expand
beyond the radio tower and being a reactive system. MNOs need better ways to automate
the radio access network and go beyond existing constraints to assure optimal use of
resources and a good user experience. SON needs to adopt machine learning capabilities
and to draw from a larger pool of data sources.

© 2019 Cisco and/or its affiliates. All rights reserved.

White paper
Cisco public

Contents The need for more SON intelligence

New challenges are arising from the tremendous growth of new data sources
The need for more SON
including customer experience systems, geolocation, crowdsourcing, big
intelligence data analytics, and the Internet of Things (IoT). Because mobility domains
such as packet core, policy, and backhaul are all affected, it drives the
SON moves into the machine-
requirement for an end-to-end automation solution.
learning era
This explosion of network performance and status data is accompanied
Why Cisco Crosswork SON? by a rise in overall network complexity because a multilayered architecture
is required to support new services. These factors contribute to the need
Learn more to bring more intelligence to a SON solution. The solution must be more
scalable and adaptable, so there’s less human intervention during the setup
and maintenance phases.

Classic rule-based algorithms for identifying suboptimal network

performance and problem correction need to be complemented with
algorithms that can identify the relevant information from extremely large
amounts of data. Additional improvement can be achieved by complementing
the “problem detection – response” solution with tools that can predict the
network state in advance before degradation happens. MNOs need to move
from a reactive SON to a proactive solution.

© 2019 Cisco and/or its affiliates. All rights reserved.

White paper
Cisco public

In addition, the predictive capabilities of machine learning Machine-learning predictive algorithms can also be used
can support the need to know about the consequences in situations where a reactive approach is enriched by
of each corrective action a SON may execute. Machine- the addition of a proactive approach. The addition of
learning SON is faster and more effective than classic prediction capabilities without real execution offers the
algorithms; it can select the optimal choice for the potential for more improvement. Prediction isolates the
required impact on network performance. possible consequences of the corrective actions that the
SON could take, such as cell parameter changes.

To become proactive, time series prediction-based

machine learning models are deployed to help foresee
network behavior and performance. These machine
learning models use cell radio configuration as well as
historical performance of the cell and co-related cells
to provide accurate predictions. Knowing about network
degradation in advance and fixing the issue before
performance degradation occurs can greatly improve the
user experience.

SON moves into the machine-

learning era
Operators of modern mobile networks face issues
related to complexity that can’t be solved by
straightforward software logic for all use cases. The
increased complexity and diversity of the networks
range from the diversification of architecture schemes
to the plurality of RAN technologies in use. Fortunately,
because of progress in advanced machine-learning
algorithms, it’s now possible to enhance existing SON
solutions by using more network data and new input
categories.

Machine-learning algorithms can be used where

there is a need to understand patterns in the network.
These algorithms can detect cells with suboptimal
performance, which might not be identified using classic
detection approaches.

© 2019 Cisco and/or its affiliates. All rights reserved.

White paper
Cisco public

Why Cisco Crosswork SON? Cisco’s machine learning SON solution maximizes
the performance gain, reduces human intervention,
Cisco Crosswork SON operates in a closed loop that minimizes errors, and ensures that your network is
includes monitoring, analyzing, adjusting execution, operating at maximum efficiency. This zero-touch
and remonitoring. In addition, it brings the advantage solution increases your network savings, improves
of supporting multitechnology, multivendor RAN the user experience, and enables fast and simple
deployments. Cisco Crosswork SON is an advanced deployment of new services.
SON solution with functions that perform dynamic
adjustment of radio access network (RAN) configuration The machine learning engine will be part of new
and control user bandwidth policy and charging rules Crosswork SON products, that include:
function (PCRF). By taking advantage of these functions, • The SONFlex framework, which allows you to develop
you can further improve service quality and system a variety of applications, such as radio applications for
capacity. RACH, load management, coverage, capacity, mobility
Cisco Crosswork SON suite of RAN automation solutions optimization, and Voice Over LTE (VoLTE) solutions.
is evolving into the machine-learning era. Thanks to • With user-centric SON, you can develop applications
its machine-learning engine, Cisco Crosswork SON that are aware of user experiences and that can
minimizes the impact of rapid changes in the network, address key quality indicators (KQIs) for users.
which could lead to network performance degradation User-centric SON enables new applications for
and poor user experiences. enterprise, private LTE, and high mobility.

The machine-learning engine in Cisco Crosswork SON • Using multi domain service optimization (MDSO),
includes new features that can proactively identify you can develop service-aware applications.
network degradation before it occurs and act to These applications address the quality of
prevents these issues. To detect the behavior patterns experience, service level agreements (SLAs),
hidden in the mass of measurements at the network, and multi domain optimization. MDSO enables
it uses the clusterization of similar network cells. This new SON applications for IoT, video, and 5G.
method identifies cells with suboptimal performance
and analyzes the root causes on these cells to optimize Learn more
performance. The predictive capabilities select the
Cisco Crosswork SON harnesses the increasing
optimal actions to maximize performance. Employing
capabilities of machine learning to incorporate
these advanced algorithms can reduce the level of
sustainable and scalable algorithms to make it more
uncertainty about how the actions would affect the radio
intelligent and adaptable. To find out more, visit
network, as well as reduce the need for multiple steps of
www.cisco.com/go/mobile
corrective action.

© 2019 Cisco and/or its affiliates. All rights reserved. Cisco and the Cisco logo are trademarks or registered trademarks of Cisco and/or its affiliates in the U.S. and other countries. To view a list
of Cisco trademarks, go to this URL: www.cisco.com/go/trademarks. Third-party trademarks mentioned are the property of their respective owners. The use of the word partner does not imply a
partnership relationship between Cisco and any other company.