Analysys Mason Hybrid S

White paper

Hybrid SON RoI in realistic

deployments of mixed 3G/4G

networks

February 2012

Mark H Mortensen (Principal Analyst, Analysys Mason)

Gerry Foster (Director of Technology Engineering,

AIRCOM International)

Ref: RX952

.

Ref: RX952 .

Contents

1 Executive summary 1

2 Recommendations 2

3 The promise of SON 3

4 SON has become both a 4G and 3G story 4

5 SON architectures 5

6 Quantification of SON benefits 7

6.1 Benefits from distributed algorithms 7

6.2 Benefits from higher-order control algorithms 7

6.3 Costs of SON software and network equipment and implementation 8

7 Return on investment under slow and fast roll-out scenarios

in hybrid SON architecture 9

7.1 Slow-roll UMTS and LTE data evolution scenario 9

7.2 Fast-roll LTE scenario 12

7.3 Cost and scope input assumptions 12

7.4 Benefits and capabilities not included in either scenario 12

8 Conclusions from the two RoI models 13

9 Implications for network planning systems 14

Annex: About the authors

Hybrid SON RoI in realistic deployments of mixed 3G/4G networks

Ref: RX952 .

Copyright 2012. Analysys Mason Limited and AIRCOM International (the authors)

have co-written the information contained herein. The ownership, use and disclosure of this

information are subject to the Commercial Terms contained in the contract between

Analysys Mason Limited and AIRCOM International. AIRCOM International has the right

to publish the information, and Analysys Mason Limited additionally maintains the rights to

the information and material for its own use.

The opinions expressed are those of the stated authors only. The authors recognise that

many terms appearing in this document are proprietary; all such trademarks are

acknowledged and every effort has been made to indicate them by the normal UK

publishing practice of capitalisation. However, the presence of a term, in whatever form,

does not affect its legality as a trademark.

The authors maintain that all reasonable care and skill have been used in the compilation of

this publication. However, the authors shall not be under any liability for loss or damage

(including consequential loss) whatsoever or howsoever arising as a result of the use of this

publication.

Hybrid SON RoI in realistic deployments of mixed 3G/4G networks | 1

Ref: RX952 .

1 Executive summary

Operators are beginning their 4G LTE deployments. As the investments needed are large,

operators have therefore become much more tactical in their LTE deployments, often upgrading to

UMTS technologies such as HSPA and HSPA+ to provide additional data capacity. The costs

associated with LTE deployments are driving an increasing number of operators towards

infrastructure sharing, such as physical towers, or even more intimate sharing arrangements.

Self-organising networks (SON) features, which allow for the automation of several tasks in

mobile networks, are being standardised by 3GPP with a goal of reducing LTE opex costs by 60%

while optimising network operation, equipment use and user experience. All leading equipment

vendors are proceeding with their plans to provide SON functions in their LTE offerings, and now

also their 3G UMTS equipment, while leading network planning and optimisation vendors are

creating their own products for adding SON features onto mobile networks.

This white paper discusses a model of the likely benefits that can be realised in a realistic network.

In the base model, the operator introduces SON features (in a hybrid architecture) in both UMTS

and LTE over a period of three years, while LTE is minimally deployed and UMTS capacity

augmentation primarily meets the increasing data needs. The paper also outlines the additional

benefits that can be gained through an aggressive LTE roll-out.

The model finds that the total cost of network deployment and operations can be

reduced by 25% or more through the deployment of a hybrid SON architecture,

leading to an EBITDA 5% larger than if the SON features were not implemented.


Ref: RX952 .

2 Recommendations

Network operators should implement SON features in their 3G and 4G networks to decrease

their opex and minimise their capex during data capacity enhancement projects. The benefits

of SON to the dominating 3G portion of their networks can be several times larger than those

for the LTE portion of their networks during the next few years. So while adding data capacity

to their networks, they should convert the 3G portion of their network to SON operations,

while deploying SON in all LTE additions and conversions. Network implementation of SON

additionally improves and unifies network operation across UMTS and LTE, which also

improves SingleRAN optimisation.

Adoption of SON Energy Saving represents a large opex benefit opportunity, in terms of

electrical power saving. Additionally, for the base stations included in the ES group, MTBF

is improved as a result of RX952overall reduction in load per unit time. However, ES adoption

requires that a number of functions be implemented, both in the network as well as in

supporting external SON control systems, such as the ability for ES stimulated T0 re-

planning.

SON features should be implemented to provide both early deployment semi-manual open-

loop operation for initial set-up, tuning and validation and control capabilities along with the

ability to later switch to full automated control. A good hybrid SON system should:

monitor distributed SON behaviour.

perform sequences of actions automated by the network engineer.

report on the status of previously implemented automatic recommendations.

provide proposals for recommendations not yet automated for selective semi-manual

control by the network engineer.

As the system deployment is tuned semi-manually, the network engineer should, over time,

activate further automation controls as repeated good recommendations are proved to be

reliable.

Communications service providers (CSPs) should determine the optimal LTE vs. HSPA and

HSPA+ implementation paths. These should take into account the marketing needs, the

spectrum available, congestion in the cell sites, and the increased benefits of SON for LTE.

With the already large SON benefits for LTE being greater than when applied to 3G, the

additional SON savings can also help CSPs pay for LTE upgrades.


Ref: RX952 .

3 The promise of SON

In 2007, mobile CSPs and vendors were starting to look at the long-term evolution of 3GPP, which

has led to the current 4G LTE standards. All parties agreed that they wanted significantly more

capacity for less cost overall, but also were adamant about two requirements:

the system had to be cheaper per bit than UMTS to buy and implement

the system had to cost significantly less to run.

The second of these requirements gave rise to features for networks to self-organise and self-

optimise. Thus, the Self-Organising Networks (SON) Work Item 3GPP Release 7 was born,

which has now been propagated forward and extended through Releases 8 to 11. The goal was a

nominal 60% reduction in LTE opex during deployment and ongoing operations.

Figure 3.1: SON

benefits during the four

steps of the mobile

network lifecycle

[Source: Analysys

Mason, 2012]


Ref: RX952 .

4 SON has become both a 4G and 3G story

Equipment manufacturers began back-porting SON features to 3G technology in earnest in 2010,

as CSPs began to understand that to meet the rising demand for data, it could be more cost-

effective for them to expand HSPA and HSPA+ high-speed data capacity on the existing 3G

infrastructure in many locations. Installation of the latest technology, LTE, would be reserved for

targeted locations where the greater spectral efficiency was required or where there were

marketing imperatives to provide significantly more data capacity per unit area.1 In addition,

applicable SON features are being added to the 3G UMTS equipment.2

CSPs are actively engaged in these data upgrades, as shown in the table below from

Analysys Masons Wireless Networks Tracker:3

Technology

upgrade

Planned In deployment or

operational

Figure 4.1: CSP projects

planned vs. in

deployment or already

operational, worldwide

[Source: Analysys

Mason, 2012]

LTE 149 49

HSPA+ 31 79

HSPA 7 169

1 See Karapandi, H. and Norman, T., Operator strategies for network evolution: the road to LTE, Analysys Mason

(Cambridge, 2009) and Norman, T., Will HSPA+ upgrades be critical as operators develop network capacity?, Analysys Mason (Cambridge, 2010).

2 In particular, ANR and scrambling code planning.

3 See Wireless Networks Tracker, 21 September 2011, Analysys Mason (Cambridge, 2011).


Ref: RX952 .

5 SON architectures

The SON standards were defined with three architectural options for their deployment:

Distributed: where the SON algorithms are deployed in the base station/controller. All

equipment manufacturers are implementing distributed algorithms for a subset of the SON

features. Different equipment vendors offer different distributed algorithms but all recommend

that as a minimum for LTE, automatic self-configuration (SC), physical cell identity (PCI) and

automatic neighbour relation (ANR) optimisation are essential distributed algorithms for early

deployment of LTE base stations and femto-cells.

Centralised: where the SON algorithms are deployed above the network infrastructure in, for

instance, a server connected to the northbound SON interfaces defined by 3GPP. IBM and

other software vendors, as well as several equipment manufacturers, in particular NSN, have

been advocating such systems.

Hybrid: where the two other techniques are combined in a two-stage control architecture. A

hybrid SON algorithm is located at the operations and maintenance centre (OMC) or a server

sitting northbound of the OMC, as in the centralised case. It manages the distributed

algorithms in synchronism with the centralised algorithms by monitoring and controlling them

over a dedicated northbound (Itf-N) SON interface to the base stations via the OMC. Currently

Itf-N interfaces are defined for several distributed algorithms such as ANR, PCI and energy

savings (ES). The limiting factor for good hybrid solutions is the adoption and exposure of

these interfaces they are well defined, but poorly supported, on much of the equipment

currently available.

Distributed algorithms typically run fast, as they are deployed close to the radio network, but

influence fewer cells per algorithm, while the centralised algorithms must run slower but can

oversee a much greater scope of cells at once and can, in some cases, manage a whole network.

The hybrid approach has been gaining in popularity among CSPs. In a hybrid architecture, the

manufacturer deploys distributed algorithms at the base stations and provides a set of open

interfaces to allow the CSP to use a control algorithm that operates on a wider geographic area on

a slower time scale. This control algorithm can also:

Provide an open loop control structure whereby reconfiguration options are proposed by the

hybrid SON system and approved (or modified) by engineering personnel. Such a control

structure is important to provide until the automatic algorithms have proven themselves both

in normal and adverse conditions although it is believed by the authors that some level of

human control will always be required.

Oversee networks from multiple vendors, gluing together the edges of the two networks and

reconciling the probable two different distributed SON vendor optimisation algorithm sets.


Ref: RX952 .

Provide an initial configuration for the distributed algorithms, since most of these algorithms

do not work well until there is sufficient traffic to feed them to make good optimisation

decisions. The control algorithm would inject plans for initial T0 configurations for ANR,

for instance, and also recognise when a distributed algorithm is not doing what it should. The

control algorithm would provide oversight to the distributed ANR and to intervene as required

to provide resets and configuration adjustments and, in extreme cases, a new T0 plan.

Track and monitor live performance data from the distributed ANR/PCI algorithms and use

this data to feed into the hybrid energy saving algorithm. This enables the ability to drive ANR

and PCI plans not just for a T0 reset case, but according to demand during different times of

day when the energy saving algorithm is seeking the best cells and PRB slots to turn off during

periods of low load.

As SON is evolving, in practice, operators and vendors alike offered many different perspectives

on the mobile infrastructure market and associated merits of distributed and centralised algorithms.

But most agree that a hybrid solution offers the most beneficial approach. In considering different

roll-out scenarios, this white paper assumes a hybrid SON architecture.


Ref: RX952 .

6 Quantification of SON benefits

The benefits of a hybrid SON system can be divided roughly into those that come from the

distributed algorithms, embedded deeply in the network equipment, and those that come from the

higher-level control algorithms.

6.1 Benefits from distributed algorithms

About two-thirds of the total SON benefits in LTE come from the distributed algorithms usually

embedded in the equipment.

Distributed ANRs

Currently, about half of the network configuration changes in a modern mobile network involve

changes to nearest neighbour cell relationship parameters. Automating these can significantly

decrease the engineering and configuration labour.

PCIs for LTE

PCIs uniquely identify a cell. SON functions can automatically assign the identifiers and

determine neighbour associations between a newly added cell and existing cells.

ANR and scrambling code planning for UMTS

ANR for UMTS manages the UMTS neighbour relations and is usually associated with a

complementary primary scrambling code planning algorithm. These algorithms are automatic

versions of what a planning tool does for an initial roll-out plan or T0 plan. These native

distributed algorithms can be combined in a hybrid set-up with an oversight algorithm to provide a

cooperative capability to reset the distributed ANR algorithm per cell. This would be required

when the ANR distributed algorithms have moved out of normal operational bounds. The error

message would be sent over the Itf-N interface to the oversight algorithm which would take

control.

6.2 Benefits from higher-order control algorithms

About one-third of the SON benefits in LTE comes from the higher-order control algorithms in the

hybrid architecture. These benefits come from a combination of increases in staff productivity,

decreases in visits to the cell sites, and a revenue contribution from decreased churn due to better

quality of service.


Ref: RX952 .

Coverage and capacity optimisation

Coverage and capacity optimisation aims to maximize system capacity and ensure there is an

optimum overlapping area between adjacent cells. Parameter settings are optimised by

cooperatively adjusting antenna settings (manually or, in the case of SON, automatically under

software control such as Remote electrical tilt RET, or Remote azimuth steering RAS) and

pilot(UTRAN)/reference(EUTRAN) power among the related cells. After initial deployment, these

and other coverage and capacity optimisation algorithm techniques should obviate the need for one

to two site revisits per year, for a notable fraction of the population of deployed base stations.

ANR oversight

Early deployments have demonstrated that there will be some cases where automated cell planning

(ACP), ANR and PCI will not always work perfectly. This is true for HSPA+ upgrades as well as

LTE secondary passes will most likely be required. To address these secondary passes at cells

and their neighbours, oversight by an ANR planning control algorithm is seen as essential to check

where the (fast and small scope) distributed algorithms are not able to operate perfectly due to

their limited network visibility. In this analysis, we assume such an oversight algorithm that can

monitor and control those cells where distributed algorithms are not working ideally. It would

supplement information from the cells with performance information from the equipment and

external sources, combining these inputs to provide further cost savings. The savings would come

from the reduction of secondary site visits after initial minimal installation set-up deployments and

also saving software adjustments to the basic ANR/ACP distributed algorithms.

Macro energy savings

During periods of low utilisation, shutting off individual cells and filling in coverage with nearby

reconfigured cells can reduce power consumption by as much as 25%, with an overall effect of

about 2% reduction in energy consumption for the network. With the cost of energy rising, and

already constituting 4% (developed markets) to 8% (emerging markets) of the annual opex costs,4

the monetary savings can be significant.

6.3 Costs of SON software and network equipment and implementation

Implementing SON on existing 3G system for 10 000 cells is assumed to require software totalling

about EUR1.3 million with an additional EUR1500 for each further 10 000 cell addition. Normal

UMTS and LTE coverage and implementation costs were assumed, as used in current engineering

models for network planning.

4 See Norman, Terry and Viola, Catherine, Transform the economics of your wireless business with infrastructure

sharing, Analysys Mason (Cambridge, 2010).


Ref: RX952 .

7 Return on investment under slow and fast roll-out scenarios

in hybrid SON architecture

Two scenarios were considered, a slow roll where the required increase in data capacity comes

primarily from expansion of the existing UMTS network and a fast roll in which LTE is

vigorously deployed to meet the increase in data capacity.

7.1 Slow-roll UMTS and LTE data evolution scenario

The slow-roll scenario is for a three-year deployment programme that enhances data capacity by

20% per annum mostly through UMTS upgrades with modest LTE deployments in about 20% of

the cell sites (but providing much less than 20% of the available capacity). It also assumes a

deployment of hybrid SON across the UMTS and LTE technologies.

Network deployment costs are shown for a theoretical large US CSP and a number of other CSPs

we have modelled of varying sizes and markets. Shown are the total costs to implement and

support the network over the three-year study period, both without SON and with SON. In most

cases, SON decreases the system implementation costs by about 25%, with the benefit to the

UMTS portion of the network being two to four times greater than the LTE portion.

Based on average modelling characteristics of mobile UMTS and LTE networks, the benefits of

SON are substantial, as shown below.

LTE SON benefits

The net benefits of SON for the LTE fraction of the network come from a combination of energy

savings, automated cell and network configuration, and from features providing human oversight

and control over the automated system. The decrease in total network yearly capex and opex,

comes from the areas shown in Figure 7.1.


Ref: RX952 .

ANR/PCI Distributed

56%

Hybrid ANR Oversight

18%

CCO7%

Energy Saving19%

Figure 7.1: Relative

benefits of SON for LTE

networks [Source:

Analysys Mason, 2012]

UMTS SON benefits

UMTS equipment is not as energy-efficient as new LTE equipment, so the benefit of energy

savings is much greater proportionally, but there is little benefit from the ANR SON feature that is

so valuable in the LTE-based networks. The net effect is that for UMTS equipment, SON features

provide somewhat less benefit as for LTE.

Note that the ANR/SC SON benefit for UMTS is projected to be low since most UMTS

deployments have fairly robust and stable scrambling code plans.

ANR/SC distributed

1%

Hybrid ANR oversight

21%

CCO35%

Energy saving43%

Figure 7.2: Relative

benefits of SON for 3G

UMTS networks

[Source: Analysys

Mason, 2012]

Applying these benefits to a model of various CSP networks provides a good view into the

potential for SON.


Ref: RX952 .

0

2000

4000

6000

8000

10000

12000

14000

16000

No SON With SON

EU

R m

illio

ns

Network opex

Network capex

Figure 7.3: Benefits of

SON on total three-year

network opex and

capex for a large US

CSP in a slow-roll

model [Source:


0

200

400

600

800

1000

1200

1400

1600

1800

2000

LargeEurope

MediumEurope

SmallEurope

LargeAPAC

MediumAPAC

EU

R M

illio

ns

Additional capex and opex w/o SON

Total capex and opex with SON

Figure 7.4: Capex and

opex three-year savings

from deploying SON for

several other model

operators [Source:


Comparing the costs with and without SON leads to the following conclusions in the slow-roll

scenario:

Without SON, total network costs (capex+opex) would be at 25% higher, or more, for

most operators.

For large operators in Europe, this figure can be as large as 50% as economies of scale

dominate in dense subscriber regions.

For a medium European operator this figure is lower because of decreased economies of scale.


Ref: RX952 .

For large US operators, the figure is close to the average for all regions and sizes of operators

as the areas of scope are very large, but the gains balance out since there are still localised

dense areas of population, and hence increased workforce savings.

Capex savings are about one-third of the opex savings in most markets, except in APAC where

staff costs are lower, decreasing the opex savings to only about twice those of capex.

Total EBITDA increase over a period of three years from implementing hybrid SON in a

realistic network is about 15% of a years EBITDA, or a +5% EBITDA increase per annum,

on average. However, as mentioned earlier, savings are typically greater for the larger

operators.

7.2 Fast-roll LTE scenario

With the benefits of SON for LTE being a bit larger than for UMTS, we further modelled a fast-

roll case with no UMTS upgrades. The analysis assumed a large US operator, rolling out LTE

with and without hybrid SON to produce a 75% LTE overlay population coverage within three

years with the additional data capacity provided exclusively by LTE (i.e. little expansion of UMTS

data). The model shows that just deploying LTE to make up the data capacity shortfall provides a

slightly better result than slow growing both UMTS and LTE data capacity.

However, when applied to other operators adopting the same fast LTE-only growth policy, this

scenario had far fewer benefits. This came from having a higher population density for the

country, as compared to the USA.

7.3 Cost and scope input assumptions

This analysis has used available operator financial end-of-year reports from 2009 and 2010 that

quote revenue and EBITDA, by business faction per country in order to assess costs and cost

savings. The analysis has taken into account the available data sets in the public domain that detail

numbers of base stations per operator for existing GSM and LTE base sites and cells of various

size classes. The analysis uses AIRCOM Internationals consultancy and permanent staffing costs

from its extensive outsourced engineering business. Lastly, the analysis uses site expansion and

new deployment costs from previous Analysys Mason reports to assess LTE and UMTS costs.

7.4 Benefits and capabilities not included in either scenario

The SON savings presented in this whitepaper include a small component of revenue retention due

to an increase in the quality of service provided by hybrid SON for the UMTS network. Additional

potential CSP revenue from the higher data rates available from LTE is not included.

Also, the further beneficial effects of SON for micro-, pico- and femto-cells have not been

included in these scenarios.


Ref: RX952 .

8 Conclusions from the two RoI models

The analysis presented in this whitepaper estimates that adopting a minimum of three coordinated

SON algorithms in a hybrid architecture for ACP, coverage and capacity and ES, in conjunction

with additional network growth for data capacity over the next few years, yields a net benefit of

about 5% EBITDA per annum.

Demographics and cost of staff per region need to be assessed carefully to decide the best UMTS

and LTE roll-out combinations per region, size of operator and ARPU per data user.

Whilst this analysis clearly recommends ES as the obvious SON algorithm to adopt for best

benefit, ES is not practical without implementing the other algorithms first.


Ref: RX952 .

9 Implications for network planning systems

The benefits of the hybrid SON model are 50% higher than those of the distributed model alone. In

addition, the hybrid SON architecture is superior because of the open-loop possibilities, the ability

to integrate with external performance systems, and the multivendor capabilities that can be

implemented.

The effect on the network planning systems market will be profound over the next five years. The

market will move away from being merely toolsets and manually driven processes, to being more

automated systems that incorporate performance, planning, policy and real-time control.

eNodeB

Neighbour

Metro

Region

sec. min. hours days months

Hybrid

SON

Network

Planning

Distributed

SON

Figure 9.1: Depiction of

the primary areas of

control for distributed

and hybrid SON, and

network planning

[Source: Analysys

Mason, 2012]

Some vendors, such as Nokia-Siemens Networks, are planning on making available all of the

distributed and multi-vendor hybrid SON capabilities itself. But it remains to be seen how that will

play out since, traditionally, multi-vendor networks have been planned and operated using tools

from independent software vendors (ISVs) that specialise in this area, not vendors that usually

advantage their own equipment.

Also, potentially, simulations will become more important as operators attempt to understand the

effects of new policies on their network investments and capacity before deploying these policies.

Of particular interest will be the cases where policy management techniques are employed to

customize the service for individual customers or customer groups. These policies will interact

with the general SON optimisation policies. Such coupled non-procedural systems are known to

exhibit behaviour that is difficult to predict, requiring simulations before new policies are

implemented or changes made, with monitoring and oversight controls afterwards.

Hybrid SON RoI in realistic deployments of mixed 3G/4G networks | A1

Ref: RX952 .

Annex: About the authors

The RoI model presented here and much of the report on the model and its implications is the work of

Gerry Foster, Director of Technology Engineering at AIRCOM International. Mark Mortensen,

Principal Analyst at Analysys Mason, worked with AIRCOM International to validate and extend the

model.

Dr Mark H Mortensen (Principal Analyst) is the lead analyst for Analysys Masons Customer

Care and Service Fulfilment research programmes. He is an expert in personalised multi-channel

CRM, the interaction of BSS and OSS systems with complex networks, and provisioning of

network and service layers. The first 20 years of Marks career were spent at Bell Laboratories,

where he specialised in enterprise-wide systems architecture and strategy, leading teams that

brought new software products to new markets and network technologies, and the interaction of

software systems with the underlying network hardware.

Mark was Chief Scientist of Management Systems at Bell Labs, has been president of his own

OSS strategy consulting company, CMO at the inventory specialist Granite Systems and a

network planning company, and VP of Product Strategy at Telcordia Technologies. He is also an

adjunct professor at UMass Lowell in the Manning School of Business, specialising in strategic

management of high technology companies. Mark holds an MPhil and a PhD in Physics from

Yale University and has received two AT&T Architecture awards for innovative communications

software solutions.

Gerry Foster, Director of Technology Engineering, joined AIRCOM in January 2008 and has

introduced several new network planning and optimisation product capabilities to the business. As

Director of Technology Engineering, Gerry works closely with the CTO and product teams to

bring new technology to AIRCOM products by leading the front-end requirements work.

Gerry has 25 years of experience in the field of mobile communications, specialising in

architectures and protocols. He spent the first five years of his career in RF/microwave component

design, and then moved to the utilities industry for five years, designing communications systems

to support technologies such as PMR data, SCADA and underwater systems. Since 1999, Gerry

has been in the mobile communications industry working on GSM/GPRS evolution at Lucent,

UMTS RNCs and Core network elements at Motorola/Huawei and now UMTS/LTE network

tools & services evolution at AIRCOM.

Gerry has led communications development, systems architecture and analysis teams and

contributed to ETSI, 3GPP and IETF standards. Gerry has a degree in Communications

Engineering, and has been a Chartered Engineer with the IEE/IET since 1989. He continues to

author patents in a number of different communications fields.

Analysys Mason Hybrid S

Documents

authors hybrid son roi

son architectures

son functions

quantification of son

promise of son

g lte deployments

hybrid son architecture

costs of son software