1) Small cells are proliferating and new business models around managing heterogeneous networks containing both small cells and macrocells are emerging. Operators are exploring strategies like network sharing and outsourcing network management.
2) Adding intelligence at the small cell level through techniques like caching popular content locally and transrating video can significantly reduce mobile network traffic loads and improve user experience by reducing reliance on backhaul bandwidth.
3) Self-organizing networks aim to automate coordination between different cell types and vendors to optimize performance across heterogeneous networks as they increase in scale and complexity.
1. White Paper
WEBSITE: www.jdsu.com/test
Optimizing Small Cells and the Heterogeneous Network (HetNet)
Small cells are proliferating and investment is growing in every aspect. So are the various new business
models surrounding the technology—managed services maintaining the HetNet are being pitched to
various operators on a global basis. In the UK, for instance, Vodafone and O2 agreed to collapse their
cell sites into one network. RAN sharing is becoming common practice, creating macrocell and future
small-cell backhaul infrastructure opportunities shared between operators.
Cable operators’ assets are well positioned to bring unique value to mobile operators seeking to deploy,
fill coverage, or data offload with the development of a small-cell offer, with Time Warner Cable and
Comcast showing interest in this space. Virgin Media UK is already offering small-cells-as-a-service
and this will be an interesting space to watch going forward. With the added benefits that small cells
bring, new charging and policy mechanisms are also being introduced. More importantly, with
backhaul costs to small cells the same as to macrocells, the cost benefit rests heavily on backhaul. With
this in mind, backhaul sharing is moving from an initial thought to definitive business case, and these
are just a few of the business models that have sprung up to date.
Taking a snapshot of the news in the small-cell industry, 2012 wrapped up with Informa’s latest report
showing that almost 98 percent of mobile operators believe that small cells are key to the future of their
mobile networks with capacity coming to the fore. In Europe, for example, Vodafone Greece launched
a location-based service driven by indoor picocells which whitelists traffic generated indoors, and this
will be an interesting market to watch, not least for the impact of macro-economic conditions but in the
value-add that location-based services provide.
On the pricing front, some interesting pricing models have come to the fore. These range from free to
high upfront fees, as shown in the Informa Telecoms & Media table below.
Table1.Pricingmodelsforfemtocellservices.
Source:InformaTelecoms&Media
Market PricingModel DeploymentExamples
Consumer
Add-onsforunlimitedcalling MoldTelecom,Sprint,Optus
Freefemtocell Softbank,Vodafone(GR),SFR
Lowupfrontfee Vodafone(UK)
Highupfrontfee Vodafone(Italy,Hungary),Verizon
Monthlyfee Sprint,Movistar,NTTDoCoMo
Enterprise Highupfrontfee Alloperators
2. White Paper: Optimizing Small Cells and the HetNet 2
Table2.Femtocelldeploymentsbytargetgroup.
Source:InformaTelecoms&Media
The latest forecast from ABI Research shows that outdoor small cells will reach 500,000 units in 2013
and that the 1 W and below small-cell class will exhibit the highest growth, representing almost two-
thirds of unit shipments in 2013, and continue to grow to overtake the higher-power 5-10 W microcell
shipments during 2014.
On the deployment side, the big news was that the number of small cells deployed overtook the total
number of macro cells in November 2012, with estimates that the number of small cells has surpassed
6 million while there are 5.9 million macrocells deployed since inception—and we are just in early stages
ofglobaldeployment.
Table3.Mobiledatatraffic.
Source:Cisco2012VNIreport
As of November 2012, 9 of the top 10 mobile operator groups (by revenue) offered femtocell services. In
Japan, NTT DoCoMo announced the launch of a dual-mode 3G/LTE femtocell, future-proofing their
investment.
TargetGroup NumberofDeployments Examples
Consumer 26 VodafoneUK,AT&T,Cosmote
Enterprise 6 T-MobileUK,NetworkNorway,OrangeFrance
Consumerand
Enterprise
8 VodafoneNZ,VerizonWireless,Sprint
Public 5 VodafoneQatar,SKTelecom,TOTThailand
Rural 1 Softbank(using satellitebackhaul)
Exabytes per Month
2011 2012 2013 2014 2015 2016
12
6
0
0.6 EB
per
month
1.3 EB
per
month
2.4 EB
per
month
4.2 EB
per
month
6.9 EB
per
month
78% CAGR 2011-2016
10.8 EB
per
month
3. White Paper: Optimizing Small Cells and the HetNet 3
Mobile traffic is predicted to grow anywhere from 20 to 50 times over the next 5 years. It is clear that
urgent action is required to meet these demands. For operators to take urgent action, they need to look
deeper, and two data points come to mind. The first is that almost 80 percent of data traffic happens
indoors, while just 10 percent of cells handle almost 90 percent of data traffic.
When you couple this information, it becomes very clear that a multi-faceted approach is needed to
solve this traffic growth. It becomes clear that optimizing those 10 percent of cells, whether macro or
small, will have a major impact to the network capability. When combined with optimizing traffic for
indoor scenarios, this will have a force-multiplier effect.
GettheMostOutofYourMacrocellNetwork
With billions already spent, operators first look at squeezing the most out of their existing macrocell
networks. There are several strategies that are getting attention to this end.
Multicarrier — If a carrier finds that they have available spectrum, re-farming of spectrum offers a cost-
efficient way to increase both coverage and capacity.
Sectorization — Allocating more sectors in one form or another reduces the need for new macrocell
sites. By adding additional sectors, operators are getting increased capacity and, to a lesser extent,
increased coverage but without densification in the macrocell network.
Figure1.Differentsectorizationoptions.
Source:NokiaSiemensNetworks
80percent
ofdatatrafficisindoors
90percent
ofdatatrafficishandledby
fewerthan10percentofcells
4. White Paper: Optimizing Small Cells and the HetNet 4
Tilt Optimization — The third macrocell optimization strategy is antenna tilt, which minimizes
interference. This has the added effect of increasing capacity at a low cost, and as such should always
be pursued before any further optimizations. It is a powerful optimization technique as it has the most
direct impact on coverage and interference parameters of the network. When you couple this with the
advent of remote electrical tilt (RET) antennas, tilt optimization lends itself well to antenna-based self
optimizing networks (SON).
Figure2.Impactofantennatiltoncoverageandinterference.
Source:NokiaSiemensNetworks
Cloud RAN — C-RAN takes its form from a distributed base-station architecture by pooling baseband
processors in a central location and distributing remote radio head ends which are connected
through dark fiber to the baseband engines. The interface between the baseband and the remote radio
has been standardized by a few industry consortia including CPRI (Ericsson) and OBSAI (Nokia).
Figure3.Distributedwirelessbase-stationarchitecture.
Source:FrankRayal.com
This approach provides marginal improvements in capacity and coverage where there is abundant fiber.
In this case, operators benefit from reduced cost, reduced power consumption, smaller footprint, and
lower maintenance cost.
Downtilting reduces
interference to neighbors
Uptilting increases
cell coverage
Outdoor
Antenna
Optical Fiber Cable
OBSAI/CPRI Interface
Remote
Radio
Head
Backhaul
Indoor
Baseband
Processing
5. White Paper: Optimizing Small Cells and the HetNet 5
SmallCells
Operators are shifting their focus to a three-pronged approach to squeezing out more capacity and
coverage:
• Moving the base station closer to the user equipment results in a higher-quality air interface, which
provides better spatial efficiency
• Spectrum increase: more spectrum is being freed up in an attempt to meet demand
• Spectrum efficiency: moving to LTE delivers better spectrum efficiency
Figure4.Athree-prongedapproachtocapacityneeds
With higher signal quality using small cells, more bits can be transmitted at the same time, which leads
to better throughput. When you combine this with new spectrum it has a multiplier effect. Couple that
with the spatial efficiency of small cells and you get the force-multiplier effect of a theoretical 1000x
capacity increase as highlighted in Figure 4. Note that for completeness, other methods from various
vendors get to the 1000x by 10x more performance and 10x more spectrum with 10x more cells.
Apart from the capacity increase, small cells enable:
• Better latency: users will experience faster download and upload times
• In-building coverage: small cells invariably provide better in-building coverage and this can
represent a significant source of revenue for network operators
• Better cell-edge coverage: small cells provide better cell-edge performance than macro cells, resulting
in better quality of experience
6. White Paper: Optimizing Small Cells and the HetNet 6
Informa published a report recently that highlighted the industry’s views on what the important factors
are concerning small cells, summarized in Figure 5.
Figure5.Factorsaffectingsmall-celldeployment
TheSelf-OrganizingNetwork(SON)Imperative
A difficult factor when introducing and managing hetnets is that an operator is potentially facing not
one dimension, but five dimensions of difficulty while trying to offer a seamless user experience.
CoexistwithMacrocells
• Therearemorehandovers,requiring
efficientandhighercapacitytohandle
thehighersignalingtraffic.
• Thereismoreneighbormanagement,
withneighborlistsacrossclustersof
smallcellsandtheirlargercousins.
CoexistwithMacrocells
• Self-organizingNetworks(SON)
capabilitiescoordinatebetweencells,
harmonizeparameters,maximizing
performanceoftheentirenetwork.
• Networkcontinuallymonitorsitsown
performance,traffictypeandsource,
adaptingitselfautomaticallytoachieve
optimalperformance,automatically
comingintoservice.
InterferenceManagement
• Asthe numberofcellsincreaseina
mobilenetwork,therearemorecell
borders,leadingtogreaterpotentialfor
interference.
• Automaticallyadjustitstransmitpower
andschedulingofresourcestoavoid
intercellinterference.
• Powerlevelsneedtobecarefullysetto
balanceinterferenceandcoverage.
Backhaul
• Carefulplanningandperformance
managementisrequiredtoavoidcreating
bottleneckswherecapacityisrestrictedby
insufficientbackhaulupstream.
VendorInteroperability
• Openinterfacesessentialtooptimize
performancebetweenmultiplevendors
• Withsuchcloseinteractionrequired
betweenthedifferentlayersofa
HeterogeneousNetwork,itisimportant
thatopenstandardinterfacesare
implemented.Theseallowdifferent
vendorsproductstobeusedindifferent
partsofthenetwork,sothatthebest
productscanbeselectedfordifferent
tasks.
Other
An operator could try to manage perhaps 100,000 small cells, and 1000s of macrocells, with some on
UMTS, some on LTE, all on different vendors gear, sharing some RAN resources, and leasing backhaul
from another player. How do you connect and manage the network and its issues? This is the big
operational challenge operators are facing when going small.
7. White Paper: Optimizing Small Cells and the HetNet 7
A SON tries to address these challenges. A 2012 Infonetics report highlights the top reasons for
implementing SON in Figure 6.
Percent of respondent operators
0% 20% 40% 60% 80%
Opex reduction
Improvement in capacity, quality,
and network performance
Small cell usage in the network
Figure6.TopreasonstodeploySON
Both the 3GPP and the Next Generation Mobile Network Alliance (NGMN) have working groups
making progress attempting to standardize features and use cases. As an example, some NGMN
requirements are listed below.
Performance management Self optimization
Real-time monitoring Fault management/correction (self healing)
Self configuration (installation) O&M-related SON (extension/upgrade)
Table4.NGMNSONfocus
SON was introduced as a 3GPP release 9 feature, and now there are over 30 use cases for SON. The
breadth of these use cases, in terms of what they address within the network, is vast enough that most
SON vendors focus on only a specific area. It will be interesting to see how the SON market matures as
with the advent of LTE and its heterogeneous nature, perhaps it seems logical that operators might end
up with multiple SON vendors, all part of the same network.
On the SON revenue front, most vendors are still in a pre-revenue stage with the likes of Intucell,
Spidercloud, and Reverb networks leading the charge. Intucell was recently acquired by Cisco for
$475M, showing the importance of this technology in this fast-moving space.
With device-assisted SON, user equipment can collect measurements on network performance.
However, standardization of this feature has been slow, with the access network discovery and selection
function (ANDSF) being the most prominent method for achieving this.
There is a growing school of thought that SON is equivalent to vendor lock-in, which may detract from
the business case for HetNets. Perhaps a hybrid of SON solutions will allow for vendor differentiation—
but at what expense to interoperability?
8. White Paper: Optimizing Small Cells and the HetNet 8
AddingIntelligencetoSmartCells
Early signs in this space can be found in running things like network and application services. Intel
recently conducted a trial on this subject and the results showed that deploying intelligence at the
access point can radically change the traffic profile. This can make the small-cell business case far more
attractive.
A summary of the benefits and methods are shown below.
Figure7.Addingintelligencetosmartcells
If an operator can radically change a traffic profile at a moment’s notice, there are two very compelling
benefits. The first is that you can cut the amount of traffic being backhauled. The second is that the user
experience can be greatly improved with faster web page and video downloads.
There are several ways the industry can achieve this.
Caching Content — Even with exponential traffic growth over the past and future years, analysis shows
that there are many situations where multiple users actually access the same content. Some examples are
popular TV shows, and a prime example is with viral videos such as “Gangnam Style” with over a billion
viewings. Geographically-relevant data such as maps and restaurant guides are similar draws. In these
situations, the network will be inundated with requests for the same content.
Caching can deliver a number of benefits if an operator can predict the content that the user will try
and pull from the network. This is the concept of predictive caching. If the carrier predicts content and
downloads it off-peak, the traffic profile can be evened out during the day, reducing load at peak times.
When the user pulls content, whenever there is a hit, it actually comes from the cache. This saves on
backhaul needs.
Caching
content
Transrate videoWireless policy
management
Cuts backhaul traffic Improves the user experienceBenefits
Method
Determiningwhat content
users will most likely
want to download
Managedoffload
Localtransrate in
the small cellHow
Running apps
closer to users
improves QoE
Caching and
offloading cut
backhaul costs
9. White Paper: Optimizing Small Cells and the HetNet 9
SMALL CELL
Predictive
Caching
Proactive Caching
Users access the same content suchas popular TV shows,
YouTube videos, sports, maps …
• Carrier predicts the content and
downloads it off-peak
• Evens out traffic profile during
the day, reduces load at peak
times
• Content downloaded once and
cached for future requests.
• Reduces backhaul traffic all day
User requests
content
First play direct
fromYouTube
Subsequent plays
from the cache
Carrier predicts content,
downloads prior to request
First and subsequent
plays from cache
The other method of caching is proactive. Here, the user downloads content from the network and, once
content is downloaded, it is stored as it traverses the small cell. Any subsequent request for that content
will be served by the small cell’s cache. This process is described below.
Figure8.Content-cachingmechanisms
Transrating Video — One forecast predicts that by 2015, video will consume 90 percent of mobile
traffic, and it surpassed 50 percent just last year. Any intelligence that can be added to optimize this
bandwidth hog would have far-reaching benefits. To this end, content-aware technology is making big
inroads into mobile operators’ strategies. Content awareness brings valuable visibility by analyzing
and understanding a packet’s contents. It also recognizes the type of application or service to which
a packet belongs. For example, the technology has the ability to see the header and the payload, most
notably Layer 4 through Layer 7. The data contained in these layers is then adapted to the user device
and modified to optimize delivery. Video is one application that can benefit from content awareness.
Inspecting packets from Layer 4 through Layer 7 reveals valuable information for mobile network
operators.
Transrating comes in a couple of different variants. The first is where the codec is transrated in a
dedicated node at the edge and then decoded right at the small cell, reducing the backhaul bandwidth
needed. Bear in mind that reducing the peak load has the greatest impact on reducing backhaul costs.
Transrating reduces a video down to a lower bit-rate codec while minimizing the impact on the video
quality. More prevalent these days is doing the transrating on the fly, meaning adapting to network
conditions (intelligent transrating). Figure 9 highlights different transrating mechanisms.
10. White Paper: Optimizing Small Cells and the HetNet 10
Figure9. VideoTransratingMechanisms
Another transrating method is where the node is still at the edge encoding on the fly, and if a
smartphone is sufficiently powerful and supports the same codecs, the transrated video is delivered to
the smartphone directly without the smart cell taking the burden of transrating.
On the business front, a recent Tellabs report forecasts that transrating can save 30-50 percent of needed
video bandwidth; this amounts to a considerable savings for operators’ backhaul costs. Later, we discuss
managed offload, where Internet traffic such as video is routed away from the core of the network and
sent through other means such as a home owner’s Internet connection.
SmallCellBackhaul
While deploying large numbers of small cells near to consumers helps solve the capacity and coverage
problem for the RAN, it also creates a new one for backhaul, which must provide connectivity at
sufficient capacity and quality of service. Backhaul is perceived as the most critical factor for a small-cell
platform, followed by the ability to self-optimize and cooperate with the macrocell network.
Figure10.Barrierstoentryforsmall-cellbackhaul
Working with building owners for site,
power, backhaul network connections
Percent of operators rating it a barrier
0% 10% 20% 30% 40% 50%
Source fiber backhaul connections
Determining if Line of Sight (LOS) or
Non-LOS (NLOS) is needed for backhaul
Availability of suitable backhaul
product form factor (size, color, shape)
Content-AwareVideo Bit Rate
Throttling
SmartVideoTransratingfor
BandwidthReduction
• Delivers video consistent withthe viewingrate
• Analyzes the encoded video stream andestimates the video content in the buffer of the subscriber device
Content-AwareVideoTransrating
• Frames and bit rate inthe video stream are analyzedandreduced without noticeably degradingviewingquality
• Network operators cansupport more users without affecting QoE
Device-AwareVideoTransrating
• Reduces the bit rate of the video to fit the specific device, saving bandwidth
Network-AwareVideoTransrating
• Real-time RF conditions are estimatedand the video bit rate adapted to matchvaryingbandwidth conditions
• When network conditions are poor, the bit rate is reduced to minimize or eliminate potential stalls while
still maintaininggood video frame quality
12. White Paper: Optimizing Small Cells and the HetNet 12
With the exception of the last 100 meters, small-cell backhaul has more in common than not with
macrocells. Key performance indicators for small cells should ideally be the same as with the macrocell
network. Apart from reusing existing POPs, other existing network resources can also be leveraged.
Small-cell traffic can be routed to the same concentrator in the network that the macrocell network
already serves. Typically, the additional traffic from the small cells or the fact its origination differs
doesn’t require additional switches or routers, at least on first pass. A recent Infonetics study shows that
86 percent of operators surveyed plan to backhaul small-cell traffic to nearby macrocell sites via a variety
of locations including buildings, streetlights, and traffic and utility poles.
Copper Microwave Fiber
North America 85% 5% 10%
Europe 25% 65% 10%
Asia 10% 40% 50%
Table6.Currentregionalbackhaulmethods
While fiber provides the most bandwidth, it cannot be cost effectively pulled to every lamp post,
at least in many markets. Therefore, various forms of microwave, non line of sight (NLOS), standard
microwave, and millimeter wave, will most often be the solutions of choice.
New,lower-costwirelessbackhaulproductswithnewfeatureswillbeneededtosupportsmallcells.Existing
microwave frequencies in the 6-42 GHz band cannot support discrete antennas, at least at the street level.
New frequency bands are being considered for wireless backhaul such as 3.5 GHz, 60 GHz, and 80 GHz.
Thisdoesnotfactorintheneedfornear-andNLOS-propagationcharacteristicsinrelationtosmallcells.
Figure13.Small-cellRFbandsofinterest
Source:RFMD
With public-access small cells typically needing only to support Ethernet interfaces, and the fact that
typically, small cells support a maximum of 20-30 subscribers compared to the low 100s for macrocells,
small-cell aggregation routers will inevitably reduce in complexity, power requirements, size, and,
most importantly, cost—a huge factor in the small-cell business case. Heavy Reading, for example, is
forecasting a reduction from a current 1U/2U solution down to <0.5U in the short term.
Iub vs. Iuh is also an important consideration, with much of the femtocell industry adopting the 3GPP
standard luh interface for linking femtocell access points to the service provider’s network. Ericsson’s
perspective is that all small cells, including femtocells, picos, and microcells should be linked by the Iub
interface, which allows them to be integrated completely into the network just like macrocells, claiming
less interference as it enables an operator to use the same spectrum for a small cell and a macrocell.
However, the Iub interface was not popular among some femtocell proponents, particularly Nokia
Siemens Networks, because its implementation is proprietary across vendors: you can’t attach a small
cell over an Iub from another vendor.
<6 GHz
NLoS LoS
~6-42 GHz
V Band E Band
60 GHz 71-76, 81-86 GHz
13. White Paper: Optimizing Small Cells and the HetNet 13
InternetOffload
Internet offload is an important response to the growth in mobile traffic. The industry has, in fact, a long
history of embracing technologies and approaches that could be categorized as Internet offload. 3GPP
has done a lot of work with this approach over the years, starting back in release 6 with specs such as
LIPA and SIPTO. Internet offload comes in several forms: WiFi, femtocell, and core network.
Figure14.Internetoffloadstandards
WiFiOffload
Ericsson’s acquisition of the carrier-grade WiFi provider BelAir networks shows how important WiFi is
to the industry. If WiFi integration is coming, the question for mobile operators is how they will adapt.
Operators may continue to let the user-driven WiFi model prevail, where people offload on their own
without much operator involvement. Or, they may take a more active role and offer managed offload
solutions where they have more control over how, when, and what traffic is offloaded.
Operators are taking a long, hard look toward tighter integration of WiFi with cellular services. The
first phase of deployment is that of hard offload. This is the current situation today, where a user simply
roams into their home and content automatically offloads. The second phase is already being used by
some operators such as T-Mobile and Orange, where the offload is based on SIM-based authentication.
In phase three, there will be tighter integration between WiFi access and the core, offering seamless
session mobility. This is where the industry seems to be heading.
FemtocellOffload
Femtocell offload differs from WiFi offload in two ways: femtocells are deployed in licensed spectrum
and they are fully integrated with a carrier’s network. The core network integration is important because
it means that the femtocells are transparent to all operator services. With femtocell offload, Internet
traffic can be selectively offloaded through the owner’s general internet connection.
I-WLAN
UMA/GAN
Direct
Tunnel
ANDSF
Femtocell LIPA
SIPTO
IFOM
3GPP Rel 6 3GPP Rel 10
WiFi Offload Femto Offload Core Network
Offload• Current user driven
connections toWiFi
• Next wave is SIM-based
Authentication(T-Mobile,
Orange)
• Next will be whereWiFi
access is integratedinto
the Core (sessionmobility)
• Coverage& Offloadmain
drivers
• A push to integrate with
WiFi AP
• LIPA / SIPTO
• Packet Switch Offload by
deployingOffload
Gateways behindRNC
• RAN AwareTrafficshaping
14. White Paper: Optimizing Small Cells and the HetNet 14
CoreNetworkOffload
Core network offload reflects selected IP traffic offload (SIPTO) standards. The intent is to distribute
packet gateways so that traffic is not concentrated on a handful of nodes.
Iu-PS offload involves deploying Internet offload gateways behind an RNC or group of RNCs. This has
the effect of splitting out Internet traffic bound for the operator’s core network. In the same vein, the
industry has started to leverage this architecture to push content caches and content optimization closer
to the end user not just in the small cell itself, but at the edge of the core where greater control can be
achieved. This is where an offload gateway is served either directly from the Internet (for example, from
an Akamai server) or from a local server such as a mobile CDN collocated with the gateway.
An unintended consequence of core network offload is that by diverting traffic from the core, it
becomes harder for the operator to meter usage, bill for traffic, and apply traffic management. This is
also an issue for content caching. A proposed solution is to make the offload gateway also function as
a traffic management device, which means integrating deep packet inspection and policy enforcement
capability mirroring the core network capability. For this architecture to be optimized, the traffic
management function needs to be increasingly RAN aware. However, this has the added consequence
that the offload gateway should have access to standardized RAN and policy management interfaces
which are still being standardized. This lets it adapt dynamically to load conditions on the network.
Summary
Optimizing current macrocell and future heterogeneous networks requires a multi-dimensional
approach. It begins with an operator cost-efficiently optimizing an existing macrocell network. The next
step is densification of the network with the addition of small cells to the existing infrastructure.
Capacity and coverage are driving this approach, and further enhancements to the resulting HetNet
should be made with the goals of cutting backhaul costs and improving the customer’s quality of
experience. The combination of using small cells and turning them into smart cells has a force multiplier
effect both on backhaul saving and quality of experience.
The importance of SON is highlighted by the multi-dimensional issues associated with densification
and HetNets. There is some industry standardization work going on, with the goal of reducing small-cell
backhaul with techniques such as offload perspective as well as local caching, video optimization, Wi-Fi
integration, and various Internet offload techniques.
If an operator is to improve a customer’s quality of experience while managing the exponential growth
in their network’s complexity, they will need a highly-integrated solution that is fully adaptable to
network conditions as they happen. Adding intelligence to the network and having RAN resources
direct their efforts to where it’s needed is fast becoming a necessity where once inflexibility ruled.