You are on page 1of 23

whI Popr

Analysis of Traffic Offload : WiFi to Rescue


Overview
Using WiFi as an alternative access network technology is not a new phenomena. Since mid
1990s, mobile industry has been looking at using WiFi as a complementary access
technology. Early interworking models relied on Authorization Authentication Accounting
(AAA) integration and WiFi access network and Packet Core connectivity. Later on with the
advent of dual-mode handsets, WiFi has become an access network alternative primarily to
enhance coverage in doors. More recently with the introduction of smartphones, WiFi has
become an essential wireless networking technology for mobile devices. Considering mobile
communications industry is heading towards tiered pricing plans, it is natural to expect mobile
devices to use WiFi more often to reduce consumption over the macro network technologies
such as Evolution Data Optimized (EV-DO), High Speed Packet Access (HSPA), Worldwide
nteroperability for Microwave Access (Wimax) and even Long Term Evolution (LTE). Current
estimates show already 40% of content delivered to smartphones is provided over WiFi
access networks [1]. For laptops, netbooks, PDAs, this number is expected to be higher.
WirelessE2E LLC September 2010 Page 1 oI 23
www.wirelesse2e.com
whI Popr
n this paper, we will examine the history and progress of WiFi integration
with mobile service provider networks, review a number of recent
interworking proposals to achieve traffic offload and provide an analysis
on the likely technology outcomes in this area.
Before DuaI Mode Handsets
Since mid 1990s, mobile industry has been looking at using WiFi as a
complementary access technology. This has gained momentum with the
ratification of 802.11a and 802.11b in September 1999. Especially
802.11b with its 2.4 GHz spectrum, allowing three non-overlapped
channels and ability to use 11 Mb/s as its peak radio link rate, it looked
very attractive to mobile service providers. While public WiFi service
providers were advertising peak rates of 1.5 or 2.0 Mb/s, mainly limited
by the backhaul link (T1 or E1) in early 2000s, macro networks were still
using Circuit Switched Data (CSD) at 9.6 kbps or Cellular Digital Packet
Data (CDPD).
During the early parts of the decade, 2001-2002, mobile communications
industry focused on bringing up 2.5G networks, with the exception of
NTT DoCoMo and Softbank in Japan that turned up their 3G networks
before the end of 2002. During this period, industry realized the potential
benefits of using WiFi as an access technology. Especially with the
launch of ntel's Centrino in February 2003 and the ratification of 802.11g
in June 2003, WiFi has gained even bigger momentum.
Since handheld terminals were lacking WiFi capability, primary focus of
the industry was allowing WiFi to be used as an alternative access
technology for laptops and possibly more esoteric devices like
Compaq/HP iPaqs. Users were expected to switch from one network to
WirelessE2E LLC September 2010 Page 2 oI 23
www.wirelesse2e.com
Unril 802.11b
rori[icorion roJio
lin| speeJ wos
limireJ ro 1 onJ 2
Mb/s
Inrel oJJeJ 802.11q
supporr ro Cenrrino
in 2004
whI Popr
another, typically manually for Packet Switched (PS) services only.
Therefore, the level of integration was primarily limited to the use of
common AAA platforms and interconnectivity of a WiFi access network to
a 3
rd
Generation Partnership Project (3GPP) Packet Core network. 3GPP
Technical Standard (TS) 23.234 which was originally published as part of
release 6 standards in 2004 covered these two aspects [2]. TS 23.234
covered multiple scenarios describing the connectivity between the WiFi
access network and 3GPP packet core network(s) including roaming,
sharing of an access network among multiple nternet Service Providers
(SPs). WiFi-Packet Core interworking was a limited success due to a
number of reasons:
Subscriber terminals (laptops, PCMCA cards) lacked the
capability of Subscriber dentity Module (SM) integration. Few
attempts were made by companies such as Nokia that had an
802.11b card with an integrated SM in the market for a short time
early in the decade. However, user adoption of such devices didn't
materialize since WiFi networks needed significant investment to
support SM-based authentication.
Billing and provisioning systems of many service providers were
designed around subscriber identity based on nternational Mobile
Subscriber dentity (MS). This tied such systems to rely on the
provisioning of SMs and use of Mobile Subscriber ntegrated
Services Digital Network Number (MSSDN).
nterconnecting WiFi access networks back to mobile service
provider Packet Core networks wasn't attractive primarily due to
limited benefits and high cost. Original 2.5G networks were
WirelessE2E LLC September 2010 Page 3 oI 23
www.wirelesse2e.com
Firsr No|io 802.11b
corJ wirn inreqroreJ
SIM wos ovoiloble
os eorlv os 2000.
whI Popr
universally deployed around the principle of interconnecting the
roamed access network back to home packet core where the
General Packet Radio Services (GPRS) Gateway Serving Node
(GGSN) or Packet Data Gateway (PDG) exist. This model was
adopted by operators early on since they had a large content
business that required customer to be connected to home
operator's wall-garden. Since all customer traffic was traversing
through the home network, operators were able to identify flows
belonging to particular customers to simplify (hide) authentication
and authorization details from the mobile users. Considering 2.5G
usage was miniscule (compared to today's 3G and 4G device
usage) and data roaming prices were quite steep, this approach
worked well for the mobile operators. On the other hand, WiFi
access network users were interested in a quick ramp to Internet.
They were not willing to pay a high cost and did not have much
interest in home operator services in its walled-garden.
During the early parts of the decade, many mobile operators that went
into the WiFi access network business ended up building a separate
network and along with a different business unit. They used WiFi as a
marketing tool that increased their brand coverage as well as a way to
prevent any competitive threats to their business for nternet connectivity.
Some operators experimented and continued to offer technology
solutions (e.g., BirdStep Technology, NetMotionWireless, ColumbiTech,
SmithMicro mobile Virtual Private Network (VPN) clients) that overlay
over their macro 2.5G/3G networks and their WiFi access networks. But
these mobile VPN solutions ended up mainly restricted to enterprise
users. Many operators solved AAA integration by mash-ups involving
their WiFi login pages with their Short Messaging Service (SMS).
WirelessE2E LLC September 2010 Page 4 oI 23
www.wirelesse2e.com
Durinq eorlv 2000s,
primorv
comperirion [or
mobile VPN
recnnoloqv were
Secure Soc|er Lover
(SSL) onJ IP
Securirv (IPSec).
whI Popr
Before Smartphones
First dual-mode terminal (with Global System for Mobile communications
(GSM) or Code Division Multiple Access (CDMA) and WiFi capability)
didn't come to market until 2005. n 2006, multiple manufacturers
including Samsung, Nokia, Blackberry introduced their dual-mode
devices. Earlier in the decade, start-up firm Kineto proposed an
alternative way of integrating WiFi and GSM. Their primary idea was to
use WiFi as a transport link between a dual-mode handset and the
operator network carrying GSM/GPRS signaling as well as voice and
data bearers. Their proposal has led to the standardization of Unlicensed
Mobile Access (UMA) in September 2004 which was later incorporated
into 3GPP TS 43.318 release 6 [3]. Following figure from UMA Forum
web page (http://www.smart-wi-fi.com) describes the interaction between
a UMA User Equipment (UE) and the network.

UMA's biggest benefit was to solve coverage problems experienced by
WirelessE2E LLC September 2010 Page 5 oI 23
www.wirelesse2e.com
Firsr exomple
Jevice in WiFi-
Allionce opprovol
lisr wos Mororolo
MPx rnor receiveJ
irs opprovol in
Auqusr 2004 onJ
FCC-rvpe opprovol
[or irs GSM roJio in
December o[ rne
some veor. Ir wos
never mor|ereJ.
whI Popr
many users, especially at home. Operators such as T-Mobile USA,
Orange, BT have become early adopters of the technology. Since
mobility complexity (user interaction) was hidden from the user, UMA has
become a mass-market product as opposed to earlier mobile VPN
integration technologies.
UMA's progress was hampered by two major factors: cost/complexity it
added to the dual-mode handset and a difficult marketing environment.
Handset cost/complexity was the natural result of building a consumer-
grade technology. Since mobility between macro network and WiFi was
the foundational tenet of the technology, handset development cost was
increased due to additional testing and integration effort. Marketing of
UMA by mobile carriers using this technology was a mixed-bag. Majority
of them emphasized the coverage benefit of UMA but they didn't put
significant marketing budgets behind the promise. nstead additional
calling packages were offered. Furthermore many operators who already
have strong in-door, at-home coverage continued negative marketing,
calling UMA an interim technology. Such operators positioned femtocell
as the long-term solution to in-door coverage expectations of their
customers.
After iPhone and Android
ntroduction of iPhone in 2007 and Android in 2008 changed the
smartphone usage dramatically. Both platforms have adopted WiFi as an
essential wireless access technology. Apple made iTouch platform with
WiFi radio only in addition to iPhone. Furthermore, Apple took a strong
stance on deciding which applications should be used on macro GSM/3G
networks and which applications should be limited to WiFi only. Apple
WirelessE2E LLC September 2010 Page 6 oI 23
www.wirelesse2e.com
As o[ Seprember
2010, UMA Forum
websire poinrs ro
onlv 6 (ocrivelv-
solJ) UMA [eorure
pnones. Ir incluJes
7 Bloc|berrv
moJels, No|io L7S
onJ o re[erence ro
oll AnJroiJ
smorrpnones.
whI Popr
with partnership with major carriers such as AT&T has developed traffic
offloading mechanisms. This has motivated AT&T to expand their WiFi
footprint substantially after the iPhone launch. Similarly, Android platform
integrated WiFi radio well into the software development kit from the
beginning. Many applications in the Android marketplace were written to
exploit the throughput, latency and zero-cost (for local content)
advantages of WiFi.
Most fundamental changes brought by iPhone, Android and Blackberry
include the consistency of having WiFi in all devices, rich operating
system support for applications that would use WiFi and ability to
automatically authenticate into public WiFi networks using Wireless
nternet Service Provider roaming (WSPr) capabilities. n addition rapidly
increasing wired broadband availability at home and office along with the
introduction of 802.11n in iPhone and Android devices in 2010 made
WiFi a more viable radio access network choice.
As of second half of 2010, neither Android nor OS (iPhone) supports
simultaneous use of WiFi and macro wireless network technologies for P
traffic. Only exception to this is the WiFi tethering capability that was
recently introduced to both operating systems. Same restriction also
exists in 3GPP standards that are based on the use of one radio access
technology at a time for data traffic. 3GPP release 10 standards are
being developed to remove this restriction to allow dynamic traffic
offloading capability. We will review different offloading techniques later
on in this paper.
WirelessE2E LLC September 2010 Page 7 oI 23
www.wirelesse2e.com
WISPr is JevelopeJ
onJ sronJorJizeJ bv
Wireless BrooJbonJ
Allionce.
whI Popr
Data Cards, 3G/WiFi Routers and Connection
Managers
Another major device group with dual mode radio is the collection of data
cards and 3or4G/WiFi routers. Equipment manufacturers have started
delivering dual mode wireless cards (GPRS or S-95 single carrier radio
transmission technology (1xRTT) with WiFi) in the early part of the
decade. Later on with the proliferation of 3G networks, as well as
introduction of Wimax more multi-mode devices came to market.
ntroduction of HSPA, EV-DO and Wimax also enabled a new device
category (sometimes referred as personal hotspot) that combines one or
more macro mobility backhaul link technologies along with WiFi Access
Point (AP) capability. Different manufacturers including Cradlepoint,
Sierra Wireless, Novatel, Huawei are some of the many manufacturers
that developed products to serve this market.
Common component of dual-mode data cards and personal hotspot
devices is a connection manager that runs on the computing platform
where the data card is attached or the embedded device providing the
personal hotspot. Connection manager is a software that hides network
selection, attachment, authentication, authorization, disconnection details
from the user. Furthermore connection managers have advice of billing,
optimization capabilities. Connection manager technology has evolved
from mobile VPN software we have discussed earlier.
Data OffIoading
n order to understand the need for data off-loading, best place to start
with is the LTE technology since currently it is the most advanced and
widely embraced macro wireless access network technology. Based on
WirelessE2E LLC September 2010 Page 8 oI 23
www.wirelesse2e.com
MiFi is o bronJ
nome ossocioreJ
wirn personol
norspor. In rne US,
ir is o rroJemor|
owneJ bv Novorel
wnereos SUK owns
rne some nome in
rne UK onJ mor|ers
o Jevice [rom
Huowei unJer rne
some nome.
whI Popr
the current predictions, the next decade will be the period when LTE is
deployed across the world to become the unifying radio access
technology. Considering the current diversity in the market, this is a
testament to the perceived advantages of LTE. Many of those promises
are well publicized:
common standard across the globe
high spectral efficiency (16.32 bits/Hz per mobile)
potential to provide up to 326 Mbit/s per user
all P to overcome circuit switched legacies
ability to use re-farmed spectrum
All of these advantages point to a much lower operating cost for
providing data services. Compared to the current cost levels (estimated
to be around $0.01/MByte) for HSPA+, LTE is expected to reach another
order of magnitude in cost reduction ($0.001/MByte) [4].
Even with this significant cost reduction, expected surge in data usage
looks overwhelming. The following figure is from Cisco's public research
published under Visual Network ndex (VN) summarizing the trends in
nternet traffic as well as predictions.
WirelessE2E LLC September 2010 Page 9 oI 23
www.wirelesse2e.com
Cisco nos
populorizeJ rne
niqner orJers o[
unir o[ in[ormorion
rnrouqn rneir VNI
sruJies. Oriqinollv
Zerrobvre (10
21
bvres) wos rne
orJer o[ moqniruJe
[or rne new ero o[
compurinq. More
recenrlv Yorrobvre
(10
24
bvres) come ro
circulorion.
whI Popr
This graph and a companion (showing the distribution of total traffic
among geographical regions) show that by 2014, total monthly traffic in
the US will reach 750 PB (750,000 TB).
Based on the most recent CTA figures, there were about 250,000 base
stations in the US as of the end of 2009. Let us assume a brand new LTE
base station is placed next to an existing other technology base station
and LTE carries only 50% of the total traffic (due to lagging deployment
of devices) by 2014. Based on these assumptions, following table shows
the severity of the capacity situation.
WirelessE2E LLC September 2010 Page 10 oI 23
www.wirelesse2e.com
whI Popr
Like any large scale network, wireless service provider networks have
fairly long tails in their traffic distribution graphs. Based on the typical
Pareto distribution (commonly known as the 80/20 rule), approximately
20% of base stations will be carrying 80% of all traffic. f the same
analysis is applied repetitively (assuming self-similarity of traffic
distribution patterns), it is possible to demonstrate that top 2.5% of all
base stations will carry close to 1/3 of all traffic.
Similar to the long-tail geographical distribution of traffic, a significant
temporal lumpiness also exists. n other words, traffic is not uniform,
instead it increases/decreases with people's sleeping, working, driving,
shopping, etc. habits. That's why instead of looking at the average
throughput, the Busiest Hour in a given month must be considered as the
engineering design target. Based on a conservative figure of 0.05% of
monthly traffic within the Busiest Hour, a throughput figure of 20 Mbps
can be computed for the unlikely scenario of all base stations being
equally loaded. f the analysis is limited to the load of top 20% of base
stations (50,000 base stations), then per base station throughput
requirement of 80 Mbps is derived. Certainly the top 6% or 2.5% of base
stations need to serve even higher levels of traffic.
By 2014, LTE with a theoretical sector capacity of 48 Mbps (in 44 MMO
downlink with 20+20 MHz deployment scenario) needs multiple base
stations and more importantly more spectrum in a significant number of
WirelessE2E LLC September 2010 Page 11 oI 23
www.wirelesse2e.com
whI Popr
base station locations. Same problem was recognized by the recent
National Broadband Plan prepared by Federal Communications
Commission [5].
Another method to overcome this capacity crunch is to offload a large
portion of this traffic to complementary wireless access networks. So far
industry has come up with two possible solutions:
Using LTE femtocells
Using WiFi
Femtocell is a viable approach as long as service providers subsidize it
and they are ready to handle the operational challenges of deploying and
supporting them. Unlicensed nature of spectrum, being a well-
established home networking technology with significant penetration,
commodity hardware and more importantly the presence of a WiFi radio
in all new smartphones, tablets and other consumer electronic devices
planned for LTE make WiFi traffic offloading as an equally viable
alternative to resolve the capacity crunch LTE networks will experience in
2014 and beyond.
n 3GPP there are three parallel efforts to address traffic offloading:
Local P Access (LPA): LIPA provides access to a
residential/corporate local network interconnected to a femtocell
(Home (e)NodeB Subsystem). Primary motivation is to provide
access to a subnet within home or an office for shared resources
like printers, file servers, media servers, displays, etc. while a
mobile device is attached to the 3GPP operator network. LIPA
capability will be introduced with 3GPP Release 10 specifications.
WirelessE2E LLC September 2010 Page 12 oI 23
www.wirelesse2e.com
InJusrrv snippeJ
200,000 [emrocells
in 2009. Yeorlv
volume is expecreJ
ro reocn 12M bv
2014 (nrrp:
berqinsiqnr.com)
whI Popr
Selected P Traffic Offload (SPTO): SPTO refers to the ability to
selectively forward different types of traffic via alternative routes
to/from the terminal device. Home network flavor of SIPTO is
dependent on using a Home (e)NodeB. In this mode, specific
traffic determined by operator policy and/or subscription
transferred to/from (e)NodeB directly to Internet/Intranet
bypassing the mobile operator access/core network. Macro
network flavor of SIPTO covers the ability to offload traffic next to
an Radio Network Controller (RNC) for 3G or Serving Gateway
(S-GW) for LTE as opposed to traversing the operator's core
network.
P Flow Mobility (FOM): FOM specifies the terminal device to
connect to two access networks (WiFi and 3G/4G) simultaneously
and forward/receive packets belonging to different flows through
different access networks. Unlike the other methods FOM
doesn't assume the presence of Home (e)NodeB.
LocaI IP Access (LIPA)
Fundamental premise of LPA is the need for the terminal device to have
access to resources connected to a local network that can be residential
or an office environment. mproved display and computational
capabilities of user terminals make them more suitable to actively
participate in home or office networking. Today, it is possible to listen to
music, view videos, edit documents on user terminals that are typically
connected to a mobile operator macro network. With current dual-mode
devices users typically turn off their macro access network radio link, turn
on their WiFi radio and connect to their local P network. LPA is currently
proposed as a potential method for making this local P network access
WirelessE2E LLC September 2010 Page 13 oI 23
www.wirelesse2e.com
whI Popr
seamless for the user.
Even though we are discussing WiFi offload options in this paper, LPA
itself is not dependent on the use of WiFi in the user terminal. nstead, it
is based on the presence of a home (e)NodeB and a co-located Local-
Gateway (L-GW) function. t is being defined in 3GPP standard 23.829
and is expected to be ratified as part of Release 10 [6].
Following figure taken from 3GPP 23.829 describes the network
elements involved to implement the LPA functionality.

Backhaul
Residential /
Enterprise
network
H(e)NB
H(e)NB
-GW
Mobile Operator
Core Network
UE
LIPA L-GW

User Equipment (UE) interconnects to (e)NodeB via the standard macro
network radio interface such as UMTS or LTE. L-GW includes a full-
fledged LTE Packet Data Network (PDN)-Gateway (P-GW) or a GPRS
GGSN functionality. Therefore it must implement packet filtering,
shaping, policing, address assignment for UE and direct tunneling
capability with the RAN which happens to be included in the (e)NodeB.
Home (e)NodeB Gateway (H(e)NB-GW) is the standard Security
Gateway (SeGW) for LTE. n case of UMTS a separate Home NodeB
Gateway (HNB-GW) is also needed as in the case of typical femtocell
deployments.
WirelessE2E LLC September 2010 Page 14 oI 23
www.wirelesse2e.com
whI Popr
A discovery mechanism for LPA capability must be built into the
architecture. Most likely option for implementation is to include the
availability of LPA capability in the RAN broadcast by (e)NodeB at home
or enterprise. Using this information, UE will request network connectivity
from MME or SGSN indicating the use of LPA by a dedicated Access
Point Name (APN) or possibly a flag.
LPA introduces a number of changes to existing network elements such
as MME, SGSN, Home Subscriber Server (HSS), RAN. Most
fundamental change is the placement of a core operator network node
(P-GW or GGSN) in the home or enterprise. That brings in additional
implementation complexity to (e)nodeB. Furthermore, it requires
signaling changes in existing protocols to enable the association
between core network nodes like MME or SGSN and UE and (e)NodeB
in the customer premises. MME/SGSN must be involved in L-GW
selection and scaling this for large number of customer premises devices
as opposed to a few S-GW/P-GW/GGSN can be a significant burden.
SeIected IP Traffic OffIoad (SIPTO)
SPTO extends what LPA aims to achieve by providing an alternate path
for selected traffic. Primary objective is to add capacity to the primary
RAN-core network path and reduce unit cost of transport. SIPTO is
covered in 3GPP 23.829 along with LIPA and will be standardized as part
of Release 10.
SPTO has been developed with two different flavors:
SPTO for Home (e)NodeB
SPTO for the macro network
WirelessE2E LLC September 2010 Page 15 oI 23
www.wirelesse2e.com
whI Popr
SIPTO for home (e)NodeB is similar to LIPA in terms of necessary
changes to network capabilities and additional functionality in the form of
a L-GW collocated with Home (e)NodeB. Considering substantial
complexity and limited benefit of SIPTO for home (e)NodeB as opposed
to LIPA, 3GPP decided to mark that as a future study item beyond
release 10.
SPTO for the macro network is a reflection of the current marketplace
where a number of packet core equipment vendors including Cisco,
Juniper, Tellabs already provide similar capability. Currently these
vendors perform the traffic break-out by intercepting the Gn interface
traffic between SGSN and GGSN and redirecting RNCs to use a direct
tunnel towards a new traffic break-out node typically located closer to
RAN. 3GPP will formalize this capability with Release 10 standards.
Following figure from TS 23.829 describes how the SPTO is achieved in
a UMTS/GPRS macro network.

CN
Gn / Gp
SGSN
RAN
L - G GSN
MS
RNC GGSN
CN Traffic
Gn Iu
Gn
SPTO Traffic
n this deployment model, operators may choose to implement a brand
new APN that is mapped to a closer L-GGSN (equivalent to L-GW of
LPA). n such a deployment model, no other changes are needed to
implement the SPTO functionality. A more generic solution that would
allow the use of a single/common APN needs adding a SPTO-enabled
flag per APN in user subscription, transferring this flag from HSS / Home
WirelessE2E LLC September 2010 Page 16 oI 23
www.wirelesse2e.com
whI Popr
Location Register (HLR) to MME/SGSN and possibly Dynamic Name
Server (DNS) based mechanism for gateway (GGSN or P-GW) selection.
SPTO for the macro network doesn't solve the fundamental capacity
crunch we highlighted earlier. Since it is a solution that allows traffic
break-out above RAN, it doesn't reduce any potential congestion on the
radio interface or the cell site backhaul.
IP FIow MobiIity (IFOM)
Unlike LPA or SPTO that are dependent on upstream network nodes to
provide the optimization of routing different types of traffic, FOM relies
on the handset to achieve this functionality. t explicitly calls for the use of
simultaneous connections to both macro network, e.g., LTE, UMTS and
WiFi. Therefore, FOM, unlike LPA and SPTO, is truly a release 10-
onward only technology and it is not applicable for user terminals pre-
Release 10. FOM is being specified via 3GPP TS 23.261 [7]. Following
diagram shows the interconnectivity model for FOM capable UE.
FOM uses an nternet Engineering Task Force (ETF) Request For
Comments (RFC), Dual Stack Mobile Pv6 (DSMPv6) (RFC-5555) [8].
WirelessE2E LLC September 2010 Page 17 oI 23
www.wirelesse2e.com
whI Popr
Since FOM is based on DSMPv6, it is independent of the macro
network flavor. t can be used for a green-field LTE deployment as well
as a legacy GPRS packet core.
Earlier on we looked at the mobile network industry attempts of
integration between packet core and WLAN networks. Common
characteristic of those efforts was the limitation of the UE, its ability to
use one radio interface at a time. Therefore, in earlier interworking
scenarios UE was forced to use/select one radio network and make a
selection to move to an alternative radio for all its traffic. Today many
smartphones, data cards with connection managers already have this
capability, i.e., when the UE detects the presence of an alternative
access network such as a home WiFi AP, it terminates the radio bearers
on the macro network and initiates a WiFi connection. Since WiFi access
network and packet core integration is not commonly implemented, user
typically loses her active data session and re-establishes another one.
Similarly access to some operator provided services may not be
achieved over WiFi. Considering this limitation both iPhone OS and
Android enabled smartphones to have simultaneous radio access but
limited this functionality to sending MMS over the macro network while
being connected to WiFi only.
FOM provides simultaneous attachment to two alternate access
networks. This allows fine granularity of P Flow mobility between access
networks. Using FOM, it will be possible to select particular flows per UE
and bind them to one of two different tunnels between the UE and the
DSMPv6 Home Agent (HA) that can be implemented within a P-GW or
GGSN. DSMPv6 requires a dual-stack (Pv4 or Pv6) capable UE. t is
independent of the access network that can be Pv4 or Pv6.
WirelessE2E LLC September 2010 Page 18 oI 23
www.wirelesse2e.com
whI Popr
AnaIysis
Three alternative architectures that are being standardized by 3GPP
have various advantages and shortcomings. They are quite distinct in
terms of their dependencies and feasibility. Following table is a summary
of comparison among these three approaches for traffic offloading.
Characteristics LIPA SIP1O macro-
network
IFOM
UE backward
compatibility
It requires minor
changes in the UE
signaling per
release-10
speciIications. Does
not depend on WiFi
capability in the UE.
It allows the use oI
legacy UE. Does not
depend on WiFi
capability in the
handset.
It requires a release-
10 UE along with
simultaneous WiFi
and macro network
capabilities as well
as DSMIPv6.
Ability to solve
the Iundamental
problem oI
capacity crunch
It depends on
Iemtocell and
provides same
beneIits Ior
oIIloading traIIic
Irom macro
network. Extending
LIPA Ior
Internet/Intranet
access will only
beneIit core network
capacity.
Since oIIloading
takes place above
RAN. it does not
have any impact on
congestion in the
RAN.
It solves the
capacity problem
eIIectively by
allowing selected
Ilows over the
alternative WiFi
access network.
Mobility support Mobility among
Iemtocells would
require sub-optimal
routing oI traIIic
back to operator
core.
No impact on
mobility support.
Full mobility
support at the Ilow
level between WiFi
and macro network.
Direct access to
home/enterprise
It allows direct
access bypassing
dependency on
operator core
network.
No direct access is
possible. UE can
communicate with
local network via
Internet-Operator
core network path.
No direct access is
possible. UE can
communicate with
local network via
Internet-Operator
core network path.
WirelessE2E LLC September 2010 Page 19 oI 23
www.wirelesse2e.com
whI Popr
Characteristics LIPA SIP1O macro-
network
IFOM
Standardization
eIIort
It requires changes
in HSS/HLR.
SGSN. MME.
(e)NodeB.
It requires changes
in HSS/HLR.
SGSN. MME.
It requires changes
in policy control.
updates in P-GW.
GGSN to implement
HA capability.
Regulatory
impact
LawIul Intercept Ior
local access is
diIIicult but LI-GW
under operator
control. it is
solvable.
No impact to
LawIul Intercept
capability.
II all traIIic goes
over HA (integrated
with P-GW or
GGSN). no impact.
Dependency on
radio access
network
It is dependent on
having home
(e)NodeB. and
modiIications to
support LIPA.
No dependency or
changes needed in
the radio access
network.
No dependency or
changes needed in
the radio access
network.
Overall
Feasibility
Its Iate Ior success
is dependent on the
success Ior
Iemtocell. It is
primarily an
enhancement to
Iemtocell
architecture.
It is straight-Iorward
and highly Ieasible
to implement even
though it doesn't
solve the
Iundamental
capacity crunch in
the radio access
network.
Implementing
DMIPv6 on UE and
HA is straight-
Iorward. Primary
diIIiculty is to
overcome concerns
oI mobile operators.
Looking at the relative strengths of the existing traffic offload proposals, it
is difficult to pick an outright winner. SPTO macro-network option is the
most straight-forward and most likely to be implemented rather quickly.
However, it doesn't solve the fundamental capacity crunch in the radio
access network. Therefore its value is limited to being an optimization of
the packet core/transport network. Some other tangible benefits would be
reduction in latency to increase effective throughput for customers as
well as easier capacity planning since transport facilities don't need to be
dimensioned for large number of radio access network elements
anymore.
WirelessE2E LLC September 2010 Page 20 oI 23
www.wirelesse2e.com
whI Popr
LPA provides a limited benefit of allowing access to local premises
networks without having to traverse through the mobile operator core.
Considering it is dependent on the implementation of femtocell, this
benefit looks rather small and has no impact on the macro radio network
capacity. f LPA is extended to access to nternet and ntranet, then the
additional offload benefit would be on the mobile operator core network
similar to the SPTO macro-network proposal. Femtocell solves the
macro radio network capacity crunch. However, the pace of femtocell
deployments so far doesn't show a significant momentum. LPA's market
success will be limited until cost of femtocell ownership issues are
resolved and mobile operators decide why (coverage or capacity) to
deploy femtocells.
FOM is based upon a newer generation of Mobile P that has been
around as a mobile VPN technology for more than 10 years.
Unfortunately success record of mobile P so far has been limited to
enterprise applications. t hasn't become a true consumer-grade
technology. ntroduction of LTE may change this since many operators
spearheading LTE deployments are planning to use Pv6 in handsets
and adopt a dual-stack approach of having both Pv4 and Pv6 capability.
Since many WiFi access networks will stay as Pv4, DSMPv6 will be the
best tunneling mechanism to hide Pv6 from the access network. Having
dual-stack capability will allow native access to both legacy Pv4 content
and native Pv6 content from major companies such as Google,
Facebook, Yahoo, etc. without the hindrance of Network Address
Translation (NAT). Considering the popularity of smartphones such as
iPhone, Blackberry and various Android phones, they will be the proving
ground for the feasibility of DSMPv6.
WirelessE2E LLC September 2010 Page 21 oI 23
www.wirelesse2e.com
whI Popr
ConcIusions
Enhancing macro radio access network capacity by offloading mobile
video traffic will be essential for mobile communications industry to
reduce its units costs to match its customer expectations. Two primary
paths to achieve this are the use of femtocells and WiFi offloading.
Deployment of large scale femtocells for coverage enhancement has
been a limited success so far. Using them for capacity enhancements is
a new proposition for mobile operators. They need to assess the
necessity of using them as well as decide how to deploy them selectively
for their heavy users.
Use of DMSPv6 as described in 3GPP work item FOM has a lot of
promise. t solves multiple problems including flow level mobility between
WiFi and macro radio access network. ts implementation has no
dependency on the radio access network. n other words, it is truly an
end-to-end solution that can be applied for multiple radio access
technologies. Next in line is to make user equipment capable of using
multiple radio links simultaneously. Both iPhone and Android already
have this capability but currently they are restricted in the type of access
(e.g., MMS over macro network while connected to WiFi) they can be
used for. Obviously this requires further enhancements to user
equipment for management of battery life, self-interference as well as
network/flow mapping criteria based on cost, quality, security, user
preference, etc. Putting all of this in a consumer-grade product is a
significant challenge. We believe this will continue to be a hard problem
to solve for years to come. When it is finally resolved it will have a
profound impact on the success of mobile communications industry by
reducing the unit cost of traffic delivery substantially.
WirelessE2E LLC September 2010 Page 22 oI 23
www.wirelesse2e.com
whI Popr
References
|1| Admob Mobile Metrics Report.
http://www.admob.com/marketing/pdI/mobilemetricsnov08.pdI
|2| 3GPP TS 23.234. 3GPP system to Wireless Local Area Network (WLAN)
Interworking. release 6.0.
|3| 3GPP TS 43.318. Generic Access Network (GAN) Stage 2. release 6.0.
|4| WirelessE2E estimate based on the industry usage. revenue Iigures.
|5| Connecting America: The National Broadband Plan. chapter 5. FCC. March
2010.
|6| 3GPP TS 23.829. Local IP Access and Selected IP TraIIic OIIload. v1.2.0.
August 2010.
|7| 3GPP TS 23.261. IP Flow Mobility and seamless WLAN oIIload. v2.0.0.
June 2010.
|8| RFC-5555. Mobile IPv6 support Ior dual stack Hosts and Routers
(DSMIPv6). June 2009.
WirelessE2E LLC September 2010 Page 23 oI 23
www.wirelesse2e.com

You might also like