Você está na página 1de 12

Nokia Siemens Networks

LTE-capable transport: A quality

user experience demands an
end-to-end approach

White Paper
1. E
 xecutive summary: The need for end-to-end transport
network planning

The rising tide of data traffic is putting Latency also has a considerable the QoS in the transport network,
transport networks under more pressure impact on user satisfaction, especially for instance.
than ever and the arrival of LTE will only in delay-sensitive applications such as
accelerate that process. This paper online gaming. If the reaction time of Synchronization is essential in all
focuses on what is needed for an LTE- the network is too long, a high-speed telecommunications networks. A
capable transport network to deliver an connection wont do much to improve number of different strategies can be
optimized end-user experience. the experience. Network providers adopted in transport networks and
can control some aspects of latency, they all have their advantages and
The first requirement is for cost- but it also depends on external factors, limitations. Hybrid systems, where a
effective capacity. Should transport such as the distance between the mix of synchronization technologies is
networks be dimensioned to meet user and the content. An approach used, are likely to be commonplace.
average demand or peak demand, for that delivers the best possible latency
example? Communications service wont go wrong, provided it is Service assurance and network security
providers (CSPs) could approach the economically viable. are the other factors that play key roles
dimensioning of their networks in a in determining the user experience in
number of ways, but the optimum Quality of Service (QoS) differentiation LTE-capable transport networks.
solution will essentially strike a balance enables CSPs to manage the
between providing the maximum performance of different streams of Ultimately, sound network planning
capacity for users and keeping the traffic. QoS can be a powerful tool for and an end-to-end approach to
transport network economically and managing the user experience, but it network operations will determine
technically feasible. must be managed end-to-end. LTE how well these emerging transport
radio QoS has to be aligned with networks perform.

2. T
 he impact of LTE on transport networks

LTE promises a whole new mobile There is a general consensus in the What is the right transformation
broadband experience for everyone, industry that only a packet-based strategy? What is the optimum
with throughput rates beyond 100 Mbit/s transport network will be able to meet target architecture and how can we
and short latency of around 20 ms the challenge. However, there are still get there?
or better. Its an experience formerly unresolved issues around transport
available only from fixed connections. and they tend to revolve around In this white paper we will focus mainly
One thing is clear, however. All the three topics: on the first question and discuss the
progress made in the radio and core How should we provide the user requirements for an LTE-capable
subsystems wont count for much experience? What throughput and transport network. Well look at how to
unless the underlying transport latency values are required and how provide the best or more precisely
architecture is ready to deliver the key can we achieve them? optimized - user experience. This
performance indicators (KPIs) required How can we do it all cost-efficiently naturally includes a look at how CSPs
to support such a lofty value proposition. and bring the price per bit down? design and implement the most reliable
transport networks.

2 LTE-capable transport: A quality user experience demands an end-to-end approach

3. C
 apacity and dimensioning

HSPA, HSPA+, LTE and LTE-A each to plan radio and transport network consider the realistic peak bitrates,
promise to deliver progressively capacity development over time. which can normally be achieved within
higher data rates, so how should the the cell. The above figure shows
underlying transport network be average cell throughput rates based
dimensioned? What is the capacity 3.2BTS capacity based on simulations, which were carried out
requirement for each base station on air interface bit rate by 3GPP considering a certain user
(BTS)? For the latter there are two distribution in the cell, terminal mobility,
basic approaches: The bottom-up interference etc.
approach is based on actual traffic On the other hand, many CSPs do not
model predictions, while the top-down have previous experience with the When calculating the total transport
alternative is based on the bitrates uptake of data services. Flat rates capacity needed per BTS, full peak
possible with different air interface and large data bundles typically make bitrate dimensioning might result in
technologies. predictions difficult. In short, if the sort values that are too high. Dimensioning
of educated guess used previously is based only on the average might
not practical, the other option is to do result in values that are too low and
3.1BTS transport capacity:
a top-down calculation based on the cause regular congestion. A good
bottom-up air-interface bitrates of different compromise might therefore be to use
radio technologies to achieve an a so-called single-peak, all-average
Network dimensioning has traditionally
estimate of the user plane traffic. The model, as shown in the next figure.
used the bottom-up approach. A traffic
following figure shows the theoretical
model is calculated for a time period
maximum bitrates available for certain In this model the user traffic requirement
based on certain assumptions. The
configurations. Note that those peak of the BTS is presumed to be either
model then produces estimates that
values are only for a single sector, so the aggregated average capacity of all
can be used to dimension the transport
a three-sector site would have to serve cells or the peak capacity of one cell.
three times these peak rates. Planners use whichever value is higher,
so that the advertised user service peak
The obvious advantage of this approach
Dimensioning a network based on rates can be momentarily supported in
is the scientific basis for the estimates,
peak rates is looking very much at the any given cell, although the advertised
which are based on experience. It
worst case and will result in over- user service rate will be only a fraction
also is independent of the actual radio
dimensioning. Its therefore useful to of the cell peak rate.
technology used and could be used

Max. peak data rate Average throughput (macro cell)
250 Uplink Downlink
40 Uplink
Mbps Mbps 30
0 LTE (2x2 MIMO), LTE (4x4 MIMO),
LTE LTE 20 MHz carrier 20 MHz carrier
20 MHz (2x2 MIMO) 20 MHz (4x4 MIMO)

Figure 1: Maximum peak vs. average data rates

LTE-capable transport: A quality user experience demands an end-to-end approach 3

an impact and the gain ultimately
levels out.
3.4LTE capacity
requirements and the
X2 Interface
One other peculiarity of LTE networks
Peak as compared to traditional 3G networks
Rate! is the X2 interface, which plays an
important role in the handover of
connections between neighboring
Cell BTSs. During the handover procedure
peak the radio link to the terminal is
eNB interrupted for a short time, typically
Cell average transport between 60 and 70ms. Downlink
All- All-Average/ All-Peak packets arriving at the BTS formerly
Average Single-Peak hosting the terminal will be forwarded to
the new BTS, connecting the terminal
Figure 2: Calculating the single-peak, all-average data rate via the X2 interface until the EPC has
switched the S1 path to the new BTS.
The final step needed to obtain the and overbooking become even more
actual bitrate required for the BTSs essential to ensure transport efficiency. In that sense, the X2 interface creates
S1 interface includes some overhead In fact, aggregation should be carried another set of traffic flows directly
calculations. The air interface overhead out close to the BTSs (for example, in between neighboring BTSs. This traffic
is stripped out and the transport MWR hub sites) to really leverage this is extremely bursty, since it occurs
and possible IPSec overheads are advantage. The example Figure 3 mostly during brief handover phases.
added in. Of course, signaling and shows how the multiplexing gain Studies show that it will normally be
management traffic should also to be depends on the number of aggregated in the range of only 3% of the total
taken into account. BTSs, based on certain assumptions. S1 traffic.

It is obvious that relatively large It is therefore a relatively minor factor

3.3 Traffic aggregation
overbooking advantages can be in planning, and the capacity certainly
Moving beyond the first link (or last gained by adopting hub-aggregation, does not require the installation of
mile) connecting the actual BTS whereas continued higher regions of dedicated physical transmission links
Pag04into the network, aggregation the network does not have as large between neighboring BTSs.

Multiplexing Gain on S1 front mile


MUX gain [%]




1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
# of aggregated eNBs

Figure 3: Multiplexing gain dependent on the number of aggregated BTSs

4 LTE-capable transport: A quality user experience demands an end-to-end approach

3.5Striking a balance dimensioning an LTE transport a compromise between providing the
between capacity network. However, in many cases maximum capacity for users and
a top-down plan based on the air keeping the transport economically and
and economy
interface peak and average rates will technically feasible. At the same time
A proper bottom-up planning process be a more feasible way forward. In aggregation in the network is essential
for transport network capacities would this context we suggest the use of the in order to leverage multiplexing gains,
probably provide the most accurate single-peak all-average model for the preferably close to the BTS sites.
(and differentiated) results when individual cell sites. This is essentially

4. L

Latency (or delay) is another factor that controllers, gateways and so on). delay introduced by the operators
affects subscribers service experience. Then there are additional delays network, which is the round trip time
From the user perspective, latency is arising from the transport network, from between the users handset and the
essentially the time it takes for a data the connectivity between the CSPs operators internet gateway. The CSP
packet to travel from the terminal via network and the internet and the time can influence and optimize this delay.
the mobile network to the content needed to reach the actual server However, the other component is the
server on the internet and vice versa. running the requested service. On top time it takes for the data to travel from
of this there may also be a queuing this gateway to the actual content
There are several components delay within any of the various nodes if server and back, and the CSP has no
affecting the final latency experienced theres any congestion. direct influence over it.
by the subscriber. Theres the systems
inherent latency that depends on the From the CSP perspective there are Latency is considered by many to be
radio technology used (BTSs and their two elements to this latency. One is the as important as the actual capacity
supported, since it governs parameters
Pag05 such as the time it takes for a
Roundtrip time* requested internet page to display. If
the reaction time is too long, a high-
speed connection cant do much to
GSM/ improve the experience. This is just
EDGE one example where latency plays a
role. Different services have different
HSPA latency requirements.
4.1Radio technology
HSPAevo inherent latency
LTE offers hugely improved inherent
latency values compared with other
radio technologies, such as 3G or even
min max HSPA. From 60 ms in HSPA, latency
0 20 40 60 80 100 120 140 160 180 200 ms is reduced to about 20 ms in LTE (all
roundtrip times).

DSL (~20-50 ms, depending on operator) Note that these values only take
* Server near RAN
into account the radio and core
components of latency. They ignore
Figure 4: Roundtrip delays

LTE-capable transport: A quality user experience demands an end-to-end approach 5

that the fact that the physical transport 4.3Latency and hence the download time for most
can (and will) contribute significantly recommendations applications (web pages, music, video,
to the overall latency. In other words, software and so on) is depending on
the low latency promised by LTE will In contrast with some of the earlier the total round-trip time (including LTE
only be experienced by the user if the radio technologies, it is normally not radio, mobile backhaul, LTE core and
underlying transport also supports the radio network systems (such as the the internet domain).
low latency. control plane) that impose practical
limits on the delay budget in LTE. This can be improved by activating
Instead its the way users experience specific options in the protocol stack
4.2Latency in transport and
different services. The experience (such as TCP Window Scaling) or
its origin typically becomes unacceptable using multiple concurrent TCP
long before network systems run sessions. The applicability of these
Propagation delay: The speed of light
into trouble. improvement options depends on the
is finite, which leads to a round trip
particular operating systems used on
time of about 1ms per 100km. This
The acceptable latency depends on terminals and servers, as well as on
shows the impact of topology on
the service type. 3GPP has indicated the actual application.
overall latency.
certain one-way delay goals for specific
Buffering and queuing delay: services in TS 23.203 (Table 1). Finally, some industry bodies have
Packet-based transport systems use issued recommendations for the
a number of buffering and queuing It is mainly online gaming, video permissible delay in mobile backhaul.
mechanisms, each of which adds conferencing and machine-to-machine These are based on the considerations
delay. Proper link planning will (M2M) applications that drive latency already mentioned. For example, the
minimize this effect. requirements. For example, they have NGMN defined a limit of 10 ms for
proposed a 50ms delay budget for two-way delay, and 5 ms if the CSP
Transmission delay: A data packet
online gaming. However it should be requires it (NGMN-optimized backhaul
takes a certain amount of time to be
noted that the focus of the 3GPP requirements, released August 2008).
transmitted based on its length and the
document is on functionality in the core
bandwidth of the connection. For large
(including service prioritization) and not However, such recommendations have
packets this can lead to delays of
the transport network itself. Therefore to be seen in context. It will be difficult
several ms over small bandwidth
a fixed (one-way) delay of 20 ms was to stick to such low delays over large
assumed for the transport network. geographies, since a transmission
Signal processing delay: The more distance of only 1,000 km completely
signal processing takes place within The experience of many services exhausts this delay budget. In this
a signal path, the longer the delays. depends more on actual latency than case, a delay budget of 40 ms from the
Therefore the sheer number of on the available bandwidth. TCP BTS to the EPC could be seen as a
processing nodes plays a role, as is used to shift a large part of the non- good compromise that still allows
does the difference between simply real time internet traffic and uses a providers to offer the most demanding
connecting on the optical level and handshake to secure transmission. real-time gaming services.
Pag06 processing actual routing operations. This means that the achievable bitrates
4.4 X2 latency requirements
Guaranteed Bit Rate Delay Budget Loss Rate Application Example The X2 interface also has its own
latency requirements. It might seem
GBR 100 ms 10-2 Conversational Voice
at first glance as if these latency
150 ms 10-3 Conversational Voice, Live Streaming requirements would very stringent.
However, it has to be considered
50 ms 10-3 Real Time Gaming that during the handover phase the
radio link to the user terminal will be
300 ms 10-6 Buffered Streaming
interrupted for a short time anyway.
Non - GBR 100 ms 10-6 IMS signaling, Control plane Any forwarding of packets faster than
between 60 and 70 ms therefore
300 ms 10-6 Buffered Streaming, TCP applications (specific service) serves no real purpose.
100 ms 10-3 Interactive Gaming, Live Streaming
Given that an LTE transmission
300 ms 10-6 TCP applications (premium bearer) network should be designed with
stringent delay targets, the X2 interface
300 ms 10-6 Default Bearer does not significantly change things. In
particular, it does not mandate the use
Table 1: Some applications are more sensitive to latency issues than others of direct inter-BTS connectivity.

6 LTE-capable transport: A quality user experience demands an end-to-end approach

4.5Minimize latency Truly end-to-end optimization for The delay outside of the operators
to optimize the user latency has to take into account a network also deserves some attention.
number of factors, including topology If content is stored literally at the other
(distances and the number of end of the world, delay values will
Any latency requirements are driven processing nodes) and the distance be very high in any case. Content
primarily by the targeted user between the EPC and the internet buffering and similar methods are
experience. In that sense, an peering point, as well as proper link therefore set to become increasingly
approach that delivers the best planning and dimensioning. important.
possible latency wont go wrong,
provided it is economically viable.

5. Quality of Service

5.1What is Quality of A complete QoS differentiation solution reservations for parts of the traffic and
Service? spans the whole network. The core prioritization will be used in scheduling.
is responsible mainly for QoS
Quality of Service (QoS) differentiation management, such as the definition Prioritization (soft QoS)
enables CSPs to segregate the flow of and dissemination of respective Queuing systems in various elements
traffic and this allows them to monitor QoS policies, including the use of enforce prioritization. In the air interface
and manage the performance of technologies such as DPI. Both the there is a packet or frame scheduler
different streams individually. Such a radio access and core networks take that prioritizes the data. In addition,
differentiation could be made on the care of QoS control, tagging respective transport resource management
basis of applications or services (with traffic packets using VLAN priority bits algorithms or multiplexing algorithms
real-time services typically being more or DSCP (DiffServ Code Points) values can be used.
critical), subscribers (for example, in the IP header, for example. QoS
with gold, silver or bronze profiles) or enforcement is carried out by the radio Queuing mechanisms will typically
operators (especially in situations such (for the air interface) and transport include strict priority (for high-priority
as transport network sharing). (for the transport network) systems. traffic) and weighted fair queuing
for the lower priority classes. The
From the transport point of view traffic number of queues that can be used to
flows can be assigned to QoS classes 5.2QoS enforcement and differentiate the traffic are important,
based on a number of parameters, transport QoS since this determines the level
such as the required packet delay, of granularity, or the number of
delay variation and packet loss. Such With all this in mind, the QoS different traffic classes that can be
parameters are typically universally requirements for a transport network distinguished. The maximum of
good in lightly loaded networks. to support LTE are thus about classes that can be differentiated with
However, as discussed before, the guaranteeing appropriate service VLAN p-bits on the Ethernet layer is
economic transport of LTE data rates levels for each service in terms of eight, so this might be considered a
will lead to a certain overbooking and packet delay, delay variation and useful number of queues.
congestion. In this environment, QoS is packet loss.
the tool that guarantees that, say, voice Resource reservation (hard QoS)
traffic packets preferential treatment The basic functions implemented in Admission control estimates whether
compared with peer-to-peer traffic. transport network elements are there will be sufficient resources for
prioritization and capacity reservations. each new connection or traffic flow.
In that respect it is useful to differentiate In IETF standards they are referred to This functionality is mandatory
between Quality of Experience (QoE) as Differentiated Services (DiffServ) when implementing guaranteed bit
and QoS. The former describes the and Integrated Services (IntServ) rate connections. Static resource
quality of the end-user experience, respectively. Element implementation reservation perhaps via a Network
while the latter is the method used to is often a combination of these Management System (NMS) is a
manage this experience. principles. There will be some resource good option in the mobile backhaul

LTE-capable transport: A quality user experience demands an end-to-end approach 7

sector, for example, by using MEF-type as defined by 3GPP (Table 2). The QCI 5.3QoS must be managed
services with a Committed Information value references a certain application end-to-end
Rate (CIR) and Peak Information type and is used within the access
Rate (PIR). network as a reference for controlling QoS can be a powerful tool to achieve
packet forwarding treatment. a QoE for the LTE end user. It can take
CSPs can deliberately limit the into account the requirements of
throughput of a connection by buffering QCI values are translated into a packet different services, as well as the SLA
the data so as not to exceed the pre- priority marking (DSCP value and/or purchased by the customer. It helps
defined maximum bit rate (shaping), or VLAN p-bits) applied by the BTSs manage resources in congested
by dropping packets that would exceed and gateways. Similarly, the control, environments, especially where theres
the maximum bit rate (policing). Traffic management and synchronization pressure on radio access and the mobile
shaping and policing can also be used plane traffic is marked to ensure it backhaul domain. QoS is an enabling
within queuing systems as congestion receives the right priority treatment on technology for a viable business case.
control mechanisms. the outgoing interfaces.
However, QoS must be managed
QoS implementation Based on these QoS markings, the consistently end-to-end. LTE radio
In a real case, as mentioned before, transport network elements in the QoS has to be aligned with the
the classification and tagging of packets path can then ensure that implementation in the transport network,
traffic is carried out both by the BTSs each packet is handled according to but the transport network also has to
and by the gateways, based on the its required forwarding behavior, for cater for the QoS needs of 3G or even
information collected in the gateways example, by assigning it to the correct more stringently 2G packet traffic.
and the policy server. queues. This can be combined with
connection admission control for the Note, however, that in the same way
For the most typical case of service- transmission network elements, adding as latency, any operator can only
based differentiation, this classification a component of hard QoS. control QoS within his own network. As
relies on the QCI (QoS class identifier) soon as traffic leaves for the internet,
the treatment is essentially best effort.
LTE Radio domain LTE Transport domain
LTE Traffic Class QCI Resource DSCP (Ethernet p bits)
Conversational Voice 1 46 (5)
Conversational Video 2 26 (3)
Real Time Gaming 3 46 (5)
Non-conversational Video 4 28 (3)
IMS signaling 5 non-GBR 34 (4)
Voice, video, interactive 6 18 (2)
Video (buffered streaming) 7 20 (2)

8 10 (1)
TCP-based (e.g. www, email, ftp, p2p
file sharing etc) 9 0 (0)

C-plane 46 (5)
M-plane 34 (4)
S-plane 46 (5)
ICMP 10 (1)

Table 2: An example for mapping radio QoS onto transport QoS

8 LTE-capable transport: A quality user experience demands an end-to-end approach

6. S

6.1Synchronization definition the synchronization clock output from slaves apply intelligent algorithms to
and requirements co-located TDM-based equipment and recover from the received packet
thus effectively relieve the packet stream the original clock information
Synchronization has always been of transport of synchronization duties. This at the Grandmaster.
vital importance in telecommunication is only possible in fully hybrid transport
networks. In mobile networks it is networks and does not cover the need IEEE1588 can provide both frequency
needed for the air interface to enable for phase synchronization. and phase synchronization. However, if
smooth handovers and for aligning used for phase synchronization, all the
coding procedures. Another obvious method would be nodes in the transport network between
to use GPS receivers at each cell the master and slave have to provide
There are two main flavors in site, effectively covering all the on-path support with so-called
synchronization. The most common is synchronization needs that could boundary or transparent clock functions.
frequency synchronization. A standard possibly arise. However, the cost might IEEE1588-2008 can run over any kind
3GPP requirement for all radio be prohibitive, and with cells getting of IP and/or Ethernet network.
technologies is to deliver an accuracy smaller and indoor-hotspot coverage
of the modulated carrier frequency of more of a requirement, it might not 6.2.2 Synchronous Ethernet
better than 50ppb for macro cells. be physically possible to use GPS Synchronous Ethernet (SyncE) is
receivers everywhere. defined in G.8261/8262/8264 as an
In addition, TDD technologies such SDH-like enhancement for transporting
as TD-LTE or WiMAX and some There are two main methods for frequency information on the physical
features such as Multimedia Broadcast packet-based synchronization: layer of an Ethernet link. In contrast
Multicast Service (MBMS) also require with IEEE1588, which is essentially
highly accurate time (or more precisely, 6.2.1 IEEE 1588-2008 a layer 3 technology, frequency
phase) synchronization. In the case of The IEEE1588 solution is standardized synchronization will be extracted
TD-LTE, the maximum timing error and is by far the most common directly from the Ethernet interface
at the air interface must not exceed implementation of packet at the BTS. Unlike packet-based
1.5s. For more detail, refer to table 3. synchronization. It consists of synchronization (IEEE1588-2008,
a Grandmaster (server) at a core NTP), the stability of the recovered
site and Timing Slaves (clients) frequency does not depend on the
6.2Synchronization options implemented in either the BTS or a network load and impairments.
with packet transport transport network element, such as a
cell site device. The master and slaves SyncE has to be implemented
There are a number of ways of communicate through a bidirectional at all intermediate nodes on the
providing high-accuracy synchronization IP protocol called PTP (Precision Time synchronization traffic path. In
information. The most obvious is to use Protocol) containing time stamps. The addition, it does not provide phase
information, so it cannot be the only
synchronization mechanism in the
Standard max. frequency max. timing case of TD-LTE, for example.
error at air interface error at air interface

WCDMA FDD 50 ppb (Wide Area BTS) No requirements. 6.2.3 A choice of three
100 ppb (Medium Range BTS) In fully packet-based networks, only
100 ppb (Local Area BTS) three mechanisms really can be used:
GPS, IEEE1588 and SyncE. Any of
LTE FDD 50 ppb No requirements.
these mechanisms will be useful for
LTE FDD. For TD LTE either IEEE1588
LTE TDD 50 ppb 1.5 s
with on-path support or GPS is the tool
of choice. Of course it is also possible
GSM 50 ppb no requirements. to combine different mechanisms.
100 ppb Pico Class BTS

Table 3: Synchronization is critical

LTE-capable transport: A quality user experience demands an end-to-end approach 9

7. Service assurance

The transformation towards packet- to deploy test traffic to measure the Similar measurements are possible on
based traffic raises service assurance throughput, delay, delay variation and the Ethernet layer per VLAN / p-bit for
challenges. Where TDM-based packet loss between two points in the measuring throughput, delay, delay
technologies had a wealth of operation, network. Different DSCP values can be variation and packet loss per service
administration, and maintenance assigned to the test traffic, allowing and priority class.
(OAM) tools available, packet engineers to measure the network
technologies mostly did not support behavior of various service classes.
them in the past. However, this issue is
now being addressed by the relevant implementations
7.2 Ethernet OAM
industry bodies and standards. Service
Practical implementations can differ
assurance capabilities are increasingly There are several Ethernet OAM-
significantly. Some or all of the OAM
being implemented in packet transport related standards available, which
functions may be implemented and the
equipment. address either a single link or
granularity of performance counters
multiple links
will vary (including real-time vs. history
There are different standards and
counters). The number of active
concepts available depending on the Link Layer OAM (IEEE 802.3ah)
counters may also be limited.
networking layer being used. They This looks at a single link and includes
typically include a set of functions functions for discovering and
Those functions can be implemented in
that enable detection of network faults monitoring the link, as well as
the BTS and core elements (particularly
and the measurement of network indicating remote node failures
important for true end-to-end OAM),
performance, as well as the distribution
or in the transport elements. It can be
of fault-related information. Ethernet Service OAM
useful to have dedicated probes in the
(IEEE 802.1ag / ITU-T Y.1731), a.k.a.
network to monitor specific points in
Service Layer OAM
7.1 IP layer OAM the network or to compare the service
These functions allow monitoring
level of leased lines with contracted
There are multiple options and protocols of end-to-end connectivity and
SLAs by placing the probes directly at
(standardized and proprietary) available performance between the nodes in an
the endpoints.
on the IP layer to provide OAM Ethernet domain. Additional functions
functionality. Some of them allow CSPs are included to support resilience.

8. N
 etwork security

8.1The need for transport tools. In addition, BTSs that were in LTE transport networks, compared
security in LTE traditionally located in secure, locked with WCDMA:
sites are increasingly set up in public
Packet traffic is vulnerable to hacker places. 1. The air interface encryption of the
attacks. Methods have evolved rapidly, user-plane traffic is terminated at
with cheap hardware providing hackers Furthermore, there are two major the BTSs, so user-plane traffic in
with high processing power and better differences that make security different the LTE mobile backhaul network is

10 LTE-capable transport: A quality user experience demands an end-to-end approach

not secured by radio network layer backhaul network and the BTS Probably the most efficient solution for
protocols. site are secure. IPSec provides a the realization of the SEG function at
2. Since the LTE network architecture comprehensive set of security features the cell site is the integration of the
is flat, other BTSs, the EPC nodes (traffic authentication, encryption, SEG in the BTS itself. This minimizes
(MME, S-GW) and other nodes in integrity protection), solving both the physical accessibility, which is especially
the core network become directly problems mentioned above. The 3GPP important in easily accessible hot-spot
IP-reachable from BTS sites. If security architecture is based on IPSec cells. It also reduces the need for
physical access to the site cannot be and Public Key Infrastructure (PKI). additional equipment and reduces the
prohibited, a hacker could connect IPSec is applied between Security site footprint. However, care should be
his device to the network port, attack Gateways (SEG), which are typically taken that any such integrated solution
these network elements and cause located at the cell site and at the has the necessary throughput to
significant network outages. border between the trusted and support LTE data rates, for instance, in
untrusted network. a highly-loaded three-sector cell base
Transport security features are station, and does not add significantly
mandatory unless both the mobile to the overall delay.

Note that such IPSec implementations

and the architecture decisions
Security necessary for the X2 interface are
Gateway (SEG)
closely connected. After all, it is
Security not only the user plane (S1) and
Gateway (SEG) management plane traffic that should
integrated in be encrypted, but also the hand-over
Flexi BTS
Server user traffic via the X2 interface.
Considering that each connection
Core requires an IPSec tunnel, security
eNB architectures can get quite complicated
in a fully meshed architecture.

Figure 6: SEGs guard the CSPs Security Domain

9. C

This paper has discussed the main interact at too many points, whether Close cooperation is therefore required
requirements for an effective and for effective synchronization, QoS between the relevant technical teams.
efficient transport network to support implementation, security or service A partner that is knowledgeable and
LTE and legacy mobile technologies. assurance. Parameters such as experienced in all aspects of network
capacity and latency have to be optimization can prove valuable for
LTE transport cannot be separated planned in an end-to-end manner, CSPs looking to combine maximum
from the LTE radio and core systems because the weakest link will be the subscriber satisfaction with a viable
because these systems have to breaking point. business case.

LTE-capable transport: A quality user experience demands an end-to-end approach 11

10. Glossary

CSP: Communication Service Provider

DSCP: Diffserv Code Point QoS tag on IP layer
EPC: Evolved Packet Core
HSPA: high speed packet access
IPSec: IP encryption methodology
LTE-A: Long Term Evolution Advanced
MME: Mobility management entity part of the EPC
MWR: Microwave radio
p-bit: Priority bit QoS tag on Ethernet layer
PKI: Public key infrastructure key sharing concept for IPSec
PTP: Precision Time Protocol
QoE: Quality of Experience
QoS: Quality of Service
S1 interface: the logical interface between BTS and S-GW and MME gateways / evolved packet core (EPC)
SEG: Security gateway
S-GW: Service Gateway part of the EPC
X2: the logical interface between neighboring BTSs, used e.g. during hand-over

Nokia Siemens Networks

P.O. Box 1
Visiting address:
Karaportti 3, ESPOO, Finland

Switchboard +358 71 400 4000 (Finland)

Switchboard +49 89 5159 01 (Germany)

Order-No. C401-00728-WP-201109-1-EN

Copyright 2011 Nokia Siemens Networks.

All rights reserved.

Nokia is a registered trademark of Nokia Corporation,

Siemens is a registered trademark of Siemens AG.
The wave logo is a trademark of Nokia Siemens Networks Oy.
Other company and product names mentioned in this document
may be trademarks of their respective owners, and they are
mentioned for identification purposes only.

This publication is issued to provide information only and is not

to form part of any order or contract. The products and services
described herein are subject to availability and change
without notice.