Professional Documents
Culture Documents
SOS:
Signalling Overload
due to Smart devices
Dr Konstantinos Stavropoulos
IOT Product Manager, Anite
FEBRUARY 2011
Contents
1
Executive summary
Introduction
Always (?) on
SOS evaluation
Appendices
Anite 2011
1 EXECUTIVE SUMMARY
An indispensable part of our daily routine, smart devices are causing many wireless networks to
suffer, and not only due to the insatiable user demand for data. These always on, signalling-hungry
devices do not communicate with the wireless network in a smart or efficient way. This signalling
issue has been amongst the least publicised up to now, and is the subject of this paper Signalling
Overload due to Smart devices: SOS.
Smart devices give the impression of being always on, whereas in reality they repeatedly wake
up to ping the network for updates and then go back to sleep. A 3G device can be in different
radio states, which can be simply described as High, Low, Standby, and Idle. In brief, the High
state supports data throughput with the most network resources reserved and the device battery
burden at its peak, while Idle is where devices stay dormant, save power and are periodically active.
Signalling accounts for as much as 60% of an operators network costs. Applications such as
mobile email or social networking demand constant updates and a smartphones signalling may
be eight times less efficient than a PC/laptop dongle, and does not bring in any extra revenue. SOS
will become an even bigger headache as smart devices get cheaper, more user-friendly, and provide
access to more/better content (apps).
SOS influences user experience too. Signalling affects network connections and battery life.
Subscribers expect a lot from smart devices and are likely to use them less or even churn if these
do not live up to expectations. Ultimately, an operators profit can be negatively affected by SOS.
Reactive methodologies such as real-time monitoring and field testing help to understand the
impact of SOS. However, they are costly and have limited preventative use. In contrast, laboratory
testing is the proactive approach that has helped accelerate the time to launch new devices in a
cost-effective fashion. Using a network simulator in the lab, it is possible to test and compare the
signalling behaviour of different smart devices and applications before they are launched.
As the SOS issue is acknowledged across the industry, operators have started to look at ways
to enhance the interoperability of smart devices with their networks by considering additional
network resources, new technology features and/or more network intelligence. More importantly,
operators want to get the ecosystem with manufacturers and application developers working
effectively, in a similar fashion to the successful operator-driven acceptance programmes that have
substantially reduced the time to market of new devices. Operators can simply not afford to ignore
this SOS.
Anite 2011
2 INTRODUCTION
Many mobile industry experts have been surprised by the remarkable success that smart devices
have enjoyed in the last few years. Having become an indispensable part of our daily routine, these
devices now dominate the news headlines on a regular basis.
As the mass appeal of smart devices increases, issues relating to their operation and performance
gain in significance. This document focuses on one of these issues, which has probably been
amongst the least publicised up to now: Signalling Overload due to Smart devices.
This paper is organised as follows: first, we look at the emergence of devices classified as smart,
then we discuss the related topic of mobile device radio states, investigate what always on
actually means for smart devices, analyse the impact of these devices on network signalling, and
finally describe the key methodologies used to evaluate and to address this impact.
Anite 2011
As wireless networks started to become more and more successful, there were very few handsets
that could be called smart. In effect, these smart handsets or smartphones were seen as a niche
market segment, which could address the needs of relatively speaking a small number of
mobile users. The status quo has now changed dramatically, with (first) the advent of mobile email
and (more recently) the unprecedented success of mobile applications or apps that the launch
of the iPhone spearheaded.
Smartphones have been the mobile device success story for the last two to three years, and the
trend is set to continue. The demand for smartphones has been driving the growth of handset
sales, and specialised vendors such as Apple and RIM now find themselves in the Top-5 mobile
device manufacturers. In addition to mobile handsets, other smart devices appear to be quite
successful: wireless tablets (such as the iPad) have been launched, with sales likely to exceed the
50 million-unit threshold as early as 2011. Yet, this smart device boom, which has led to 600
million mobile broadband subscriptions in 2009, has come at a price.1
As one of the main drivers behind the mobile data explosion phenomenon, smart devices are
also the reason why many wireless networks worldwide are now suffering, and not only because
of the insatiable user demand for data. Indeed, these always on, signalling-hungry devices are
not currently communicating with the wireless network in a smart or efficient way, which is
causing concern for many 3G network operators. The situation may actually deteriorate with
next-generation mobile systems (such as LTE) and when wireless devices are embedded in every
consumer machine.
It therefore seems that, at least with regard to network signalling, smart devices may have been
smart-looking, fashionable gadgets that have dazzled the public rather than the intelligent, efficient
contraptions that their naming implies. But is the Signalling Overload due to Smart devices an
issue of real concern, an SOS for mobile network operators, or an exaggerated myth?
For simplicity, the term smart device has been used in this document to refer to smart mobile devices, irrespective of their handset/
tablet/other nature. The term does not include mobile network dongles or embedded PCI cards.
Anite 2011
2
This document focuses on 3G, as it is WCDMA/HSPA networks that have driven the proliferation of smart devices by being able to
support data throughput rates comparable to those of fixed-line networks.
Anite 2011
Figure 1 illustrates these different states in a simplified fashion. It also depicts the highlevel relationship of each state with the device power (battery) and data (throughput/speed)
performance as well as the burden on network resources.
DCH / HSPA
FACH
Power
URA / PCH
IDLE
IDLE
STANDBY
LOW
HIGH
CONNECTED
Data Rate | Network Resources
Figure 1
A simplified radio state transition diagram for mobile (3G) devices
In summary, with regard to device power, data throughput and network resources, the state of the
mobile device can be:
High, to achieve the highest possible data throughput (and shortest latency), with the most
network resources reserved and the device battery burden at its peak
Idle, to stay dormant and save power by only being periodically active, with data transmission
not supported.
So, a mobile device will generally switch from one state to another as depicted in Figure 1,
depending on the use scenario. Typically, transition to a more active state will be triggered by
data packets (with the High or Low state selected based on data volume), and transition to a less
active state will be triggered by device inactivity (based on an Inactivity timer). It is this state
transitioning that causes increased signalling from smart devices, particularly when the device is in
sleep mode and needs to wake up and move to the High state.
The original aim of the 3GPP standard was for devices to move to the power-efficient Standby
(PCH) rather than the Idle state after data transmission is over. The Idle state was to be avoided as
transitions to the High state require a packet connection setup, and hence increased latency and
network signalling. However, the Standby state has not been deployed in many mobile networks,
while the configured Inactivity timers for DCH and FACH are relatively long. This is why fast
dormancy, which is referred to in the next section, has been considered by some device vendors.
Anite 2011
5 ALWAYS (?) ON
The emergence of smart devices has monopolised the industry headlines, mainly with regard to
user data throughput and network capacity demands. To a large extent, this was inevitable as
mobile networks were not designed to support the currently experienced data tsunami. Much
emphasis has also been put on the always on mode of operation of smart devices. What is often
not appreciated though is that these devices give the impression of being always on, whereas they
repeatedly wake up to ping the network for updates and then go back to sleep.
The use pattern for smart devices can be summarised as follows:
Users tend to pick up their smart device for a short period and in a highly unpredictable
manner, to make use of an application that requires manual action (e.g. to browse the Internet
for the latest news or find the nearest restaurant to go for lunch)
Mobile applications rely on updates via network server polling, often at user-specified internals
that can be as short as a minute, while push-enabling applications (including Instant
Messenger) require a TCP connection and the use of heartbeat or keep-alive messages, as
often as every 30-60 seconds.
It is this constant signalling need that led some device vendors to consider fast dormancy before
the Standby state was deployed. This device-proprietary feature overrode the long (batterydraining) Inactivity timers, by going directly from High to Idle quickly after data transmission
ended. The idea behind fast dormancy is simple: the device sends a Signalling Connection Release
Indication or SCRI to simulate a signalling connection failure, release the RRC connection and move
to the Idle state, where power consumption is low.
However, this direct transitioning proved problematic, due to the non-standardised, devicedependent implementation of fast dormancy3 and the fact that the simulated signalling
connection failures could not be distinguished from actual failures. Fast dormancy increased
signalling by increasing the amount of Idle-High state switching, which requires the setup of a
packet connection. It also made it impossible for the device to stay in the preferred (Low) state.
More comments on Fast Dormancy as a 3GPP-standardised feature are made later in the document.
Anite 2011
The constant updates that many popular applications (such as mobile email or social networking)
require from the network are the reason why smart devices are often described as chatty. Indeed,
each such update can generate as many as 30 signalling messages, equivalent to what a voice call
would require. In these terms, it is easy to understand what kind of an impact a device that asks for
updates every minute or so would have on the mobile network. More importantly for operators,
such updates cost in terms of signalling currency, but do not bring in any extra revenue (at least,
for the time being).
According to mobile network operators figures, a smartphone generates more signalling messages
per megabyte (MB) than a PC/laptop dongle. The average smartphone user consumes from 10
to 25 times less data than the average mobile broadband user (with the introduction of LTE, this
figure will probably be adjusted by a factor of two to three. At the same time, mobile dongles only
connect to the network three times as often as smart devices. In other words, a smart device may
make eight times as many connection attempts per MB as a dongle, i.e. may be eight times less
efficient than a dongle in terms of signalling.
Figure 2 below illustrates this in a simplified normalised fashion. In this figure, the data (in MB),
number of connection attempts and connection attempts /data (MB) for smartphones and dongles
are presented versus the respective normalised maximum value (this value is not the same or of
the same unit for each depicted graph).
Normalised Max
Normalised Max
DATA
SMARTPHONE
Normalised Max
CONNECTION ATTEMPTS
DONGLE
SMARTPHONE
DONGLE
CONNECTION ATTEMPTS/DATA
SMARTPHONE
DONGLE
Figure 2
A simplified normalised illustration of how smartphones and dongles rate in real-life networks
with regard to: data (MB) consumption; number of connection attempts; number of connection
attempts per data (MB)
Operators now see signalling traffic in their networks outgrow data traffic by 30%50%. And, as
the number of chatty smart devices increases, so will the impact on network signalling.
Of course, the effect on different networks will vary, and there may be many mobile networks
where Signalling Overload due to Smart devices (SOS) has not been noticeable, yet.
Anite 2011
With the European Union ready to impose a cap on roaming fees, a leading mobile operator
recently announced a flat-rate data package for smartphone customer roaming in Europe. As
more and more operators are likely to follow a similar strategy, the issue of signalling
overload may thus be exported to countries where mobile networks are able to
cope with local demand but not with additional smart devices. And this could well lead to
many unhappy mobile users (even though the reason will not be roaming data fees).
The number of applications for use with smart devices has increased rapidly in just a few years,
and is set to increase further. Apple is now claiming 300,000 mobile apps for the iPhone, while
competitor platforms/devices are doing their best to catch up. With more applications trying
to connect to the mobile network, more signalling issues are likely to occur.
Anite 2011
The use of social networking applications is rising. There are now 200 million mobile Facebook
users, who are typically twice as active as fixed-line Facebook users. According to GSMA, 50%
of the UK mobile data traffic is social networking related. More importantly, many applications
may involve more than one person, especially as users start using their smart device in the
same way they would use their PC/laptop.
As broadband is to become a human right, support for mobile broadband may need to be
extended in many countries. Although not limited to smart devices, such a development could
be a challenge for mobile networks that have not faced problems up to now.
It is also important to note the commercial impact of signalling overload. According to leading
network infrastructure vendors, as much as 60% of the network resources may be dedicated
to supporting connection attempts and only 40% to data throughput. In economic terms, the
signalling related expense may thus account for 60% of an operators network cost. So, while the
amount of data that a mobile user consumes is driving the operator revenues, it is the number and
length of times that the users device gets connected to the network that actually cost more.
The impact of smart device signalling is not restricted to network operation and efficiency. User
experience is also of great significance, as signalling can directly/indirectly affect the mobile
subscriber, from the ability to connect to the network to how long the device battery lasts for.
Subscribers expect a lot from their smart devices, and are likely to be disappointed and use
them less or even churn if these do not live up to expectations. Ultimately, the bottom line for
operators is inevitably related to signalling overload.
Signalling Overload due to Smart devices is a significant issue for mobile network operators. Of
course, the SOS urgency will differ from operator to operator and from device (or application) to
device (or application). Operators are now interested in methodologies and tools that will help
them quantify the issue for different devices/applications in their network.
Anite 2011
10
7 SOS EVALUATION
Reactive or proactive methodologies can be used to understand the impact of smart devices
on the signalling health of the mobile network. The most representative ways to measure the
signalling temperature of the network or run a signalling check-up are briefly described below.
Real-time monitoring
Mobile network operators have been interested in real-time or near-real-time monitoring systems
for a while. From cell focused network performance management (based on post-processing
cell measurements), to call/user specific evaluation tools (based on detailed data from network
traces and/or GPS), such systems have been deployed worldwide with considerable success.
These monitoring schemes are also related with the elaborate network optimisation (and selfoptimisation) mechanisms that operators are now looking into, as part of next-generation systems.
Real-time monitoring systems are useful in identifying various network issues, including signalling
overload due to smart devices. However, such systems are generally costly and complex, due to
the large amount of considered data and the processing speed they may need to provide. More
importantly, their use for SOS prevention is limited, as they rely on a posteriori analysis. In other
words, real-time monitoring systems cannot prevent the launch of devices/applications that may
wreak network havoc. Even if test devices are considered for this purpose, these could have a
negative impact on the live mobile network and on its users.
Field testing
Field testing (also known as drive testing, if performed outdoors) has been a popular way for mobile
network operators to understand how their network is performing. Field testing is a useful tool
especially when a new and not so well tested or mature technology is considered, such as LTE.
Field testing can be used by operators to check the signalling performance of smart devices too.
Indeed, the information that is recorded in a field trial can help operators understand exactly how
a device/application is behaving in their network. However, field testing is subject to the dynamic
network conditions and is not statistically bullet-proof. More importantly, similar to real-time
monitoring systems, field tests can be costly, have limited preventative use and may also affect the
performance of the live network.
Anite 2011
11
Lab testing
Lab (laboratory) testing has been the proactive approach to testing mobile devices/applications
for a number of years. Lab testing has enabled operators and device manufacturers to test devices
in a controlled and repeatable environment, which is immune to statistical uncertainty. Based on
systems that simulate a cell or a number of cells, lab testing has helped identify device issues early
and accelerate their time-to-market in a cost-effective fashion.
Figure 3 below depicts in a simplified manner the setup of the leading network simulator solution
for device interoperability testing. The simulated network system shown at the top can be driven
locally or remotely via a PC/laptop and is connected to the mobile device under test via an RF
cable. As shown in Figure 3, the device interoperability tests can also be run in an automated mode
via remote control and without manual intervention.
ANITE SAS
RF Connection
Remote Control
Device Control
Automation PC
Figure 3
A simplified (Anite SAS) system diagram for smart device interoperability testing in the lab
This setup can be extended to consider RF fading, and hence simulate the dynamic mobile network
environment in an even more realistic fashion. In addition, mobile applications can be tested by
connecting the system to internal or external data servers. In these terms, it is possible to quantify
the impact of chatty smart devices on network signalling, and test devices/applications thoroughly
with regard to SOS, before these are launched.
Anite 2011
12
Using a network simulator, different smart devices can be evaluated and their behaviour in terms
of signalling measured as part of comprehensive device performance/comparison tests. Network
simulation can reveal how long a smart device stays in a particular state, how many transitions
occur in representative test scenarios for commonly used mobile apps, and ultimately the potential
signalling impact of the device and its applications on the real-life mobile network.
Such signalling tests can be considered together with battery, acoustic quality, latency, data
throughput or other tests that have been run in the lab for a while to make the assessment of
smart devices more representative. This kind of testing is also expected to become part of the
device acceptance schemes that many Tier-1 operators have introduced to enhance the quality of
smart devices launched on their network and gain the advantage over their competitors.
In brief, lab testing can help operators assess the signalling impact of smart devices before they
are deployed and with no/minimal need to test them on the live network. In terms of cost versus
benefit, lab testing appears as the most attractive means to understand if the introduction of
a new device/application will have an adverse effect on the mobile network and to decide on
whether action should be taken at an early stage.
The typical actions that a mobile network operator may take to ensure that the smart device
boom will not lead to the network going bust are described in the next section.
Anite 2011
13
4
In accordance with 3GPP, Fast Dormancy can be supported by devices that are not Release 8 compatible, as it involves limited changes
to the relevant radio protocols. It is also important to point out that there are more 3GPP Release 7 and Release 8 features of interest in
terms of signalling (including Cell PCH) or battery life efficiency that have been or will be implemented in mobile networks,
and which are not commented herein.
Anite 2011
14
Anite 2011
15
APPENDICES
About Anite
Anite provides a comprehensive range of critical IT solutions to the Wireless and Travel sectors
across the globe. Listed on the London Stock Exchange, Anite develops and implements software
as well as provides consultancy, systems integration and managed services to ensure that our
customers operate effectively and securely. By using the latest technologies to deliver quality and
cost-effective solutions, Anite meets customers specific requirements and realises tangible results
for its clients.
Anite Telecoms, a subdivision of Anite plc, is a global market leader with over 18 years of
experience in providing cutting-edge technology within the handset testing industry to device
manufacturers, operators and test laboratories. As an established leader, Anite was the first
company to verify LTE conformance test cases in 2009. With highly flexible and reliable software
solutions, Anite is known as an innovative, agile and responsive partner to the top players in
the Telecoms industry. With a diverse team focused on making a difference and exceeding
expectations, Anite draws upon worldwide expertise and is enthusiastic about the future.
Headquartered in the UK, Anite employs around 500 staff in 13 countries across Europe, America,
Asia and the Middle East. For more information, please visit www.anite.com
ABBREVIATIONS
2G 2nd-Generation
3G 3rd-Generation
3GPP 3rd-Generation Partnership Project
DCH Dedicated CHannel
DL DownLink
FACH Forward Access CHannel
GPS Global Positioning System
HSPA High-Speed Packet Access
LTE Long Term Evolution
M2M Machine-to-Machine
MB MegaByte
PC Personal Computer
PCH Paging CHannel
PCI Peripheral Component Interconnect
RRC Radio Resource Control
SAS Anites network simulator solution (originally introduced as Stand-Alone Simulator)
SCRI Signalling Connection Release Indication
SOS
Signalling Overload due to Smart devices
TCP Transmission Control Protocol
UL UpLink
URA UTRAN Registration Area
UTRAN UMTS Terrestrial Radio Access Network
Anite 2011
16
APPENDICES
FIGURES
Figure 1 : page 6
A simplified radio state transition diagram for mobile (3G) devices.
Figure 2 : page 8
A simplified normalised illustration of how smartphones and dongles rate in real-life networks
with regard to: data (MB) consumption; number of connection attempts; number of connection
attempts per data (MB.)
Figure 3 : page 12
A simplified (Anite SAS) system diagram for smart device interoperability testing in the lab.
REFERENCES
WCDMA for UMTS HSPA Evolution and LTE, H. Holma and A. Toskala, 5th Edition, 2010, Wiley
www.anite.com/wireless
SOS: Signalling Overload due to Smart Devices, represents the view of Anite as at the date of publication. The whitepaper is provided
as is and for the purpose of information only. Anite does not make any warranty of any kind, expressed or implied, as to the information
presented within the document, and does not undertake to notify recipients of this whitepaper of any subsequent changes to the facts
and opinions included.
Anite 2011
17