the future of radio network load balancing white paper
DESCRIPTION
rfTRANSCRIPT
November 2011
White Paper
The Future of Cellular Radio Network Load Balancing
A special report for mobile operators and optimization engineers
White paper: The Future of Cellular Radio Network Load Balancing
2
Executive summary The RF conditions in the cellular radio network were never static. For load balancing
functions in the predictable world of voice and low volume data, however, this
approximation was close enough. The introduction of smartphones and mobile data
devices brought an exponential increase in mobile data usage, in turn increasing the
unpredictability of subscribers’ demand which causes loading on network resources.
A handful of subscribers, using bandwidth-hungry services, can
push cells to congestion anywhere, anytime.
Traditional optimization methods are slow, based on long term network statistics, and
have a long turnaround time – basically changing from one static network
configuration working point to another. Iterations must be followed by a labor-
intensive verification stage. Operators who chose to maintain these methods might
find themselves facing underutilized resources, premature expansion costs to support
peak loads, and customer dissatisfaction resulting from overcrowded networks.
In order to support the changing network needs, operators require a fully integrated
automated load balancing application with a built in feedback mechanism. The radio
engineers can leave the tedious roles of manual optimization to a machine, and focus
on defining network policies, performance goals and performing radio planning
activities.
A configuration
that was optimal
early this morning
could fall short
before lunchtime
“
”
White paper: The Future of Cellular Radio Network Load Balancing
3
Dealing with dynamic networks Live networks have dynamic RF traffic patterns that change throughout the week, and
over the course of a day. Changes in behavior of voice verses data usage, roads and
business areas compared to adjacent residential areas.
Unexpected load imbalances due to massive gatherings, cell malfunction or
introduction of new cells in an area, all effect the load distribution, and are rarely dealt
with as soon as they occur.
Since dealing with such dynamics is impossible from practical engineering practice
perspective, a preventive RNP (Radio Network Planning) approach is normally taken.
By this approach, cells are dimensioned to handle the busy hour traffic, regardless of
what is occurring in others cells in the area.
The high cost of imbalance
The sunk cost of underutilized resources In Static networks with no time-sharing of geographically distributed resources, costly
resources supporting peak traffic are very often unused.
In some cases, there might be underutilized cells nearby – resources that the operator
already paid for, i.e. sunk costs.
This situation is further emphasized in cases of bursty traffic demand and high
variability of the active users’ locations. Furthermore, exponential increase in demand
for mobile broadband increases this gap greatly.
Denied admissions pattern in two ajacent cells. Note that the peak load does not follow the same time pattern.
At certain times,
underutilized
resources already
paid for might be
available next to
loaded cells.
“
”
White paper: The Future of Cellular Radio Network Load Balancing
4
The mobile data crunch Mobile operators today are facing an avalanche of demand, driven by the mobile data
crunch - fast penetration of smartphones and mobile broadband. The impact is
colossal. According to a mobile data usage study conducted by Cisco, an iPhone
generates as much traffic as 96 non-smart phones; a tablet generates as much as 122
non-smart phones and a laptop with a data card consumes as much traffic as 515 non-
smart phones (Cisco, 2011).
Mobile data traffic prediction by Cisco
The operational cost of networks increases to meet the increasing demands. According
to a network cost analysis conducted by Informa telecoms & media, network costs are
expected to increase by 30% in the next 2 years, and continue in this pace (Informa
Telecoms & Media, 2011b).
To support this increase in traffic would require a similar increase in resources, causing
more increase the underutilized resource gap.
Increased uneven distribution of bandwidth demands
increases this gap greatly “The top 1 percent of mobile data subscribers generate over 20 percent of mobile data
traffic, and the top 10 percent of mobile data subscribers now generate approximately
60 percent of mobile data” (Cisco, 2011). The impact of the concentration of usage on
network planners is that it is becoming increasingly difficult to predict load patterns in
both time and place.
This, in turn, magnifies the resource gap in unpredictable amounts.
A handful of
subscribers, using
bandwidth-hungry
services, can push
cells to congestion
anywhere, anytime.
”
“
White paper: The Future of Cellular Radio Network Load Balancing
5
The “Empty bus” syndrome:
Expansions performed to maintain high level QoS at peak
usage times In order to support the peak traffic, radio planners’ dimensioning rules dictate adding
new cells or resources according to the peak traffic in the busy hours measured per
cell. In unbalanced networks, the load is uneven, and the busy hours are not the same
in all their cells, and there might be underutilized available resources already paid for
in the vicinity of a certain loaded cell.
Overloaded cells cause customer dissatisfaction, increased
churn, and lost revenue The option of leaving the network unbalanced, without expansion will risk
congestions, call setup failures and reduced data QoE at peak traffic conditions.
Leaving the network unbalanced without expansion will limit the data throughput
available to subscribers at peak time, lowering subscriber satisfaction, and possible
loss of revenue.
Existing solutions to handle load balancing Current network optimization processes are handled manually by radio engineers.
The granularity of existing optimization cycles is quite large, and can take days or even
weeks – thus making long term adjustments which are normally based on large scale
time averaging of traffic loading. By their very nature, such solutions can fit long term
or predicted load issues, and will, at best, provide a passible compromise between the
needs of different areas.
What types of solutions are currently available?
Decision supporting tools Using decision-supporting tool to perform the optimization calculations such as
required expansions, RF parameter changes, and the predicted impact on
performance, improves the capability of taking more inputs into consideration.
However, these types of tools provide reports – not actions, and are prone to error
due to the high degree of sensitivity to initial conditions. The radio engineers are still
left with the tasks of verifying the resulting recommendations, updating the OSS /
NMS, and checking the results. This is an open-loop solution, where the entire end-to-
end process includes manual stages to complete.
Local load balancing between carriers Some equipment vendors offer solutions of inter-frequency load balancing. These
solutions can balance loads between carriers - generally co-sectors in the same base
station. Traditional
optimization works
through big
periodical jumps
from one static
working point to
another
White paper: The Future of Cellular Radio Network Load Balancing
6
These solutions, while efficient in resolving localized load imbalance cases, do not
provide a solution for a balancing the load in a cluster of cells, and require an
infrastructure of multiple carriers in each sector.
Wi-Fi or Femto offloading For local areas with consistent capacity problems, operators can elect to offload the
data portion to a local Wi-Fi or femtocell. However, according to a recent report by
Informa, in order to extract value from Wi-Fi offload mobile operators will require
carrier-grade Wi-Fi networks that are more tightly integrated into the operator’s
network and back office environments than at present, and deployment of which will
incur significant costs (Informa Telecoms & Media, 2011a).
Additionally, both Femto and Wi-Fi offloading are complex techniques which require
high level of backoffice configuration management, installation and supporting
equipment, on the network side (such as Femto Gateways) or in the UE side (the client
has to support controlled Wi-Fi offloading) etc. These reasons make those techniques
non trivial and not suitable for every operator.
In any case, this solution can only add capacity to a fixed location, in the vicinity of the
Wi-Fi AP of the Femto itself, and does not provide a solution to congestion situations
that change over time and place.
Furthermore, by passing the data to a separate network, the operator faces a potential
loss of revenue, and has less control over the QoS and SLA’s toward the customers.
With Femtos current 3GPP architecture, the operator needs to decide if to dedicate a
separate carrier for the Femto deployment (which reduces the utilization of such
carrier which could have been used in the macro network) or work in an intra-
frequency mode, which has its drawbacks in the form of need to manage the Femto –
Macro layers coexistence.
LTE offloading Another solution is offloading heavy data traffic to an LTE network. However, this
elephant-gun approach is not financially justifiable for most operators for several years
to come (Informa Telecoms & Media, 2010).
In reality, many of these solutions were not designed to contend with the degree and
extent of variation in usage that we see unraveling today. Another important factor in
the transition to LTE is the ecosystem maturity. Not only does there need to be solid
UE support for multi-mode GSM/UMTS/LTE but there has to be a viable penetration
rate of such devices to the users population in order to claim that LTE can be an
affective offloading solution. Additionally, spectrum resources are always an issue,
and, at least in the first phases of deployment, since UMTS carriers cannot be
evacuated, new spectrum will be needed to activate LTE, which means more
investments from the operator side.
White paper: The Future of Cellular Radio Network Load Balancing
7
The 3G SON solution: Fully automated load
balancing To face these realities mobile operators need enhanced functionality in the existing
UMTS infrastructure that can respond to demand patterns as they form and change. A
network configuration that was optimal early this morning could fall short before
lunchtime; what is right for a certain cell could be all wrong for its neighbor.
Enhanced network responsiveness by full automation The best network engineers, working with
the finest tools can probably make no more
than 100 - 200 optimization adjustments per
month. Automated 3G SON (Self Optimizing
Network) load balancing or other
applications can do many thousands of
adjustment a day in a network, allowing the
engineers to focus on radio planning
engineering and design tasks. The endless
grunt work of tweaking, analyzing and
tweaking again is more efficiently handled
by machines.
Automated RF shaping to increase the efficiency of the
network By means of automated RF shaping, a system can modify the footprint of the
surrounding cells to the current usage demand and match the subscriber distribution
to the available resources. Using RF shaping increases the efficiency of the network,
and increases the utilization of existing resources.
Can this be truly automated? The impact of performing many changes can be disastrous if those changes are not
carefully monitored. This is the purpose of the SON’s feedback mechanism which
verifies the effect of the changes on the network.
SON revolutionizes the level of automation in operations and maintenance and
significantly decrease the OPEX associated with operations and maintenance. OPEX
Savings estimated at 65-80% (Motorola, 2009).
In order to be fully automated, a SON system must change the network configuration
in small scale, cell level iterations, verify the quality of the change performed, and
continuously compare the performance metrics to the policy targets of the operator.
The SON cycle
White paper: The Future of Cellular Radio Network Load Balancing
8
Case study #1: 3G-SON load balancing lowers power
load by 20%
This case study demonstrates an activation of an RF shaping based intra frequency load balancing SON
application on a busy cluster. Activating a 3G automated load balancing application on a cluster of sites
lowered the radio resource load on the site by 20%, transferring the load to nearby sites with shared coverage.
Once the application was deactivated the load returned to the previous values.
UEs can be moved between cells by means of RF shaping – decreasing the size of the loaded site, and
increasing the size of the neighboring cells. The cells’ relative sizes can be continuously modified to fit the
current load conditions in the area covered as these conditions change.
Before load balancing After load balancing
RF shaping moves subscribers from the loaded cell to neighboring cells
0%
20%
40%
60%
80%
100%
20:00
20:15
20:30
20:45
21:00
21:15
21:30
21:45
22:00
22:15
22:30
22:45
23:00
cluster power load
NBR #2
Loaded site
NBR #4
Average Neighbors 1,3,5
Load Balancing
activated
White paper: The Future of Cellular Radio Network Load Balancing
9
Case study #2: 3G-SON load balancing lowers admission
rejections by ~100%
This case study demonstrates the ability of a 3G SON load balancing application to lower the denied
admissions. Compared to busy times on different days, when the load balancing application was activated on
the cells the number of denied admissions per PM report dropped to 0.
What to look for in an ideal load balancing application The lists below details what to look for in an ideal automated load balancing application, and how it can deal
with the load challenges of modern cellular networks
Fully automated end-to-end balancing solution, shifting traffic between cells, based on availability,
congestion, and blocking of radio resources
Support for intra-frequency load balancing in a cluster of sites as opposed to single RBS balancing between
carriers
Contains an automatic feedback method of verifying the impact of the adjustments, and correcting them if
necessary
Based on real time performance and loading conditions and not extrapolated historic averages
Rapid response time – identifying and correcting interference issues as they occur
Standard implementation and vendor agnostic solution to avoid implementation surprises.
Configurable to set performance goals and policies
An optimization cycle of minutes, rather than days or weeks
Graphic interface with real time traffic statistics and current RF conditions
0
50
100
150
6:45
7:15
7:45
8:15
8:45
9:15
9:45
10:15
10:45
11:15
11:45
12:15
12:45
13:15
13:45
14:15
14:45
15:15
15:45
16:15
16:45
17:15
17:45
18:15
18:45
19:15
19:45
20:15
20:45
21:15
21:45
22:15
22:45
# o
f D
en
ied
Ad
dm
issi
on
s
Denied Admissions
01/06/11
08/06/11
15/06/11
Activated load balancing
White paper: The Future of Cellular Radio Network Load Balancing
10
Conclusions Unbalanced networks are increasing operators’ annual costs in both lost revenue, and sunk cost of
underutilized resources. The mobile data crunch is creating sharper, more localized dynamic traffic patterns
than ever before.
The current manual solutions including decision supporting tools cannot meet the increasing need of load
balancing in terms of reaction time and accuracy.
Automated 3G-SON load balancing solutions can identify loads in near-real time, change the RF footprint of
cells to shift users from loaded cells to unloaded cells, and verify the impact on the network, all without human
intervention.
About Intucell Intucell delivers the world’s most advanced SON solutions in the market today. The company’s SON systems
are deployed by a number of leading mobile operators worldwide. Powered by real time network
visibility, Intucell’s systems automatically tune the network to actual conditions as they develop and change.
Intucell is a private international company backed by blue-chip investors, with offices in the United Kingdom
and Singapore and an R&D center in Israel.
To find out more about how Intucell can help you meet your network goals, visit www.IntucellSystems.com, or
contact our representatives at [email protected]
Sources Cisco. (2011, Feb.). Cisco Visual Networking Index: Global Mobile Data traffic Forecast Update, 2010-2015.
Retrieved from Cisco:
http://newsroom.cisco.com/dlls/ekits/Cisco_VNI_Global_Mobile_Data_Traffic_Forecast_2010_2015.pdf
Informa Telecoms & Media. (2010, November 16). LTE world. Retrieved from UK mobile broadband network
upgrade to LTE not economically viable until 2015: http://lteworld.org/uk/uk-mobile-broadband-
network-upgrade-lte-not-economically-viable-until-2015
Informa Telecoms & Media. (2011a, January 4). Mobile operators need offload to be smart and cost effective.
Retrieved from Telecoms.com: http://www.telecoms.com/23817/mobile-operators-need-offload-to-be-
smart-and-cost-effective/
Informa Telecoms & Media. (2011b, March 30). Modeling mobile broadband networks costs: LTE and offload
case studies. Retrieved from Informa Telecoms & media:
http://webinars.informatm.com/2011/03/31/modelling-mobile-broadband-network-costs/
Motorola. (2009). LTE Operations and Maintenance Strategy Using Self-Organizing Networks to Reduce OPEX.
Retrieved from Motorola Solutions:
http://www.motorolasolutions.com/web/Business/Solutions/Industry%20Solutions/Service%20Provider
s/Network%20Operators/LTE/_Document/Static%20Files/LTE%20Operability%20SON%20White%20Pap
er.pdf