Aalborg Universitet

Multi-Layer Traffic Steering

Fotiadis, Panagiotis; Polignano, Michele; Gimenez, Lucas Chavarria; Viering, Ingo; Sartori,
Cinzia; Lobinger, Andreas; Pedersen, Klaus I.
Published in:
2013 IEEE Vehicular Technology Conference (VTC Spring)

DOI (link to publication from Publisher):

10.1109/VTCSpring.2013.6692643

Publication date:
2013

Document Version
Publisher's PDF, also known as Version of record

Link to publication from Aalborg University

Citation for published version (APA):

Fotiadis, P., Polignano, M., Gimenez, L. C., Viering, I., Sartori, C., Lobinger, A., & Pedersen, K. I. (2013). Multi-
Layer Traffic Steering: RRC Idle Absolute Priorities & Potential Enhancements. In 2013 IEEE Vehicular
Technology Conference (VTC Spring) IEEE. I E E E V T S Vehicular Technology Conference. Proceedings,
DOI: 10.1109/VTCSpring.2013.6692643

General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

? Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
? You may not further distribute the material or use it for any profit-making activity or commercial gain
? You may freely distribute the URL identifying the publication in the public portal ?
Take down policy
If you believe that this document breaches copyright please contact us at [email protected] providing details, and we will remove access to
the work immediately and investigate your claim.

Downloaded from vbn.aau.dk on: oktober 14, 2018

 Multi-Layer Traffic Steering: RRC Idle Absolute
Priorities & Potential Enhancements
Panagiotis Fotiadis1, Michele Polignano1, Lucas Chavarria1, Ingo Viering2, Cinzia Sartori3, Andreas Lobinger3, Klaus Pedersen4
1 2 3 4
Aalborg University Nomor Research GmbH Nokia Siemens Networks Nokia Siemens Networks
Department of Electronic Systems Munich, Germany (DE) Research Radio Systems, Munich (DE) Research Radio Systems, Aalborg (DK)
{paf,mpo,lcg}@es.aau.dk [email protected] {name}.{surname}@nsn.com {name}.{surname}@nsn.com

Abstract— This paper investigates the potentials of traffic UE RRC state. Both RRC Connected and Idle TS mechanisms
steering in the Radio Resource Control (RRC) Idle state by are strongly coupled with the respective mobility procedures,
evaluating the Absolute Priorities (AP) framework in a multi- whereas decisions at the state switching are performed via
layer Long Term Evolution (LTE) macrocell scenario. Frequency redirection information that instructs the UE to connect or
priorities are broadcast on the system information and RRC Idle
camp on a specific carrier (Table I).
users can be steered towards higher priority carriers whenever
coverage allows it. However, such an approach may overload the Prior state-of-the-art focuses on utilizing TS primarily for
prioritized layers. For that purpose, an enhanced scheme is load balancing purposes. In [4-5], RRC Connected schemes
proposed, where priorities are adjusted on a user basis and are are applied for biasing handover events in favor of under-
provided to the terminal via the connection release signaling. The utilized cells whereas a theoretical approach on TS exploiting
priority adjustment is based on both the Composite Available the connection setup phase is adopted in [6]. Compared to
Capacity (CAC) and the radio conditions of the candidate layers. RRC Idle, the aforementioned techniques react faster to load
Compared to broadcast AP, the proposed scheme achieves better fluctuations and resolve potential overload conditions.
load balancing performance and improves network capacity, However, without a proper idle mode policy, the cost in
given that the User Equipment (UE) inactivity periods are not
signaling overhead might be large. Examples of RRC Idle
significantly long. Finally, better alignment between the RRC
Connected and Idle mobility procedures is observed, guarantying studies can be found in [7-8]. Nevertheless, those former
significant decrease of handovers/reselections and potential studies only analyze the impact of Idle Mode mobility
battery life savings by minimizing the Inter-Frequency (IF) management on the UE power consumption.
measurement rate in the RRC Idle. In this paper, we focus on the 3GPP-defined Absolute
Priorities (AP) framework [9] and evaluate its performance as
Keywords-LTE, Load Balancing; Mobility; Absolute Priorities; a load balancing mechanism via the RRC Idle state. The study
Self Organizing Networks (SON); Radio Resource Control (RRC) is performed on a co-sited LTE macrocell scenario consisting
I. INTRODUCTION of 3 carrier frequencies at 800 MHz, 1800 MHz and 2600
MHz respectively. According to AP, frequency priorities are
According to traffic predictions [1], mobile broadband is broadcast on the system information and UEs reselect to the
growing exponentially and communication service providers higher priorities layers whenever coverage allows it. The
should upgrade their deployments accordingly, in order to broadcast scheme (common frequency priorities for all UEs) is
meet the future capacity requirements. Network evolution compared against a developed SON-based algorithm that
dictates for the migration towards multi-layer deployments dynamically adjusts the priorities on a UE resolution during
consisting of multiple overlaid networks with different the connection release procedure.
characteristics in terms of carrier frequencies (pathloss The paper is organized as follows. Section II outlines the
properties), Radio Access Technology (RAT), cell sizes, etc. basic TS schemes depending on the RRC state. In section III,
However, this diverse communication environment introduces
additional Operation Administration and Maintenance (OAM)
Table I. TRAFFIC STEERING SCHEMES
complexity, requiring self-optimizing traffic management
techniques for efficiently exploiting the vast pool of network RRC State Method
resources. In principle, the autonomous traffic management is
via forced handovers and/ or dynamic
only one of several automation deployment elements that the Connected
mobility parameters tuning (range extension)
3rd Generation Partnership Project (3GPP) has defined within
the Self-Optimizing Organizing Networks (SON) framework Idle via basic cell biasing and priority thresholds
[2] (i.e. mobility optimization, interference management,
energy saving, etc). Idle-to-
Traffic steering (TS) is defined as the ability to control and via redirection to the proper layer during the
Connected
connection setup phase
direct traffic to the best suitable cell/layer [3]. Specific criteria Transition
such as cell load, UE speed, power consumption, UE
capabilities and backhaul capacity could be utilized for Connected-to- via redirection to the proper layer during the
adaptive TS decisions depending on the operator’s policy. Idle Transition connection release phase
Additionally, a further classification can be done based on the

978-1-4673-6337-2/13/$31.00 ©2013 IEEE

the AP framework is described along with the UE-dedicated UE connects to current
priority adjustment proposal. Simulation assumptions are cell if no TS at
connection
presented in Section IV, followed by the numerical results in establishment is applied
Section V. Finally, Section VI concludes the paper.
RRC RRC
II. TRAFIC STEERING & RRC MODE ALIGNMENT Connected Idle

Fig. 1 depicts the interaction between the UE RRC states

and the potential schemes in terms of steering a user towards a UE camps on current
cell if no TS at
cell. The curved arrows (initiating from/ ending at the same connection release is
state) represent the handovers and reselections that a terminal applied
performs while being active or idle respectively.
TS in RRC Idle is rather challenging as the network has less Network Knowledge: Network Knowledge:
accurate information about the UE location, compared to RRC Load, cell, QoS, radio Minimal, only in terms of
channel conditions, etc Tracking Areas (TA)
Connected. In principle, cells are grouped in Tracking Areas
(TA) and idle users autonomously reselect to nearby eNodeBs Figure 1: Interworking between RRC states and TS policies
(eNB) according to the provided information that is available
on the broadcast channel (priorities, thresholds, minimum (1)
acceptable signal power/ quality, etc [9]). Hence, reselections
between cells that belong to the same TA are transparent to the - Criterion 2: A UE reselects to a lower priority cell N if:
network. Despite the aforementioned network knowledge
limitation, optimized RRC Idle management can provide
& (2)
significant gains in terms of control signaling minimization by
eliminating idle-to-connected ping pong events. An idle-to-
connected ping-pong is declared whenever a user that switches where ServingMeas and TargetMeasN correspond to the signal
to connected, is immediately handed over to a different cell power/quality measurements of the currently camping cell and
either due to radio conditions or load balancing purposes [3]. the IF target cell N respectively. Whenever (1) is satisfied for
RRC Idle policies can be supplemented with redirections at more than one cell that belong to frequencies with different
the connection establishment. In such a manner, TS decisions priorities, the UE reselects to the layer with the highest
can be further enhanced by exploiting the load information at priority. Absolute thresholds (ThreshXhigh, ThreshXlow,
the UE state switching. Nevertheless, additional latency in the ThreshServingLow) and frequency priorities can be provided to
connection setup time is introduced, as the user has to re- the UE by either the system information (broadcast) or via the
initiate the connection establishment procedure at the RRC CONNECTION RELEASE message. Note that for the
redirected carrier. In principle, RRC Connected TS policies current study, measurements are performed in terms of
are the most effective load balancing mechanisms, since the Reference Signal Received Power (RSRP).
network is fully aware of the most important TS-related key By using the broadcast method, priority assignment is done
performance indicators such as cell load, user throughput, on a per-frequency basis and users are steered to higher
channel conditions, etc. Thus, the eNB can react faster to the priority layers whenever sufficient coverage is available. An
load fluctuations and perform either the necessary load-based example case is shown in Fig. 2, where the high frequency
handovers or adjust mobility parameters in favor of less layer (orange) is assigned higher priority than the lower
utilized neighboring cell. frequency one (purple). Basically, the ThreshXServingLow
In order to improve the autonomous UE cell reselection in threshold determines the coverage of the orange layer and
the RRC Idle state, and minimize the signaling overhead therefore AP can guarantee that cell center users will camp on
introduced by both redirection and handover events, network the high frequency layer regardless of the different pathloss
knowledge during the connection release phase could be properties of the two layers. However, since the network is not
utilized, as the RRC CONNECTION RELEASE message can fully aware of the UE distributions in the RRC Idle mode, it is
provide explicit RRC Idle mobility management parameters
on a UE basis [10]. Furthermore, redirection information can
also be included, requesting the terminal to reselect to a
High AP Low AP
different cell for load balancing purposes.
III. ABSOLUTE PRIORITIES & POTENTIAL ENHANCEMENTS
A. Absolute Priorities Framework
AP is a priority-based scheme for Inter-Frequency (IF)/
RAT Idle mode TS and mobility management. For the cell
reselection criteria, those are defined as follows:

- Criterion 1: A UE reselects to a higher priority cell N if: Figure 3: Example of multi-layer deployment
possible that a large amount of users is camping on the orange presented in Algorithm 1.The additional coverage check in
layer. Consequently, all these users will be served by the same terms of RSRP ensures that unnecessary IF measurements can
carrier when they switch to Connected mode, a fact that may be avoided in regions where a carrier frequency is not
result in overload. RRC Connected TS or TS at the connection available. The priorities of all available carriers are adjusted in
setup could resolve the inefficient resources utilization; a descending CAC order, assigning the highest one to the layer
however at the cost of excessive control plane signaling, as AP with the largest CAC value. The required IF measurements for
constantly pushes users to the high priority layer. coverage availability might already be present due to the
Finally, UE power consumption considerations must be also previous measurements that the UE was performing while
be taken into account. Generally, idle UEs go into a wake-up being in the RRC Connected state. If not, they could be
state at every Discontinuous Reception (DRX) cycle in order requested explicitly via a measurement trigger before
to listen to the paging channel and measure nearby cells. Intra- connection release. Finally, the updated frequency priority list
frequency measurements are triggered whenever the serving {pf} is provided to the UE via the connection release
signal power falls below the SIntraFreqSearch threshold: signaling.
3 IV. SIMULATION ASSUMPTIONS
whereas the IF measurement rate depends on the frequency The evaluation of the proposed scheme is performed on a
priorities. More specifically, the ThreshServingLow and LTE deployment consisting of three co-sited macrocell layers
SPrioritySearch thresholds control the IF UE measurements, for deployed at 800 MHz, 1800 MHz and 2600MHz, with an
lower and higher priorities respectively. Typically, the inter-site distance (ISD) of 1732m. The bandwidth allocations
SPrioritySearch is higher than the ThreshServingLow in order to are 10MHz, 10MHz and 20MHz respectively. Packet arrival is
steer terminals to the high-prioritized frequencies. However, modeled as a Poisson process, and UEs switch to idle
this can be extremely inefficient from a UE battery life point whenever downlink transmission buffers are emptied. The size
of view in areas where high priority carriers are not available per burst is negative exponentially distributed with a mean
(e.g. cell edge regions). value of 400 kbits. For all simulations, both the number of
B. SON-based Dedicated Priorities at the Connection Release UEs and the mean packet size are kept constant. Hence, the
offered load per macro sector area is adjusted by spanning
The aforementioned drawbacks of broadcast AP can be over different packet interarrival times. Terminals are moving
resolved if priorities are assigned on a UE resolution. In such a in straight line trajectories at 3 km/h.
manner, users at different locations can have different Since the priority adjustment is done at the connection
priorities per frequency. Since the RRC CONNECTION release, significantly long session interarrival times might
RELEASE message includes the option of updating the degrade the algorithm performance as the decision is based on
frequency priorities, expanding the framework on a dedicated
UE resolution is possible. Table II. SIMULATION ASSUMPTIONS
The proposed scheme dynamically adjusts the priorities of
the candidate frequencies based on both their load and Network Layout Hexagonal grid, 7 cell sites, 3 sectors per site
coverage conditions. For the load estimation, the concept of ISD 1732 m
Composite Available Capacity (CAC) [11] is utilized, Pathloss Models 3GPP
declaring the amount of resources that each cell cf (index f
denotes the frequency) is willing to offer for load balancing. A 800 MHz (10MHz) AP = 5 (Low)
Carrier Frequency
1800 MHz (10MHz) AP = 6 (Mid)
detailed description of the priority adjustment scheme is (Bandwidth)
2600 MHz (20MHz) AP = 7 (High)
43 dBm (800 MHz and 1800 MHz)
Algorithm 1 Absolute Priorities Adjustment Scheme Transmission Power
46 dBm (2600 MHz)
for f = 1 to number of carrier frequencies do Number of UEs 200 UEs per macro sector area @3km/h
Traffic Model Finite Buffer (mean burst size = 400kbits)
c argmax , , && , Offered Traffic {14, 18, 22, 26} Mbps per macro sector area
ThreshServingLow
-110 dBm {-110dBm, -106dBm}
end for {XHigh, XLow}
sort { } set in descending CAC order and derive the Connected: -110 dBm (A2 event [10])
IF Measurement
corresponding carrier frequency set {f} Idle: ThreshServingLow = -110 dBm
thresholds
SPrioritySearch = -60 dBm
adjust priorities{ }accordingly Æ assign the highest
Intra-Frequency Connected: UEs are always performing
priority to the first frequency/ element of set {f} Measeruement
and continue on a descending priority order Theshold Idle: SIntraFreqSearch = -106 dBm
if |{f}| < number of carrier frequencies then
Connected:5 msec every 40 msec
find the unavailable frequencies and move them to Measurement rate
Idle: 1.28 sec (DRX cycle)
the end of the set {pf}by assigning to them the
remaining lower priorities based on the relevant CAC Handover Offset 3 dB / 5 dB (Intra/ Inter-HO) (A3 event [10])
information TTT Window 0.4 sec (Intra-HO) / 0.5 sec (Inter-HO)
end if
Trelesection Timer 1.3 sec
return { }
L3 Filtering Factor 4
outdated load information. In order to simulate this effect and
provide a low algorithm bound, an extreme additional case is 2500
considered, where the priority adjustment is done randomly Broadcast
Random
without any load/ coverage knowledge. Note that except for SON-based
2000

Avg. UE Throughput [Kbps]

the RRC Idle TS mechanisms, no other TS scheme is applied.
In order to define CAC, traffic is treated in an elastic
manner by defining a minimum bit rate requirement minBR. If 1500
Ru is the achievable instantaneous throughput of user u, and fu ~10 %
the corresponding resource share, then the actual amount of
resources required to meet minBR is given as follows: 1000
~65 %
·
(4) 500

By assuming that the remaining resources can be utilized

0
for load balancing, the CAC of cell c is modeled according to 14 18 22 26
Offered Traffic [Mbps / Macro sector area]
(5), where NPRB,c is the cell bandwith in Physical Resource
Blocks (PRB), and NPRB,min the lowest available bandwidth in Figure 4: Avg. UE Throughput versus Offered Load
the system . Fig. 3 illustrates the user distribution as a function of load
∑ for the different simulated schemes. For the broadcast case,
,
,0 100% 5 the user distribution is insensitive to load, as it is primarily
, , affected by the absolute RSRP thresholds. Note that the
1800MHz band is left unutilized due to its lower priority
The bandwith scaling factor emulates the 3GPP-defined
compared to the 2600MHz carrier. Therefore, even if (1) is
Cell Capacity Class Value (CCCV) [11]. MinBR is set to 1024
satisfied for both carriers, all cell center UEs reselect to the
Kbps whereas the target operational load, ρtarget, is set to 0.8.
2600MHz layer. On the other hand, the random dedicated
V. RESULTS scheme tends to overload the 800MHz layer as the offered
traffic increases. This behavior is explained by the fact that
The performance of the different schemes is evaluated in except for cell edge users (they will reselect to the 800MHz
terms of the user distribution over the different carriers (both layer regardless of the priority assignment simply due to radio
RRC Idle & Connected, averaged over the whole simulation conditions), cell center users start camping on the 800MHz
time), average user throughput and mobility events rate for band due to the random assignment. Hence, session
different load conditions. Finally, the percentage of DRX completion times become significantly longer and gradually
cycles performing IF measurements in the RRC Idle state is the RRC Idle state vanishes, “trapping” all users in the RRC
recorded as well. Connected.
SON-based dedicated priorities offload significantly the
Broadcast 2600MHz carrier primarily towards the 1800MHz one as the
2600 MHz
1
1800 MHz
offered traffic increases due to the load-aware nature of the
0.75 800 MHz priority adjustment algorithm, and cell center are now better
0.5 distributed between the two higher frequency bands. An
0.25 important observation is the fact that the user distribution
0 tends to converge to the optimal 0.5/0.25/0.25
U E D is tr ib u tio n (Id le + C o n n e c te d )

14 18 22 26
ratio since the 2600MHz layer has double bandwidth.
1
Random Dedicated However, this is not achieved due to the slower adaptation
0.75
of the RRC Idle mode to the load fluctuations, as users have to
0.5
switch to idle in order TS to be applied.
0.25
The avg. UE throughput versus the offered traffic is shown
0
in Fig. 4. The random scheme shows the worst performance
14 18 22 26 since users are connecting to the 800MHz carrier ignoring the
SON-based Dedicated double bandwidth of the 2600MHz band. The better load
1 distribution across the network layers is further proven by the
0.75 capacity gains that the SON-based scheme provides over the
0.5 broadcast AP case. In principle, SON-based dedicated
0.25 priorities improve the network capacity by ~10%, whereas in
0
14 18 22 26 terms of avg. UE throughput, a ~65% gain is observed for the
Offered Traffic [Mbps / Macro sector area] case of 26 Mbps offered load. It is rather clear that SON-based
priorities would minimize the signaling that additional TS
Figure 3: UE Distribution versus Offered Load mechanisms (i.e. load-based handovers/ redirections) would
 0.04
RRC Connected Handover Rate measurements occur and UE battery life savings can be
Intra-Freq achieved. The noticeably higher IF measurement levels for
0.03 Inter-Freq
Load Increase the broadcast case are primarily caused by cell edge users
M ob ility Event R ate [Even ts /use r/s ec ]

0.02 camping on the 800MHz frequency and perform IF

measurements at every DRX cycle. Note that for the random
0.01
dedicated scheme and offered load conditions above 14 Mbps,
0 the low IF measurement rates are due to the lack of RRC Idle
Broadcast Random Dedicated SON-based Dedicated
(a) samples.
RRC Idle Reselection Rate
0.04
Intra-Freq VI. CONCLUSIONS
Load Increase
0.03 Inter-Freq In this paper, a novel SON-based UE-dedicated priority
0.02 adjustment scheme has been proposed for enhancing the AP
framework by utilizing the RRC connection release signaling.
0.01
Compared to the conventional AP broadcasting, the developed
0 mechanism achieves better load balancing, improving user
Broadcast Random Dedicated SON-based Dedicated
(b) experience and network capacity. However, long UE inactivity
periods can degrade the performance, since the TS decision is
Figure 5: (a) Handover and (b) Reselection Rate versus Offered
Load taken at the connection release. For that purpose, additional
features such as dedicated priority validity-timers and UE
have generated if they were applied on top of them, as the
behavior prediction based on call history could additionally be
RRC Idle cell resection procedures have been improved.
applied. Note that under no circumstances, sole idle mode TS
Consequently, less RRC Idle wrong decision would have had
could outperform TS at the connection setup or in the RRC
to be corrected at either the connection establishment phase or
Connected due to its slower adaptation to the load variations.
RRC Connected mode.
However, an optimized idle mode policy can provide better
The benefits of proper alignment between the RRC Idle and
alignment to the TS/ mobility management procedures that
Connected are depicted in Fig. 5, where both handover (Fig.
take place in the different RRC states, resulting in significant
5a) and reselection rates (Fig. 5b) are presented. Using the
signaling decrease. Future work includes the algorithm
SON-based priorities, the signaling overhead generated by the
evalution in a Heterogeneous Network (HetNet) along with its
mobility events is decreased significantly due to the fact that
interaction with additional TS policies at different RRC states.
idle-to-connected ping-pongs (and vice versa) are minimized.
Hence, the observed gains derive from the IF handover/ VII. REFERENCES
reselection rate reduction. Note that for the random scheme, IF
[1] “Cisco Visual Networking Index: Global Mobile Data Traffic Forecast
events are almost eliminated as all users are connected to the Update, 2011-2016”, Cisco Public Information, February 2012.
800MHz carrier, thus only intra-frequency ones are performed [2] 3GPP TR 36.902, “Self-configuring and self-optimizing network cases
at the cell borders. Finally, the severe overload that the and solutions (Release 9)”, v9.3.1, March 2011.
800MHz layer experiences at higher offered traffic conditions [3] S. Hamalainen, H. Sanneck, and C. Sartori, “LTE Self-Organizing
is further proven by the elimination of reselections, as all users Networks (SON): Network Management Automation for Operational
are “trapped” in the RRC Connected state. Efficiency”, Wiley 2012.
Finally, the percentage of DRX cycles performing IF [4] A. Lobinger, S. Stefanski, T. Jansen, and I. Balan, “Load Balancing in
Downlink LTE Self-Optimizing Networks”,IEEE Vehicular Technology
measurements in idle mode is illustrated in Fig. 6. The SON- Conference, May 2010.
based scheme outperforms the broadcast one, since UEs [5] Peng Tian, Hui Tian, Jianchi Zhu, Lan Chen and Xiaoming She, “An
almost always camp on the higher priority carrier; hence no IF Adaptive Bias Configuration Strategy For Range Extension in LTE-
Advanced Heterogeneous Networks ”, IEEE International Conference
on Communication Technology and Application, October 2011.
25
%DRX Cycles in RRC Idle with IF measurements [%]

14Mbps [6] Niels T.K. Jørgensen, Daniela Laselva and Jeroen Wigard,“On the
18Mbps Potentials of Traffic Steering Techniques between HSDPA and LTE”,
22Mbps
20 26Mbps IEEE Vehicular Technology Conference, May 2010.
[7] “Standby Time Analysis for Mobile Devices in Multi-carrier WCDMA
Networks”, Qualcomm Incorporated, May 2010.
15
[8] Tomasz Mach, Rahim Tafazolli,“Battery Life Idle Parameter
Optimization of UE in Self Organizing Network”, IEEE International
Symposium on Personal Indoor and Mobile Radio Communications,
10
September 2011.
[9] 3GPP TS 36.304, “Evolved Universal Terrestrial Radio Access (E-
5 UTRA); User Equipment (UE) procedures in idle mode”, v11.2.0,
December 2012.
[10] 3GPP TS 36.331, “Evolved Universal Terrestrial Radio Access (E-
0
Broadcast Random Dedicated SON-based Dedicated UTRA); Radio Resource Control (RRC); protocol specification”,
v10.0.0, December 2010.
Figure 6: Percentage of DRX cycles performing IF [11] 3GPP TS 36.423, “Evolved Universal Terrestrial Radio Access (E-
measurements in RRC Idle UTRA); X2 Application Protocol”, v10.5.0, March 2012.