Nokia Solutions and Networks

Smart Scheduler
NSN White paper
February 2014
nsn.com
Page 2
CONTENTS
1. Introduction 3
2. Smart Scheduler Features and Benefts 4
3. Smart Scheduler wit Explicit Multi-Cell
Coordination
10
3.1 Distributed RAN with X2+ and
non-ideal backhaul
10
3.2 Distributed RAN with slow centralized
scheduling and non-ideal backhaul
11
3.3 Centralized RAN (C-RAN) 11
3.4 Enhanced Inter-Cell Interference
Control (eICIC) with co-channel
small cells
12
4. Further Evolution of LTE Scheduling 14
5. Summary 15
6. Abbreviations 15
nsn.com
Page 3
1. Introduction
As of January 2014, Long Term Evolution (LTE) has been successfully
deployed by more than 250 operators, with more than 200 million
customers enjoying high mobile broadband data rates. LTE in FDD and
TDD mode (TD-LTE) is designed for a so-called frequency reuse of one
where all the cells use the same frequency. Reuse of one provides the
highest network efciency and enables high data rates close to the
base station.
The challenge with reuse of one is the high inter-cell interference
when the terminal (User Equipment UE) is located between two cells.
The data rate over the cell area is illustrated in Figure 1. Boosting the
cell edge performance is the main motivation of Smart Scheduler.
Smart Scheduler can also enhance the average data rates and system
capacity by considering signal fading and interference in packet
scheduling decisions. Smart Scheduler algorithms, benefts, impact on
the network architecture and further evolution are discussed in this
white paper. If not otherwise explicitly stated, all statements are valid
for both LTE (in FDD mode) as well as for TD-LTE.
Frequency f1
Cell A Cell B UE
High data rate
close to BTS
Data
rate
Frequency f1
Low data rate
at cell edge
Fig. 1. Frequency reuse of one creates high inter-cell interference
nsn.com
Page 4
2. Smart Scheduler features
and benefts
LTE radio technology is highly standardized by 3GPP but only with
regard to the interfaces the network algorithms including link
adaptation, power control and packet scheduling are not standardized.
Therefore, there can be diferences in network performance due to
the diferent algorithms being used by diferent vendors. The most
relevant features and benefts are described in this section. Packet
scheduling can use diferent input information for resource allocation
and for interference coordination:
Channel Quality Information (CQI) from UE to BTS for downlink
scheduling.
Sounding Reference Signal (SRS) measurements and interference
measurements in the frequency domain for uplink scheduling.
Load and other information exchange over the X2 interface
between base stations. X2 interface in Release 8 allows some
exchange of information between the base stations, but further
extensions will be discussed in 3GPP and can also be added
proprietarily.
Quality of Service (QoS) parameters from the packet core network
These diferent input information options are illustrated in Figure 2.
Cell A Cell B UE
Gateway
QoS QoS
Channel quality
information (CQI)
Coordination over X2+
Fig. 2. Input information for coordinating the resource usage
nsn.com
Page 5
Smart Scheduler can utilize the diferent input values to optimize
packet scheduling and link adaptation. LTE allows considerable
freedom to defne allocations in the time, frequency and power
domains. A number of diferent features are required for the diferent
use cases. The same features are utilized both in Frequency Division
Duplex (FDD) and Time Division Duplex (TDD) based LTE. The Smart
Scheduler utilizes the following main features:
Frequency Selective Scheduling (FSS) improves performance in
the case of frequency selective fading and fractional inter-cell
interference. FSS consists of Channel Aware Scheduling (CAS) and
Interference Aware Scheduling (IAS). The feld measurements show
+30% gains for the cell edge data rates.
Interference shaping further improves the efciency of the inter-
cell interference avoidance by FSS. When the cell loading is low,
the number and set of physical resource blocks is adapted only
slowly according to trafc fuctuations. This approach makes it
more efcient for the adjacent highly loaded cells to avoid inter-
cell interference based on UE CQI reporting. Studies show gains
exceeding 100%.
QoS diferentiation improves cell edge performance by allocating
more resources for users in weak channel conditions. QoS can be
utilized to maintain the data rate, for example for video streaming
services. Further fexibility is obtained by using operator specifc
QoS Class Identifer (QCI) values. The minimum guaranteed cell
edge data rate can be obtained also by Nominal Bit Rate (NBR)
which works even without guaranteed bit rate QoS classes. Cell
edge prioritization has only a minor impact on the cell aggregate
throughput capacity, in typical case 30% cell edge throughput
improvement can be obtained at the cost of 5% cell throughput
capacity. The capacity measured in number of satisfed subscribers
is still higher.
Interference aware uplink power control considers the adjacent
cells when allocating the uplink transmission power. The feature
minimizes inter-cell interference and helps to boost uplink
data rates.
Intra-frequency load balancing helps when the load in the adjacent
cells is not balanced. The idea is to modify handover parameters
based on the information exchange of the X2 interface. If there
are double the users in the adjacent cell, the intra-frequency load
balancing can improve the cell edge data rate by 30%.
nsn.com
Page 6
Multi-cell scheduling can reduce the power levels (muting or related
variants) in adjacent cells to minimize the interference. The multi-
cell scheduling coordinates resource allocation between multiple
cells in time and in frequency, using a selection of users and power
levels in multiple cells to combine the benefts of frequency-
selective scheduling and spectral efciency gain due to reduced
interference. The coordination happens between the sectors
of one base station, or over the X2 interface between the base
stations. Multi-cell scheduling can improve cell edge performance
by 20%. Multi-cell scheduling requires inter base station time
synchronization. TD-LTE base stations need to be synchronized
while the synchronization of LTE FDD base stations is not mandatory
and is typically not used by operators for FDD deployments. Note
also that the reference signals are overlapping in adjacent cells in
a synchronized network. Therefore, UEs should preferably support
cancellation of common reference signals for better performance.
The Smart Scheduler use cases, features and gains are shown in Figure
3 and Figure 4. Figure 4 shows the gains of the individual scheduling
functionalities when used jointly. More gain can be obtained in HetNet
scenarios with eICIC.
100%
Cell edge Average
Intra-frequency load balancing
Multi-cell scheduling
Nominal bit rate and QoS
Frequency selective scheduling
20%
40%
60%
80%
120%
0%
Fig. 4. Smart Scheduler downlink
data rate gains with non-ideal
backhaul
Fig. 3. Smart Scheduler use cases
and solutions
Use case Feature
Fractional inter-cell interference
Lower
gain
Highest
gain
Unbalanced loading between cells
HetNet
Minimum cell edge rate required
Fractional inter-cell interference
Frequency selective fading
Multi-cell scheduling
Intra- and inter-frequency load balancing
eICIC
QoS differentation and nominal bit rate
Baseline scheduler
FSS including Interference Aware
Scheduling (IAS) and Channel Aware
Scheduling (CAS)
nsn.com
Page 7
Fig. 5. Frequency Selective Scheduling to minimize fading impact
Lets now analyze Frequency Selective Scheduling (FSS) the most
important part of the Smart Scheduler. The multipath propagation in
the mobile environment makes the fading frequency selective. The
typical coherence bandwidth of the macro cell channel is 1-2 MHz,
therefore, there are faded and non-faded frequencies within one LTE
carrier. LTE radio uses Orthogonal Frequency Division Multiple Access
(OFDMA) in the downlink and Single Carrier Frequency Division Multiple
Access (SC-FDMA) in the uplink. Therefore, FSS allows use of those
parts of the carrier (called Physical Resource Blocks) not faded for the
transmission. The concept is illustrated in Figure 5. Information about
channel fading can be obtained from UE CQI reports in downlink and
from Sounding Reference Symbols (SRS) in uplink.
Transmit on those resource blocks that are not faded
Carrier bandwith
Resource block
Frequency
nsn.com
Page 8
Interfering cell Target cell
F
r
e
q
u
e
n
c
y
CQI 1 (low)
CQI 2 (high)
CQI 3 (high)
CQI 4 (low)
CQI 5 (low)
CQI 6 (high)
CQI 7 (high)
CQI 8 (high)
Fractional load
in adjacent cell
UE reports
subband CQI
Frequency selective
scheduling
UE A
No transmission
Transmission in adjacent cell
Transmission to UE A Transmission to other UEs
Fig. 6. Frequency Selective Scheduling (FSS) to minimize
inter-cell interference
FSS can also be applied to avoid inter-cell interference. An example
is shown in Figure 6 where the interfering cell is partially loaded. The
UE is connected to the target cell but receives strong interference
from the adjacent interfering cell. The UE reports sub-banded CQI
values in the frequency domain to the target cell. Low CQI values are
reported on those sub-bands where the interfering cell has on-going
transmission while high CQI values are reported in other sub-bands.
The target cell with FSS tends to allocate those downlink physical
resources blocks to the UE where the interference is lowest. The
other resource blocks in the target cell can be allocated to other UEs
that do not receive interference from the adjacent cell. Benefts of
FSS include:
Efective inter-cell interference coordination without the need for
explicit inter-BTS coordination
Utilization of UE CQI reports for interference mitigation and without
the need for coordination signaling between the base stations
Improved cell edge data rates as well as total cell capacity.
nsn.com
Page 9
Fig. 7. Field measurements with
FSS in downlink
Cell edge throughput
M
b
p
s
M
b
p
s
Cell capacity
FSS o FSS on
2.5
3.0
0.5
1.0
1.5
2.0
3.5
0.0
25
30
FSS o FSS on
5
10
15
20
35
0
As part of the Smart Scheduler concept, the underlying link adaptation
function is critical for the success of features such as FSS. The quality
of reporting from each active terminal is always monitored and
compensation is constantly conducted in order to improve the value
of the scheduler decisions. With such methods, NSN has in numerous
commercial LTE networks shown the practical value of FSS. An example
feld measurement result with 10 MHz bandwidth is shown in Figure 7.
Fig. 8. Interference
shaping for more efcient
interference avoidance
Interference shaping further improves the efciency of the inter-cell
interference avoidance by FSS. Interference shaping is illustrated in
Figure 8. When the cell loading is low, the number and set of physical
resource blocks is adapted only slowly according to trafc fuctuations.
This approach makes it more efcient for the adjacent high loaded
cells to avoid inter-cell interference based on UE CQI reporting. The
studies show major gains for the cell edge data rates in those cases
where the loading is unbalanced between the cells: the gains can
exceed 100%. High gains can be achieved in the distributed solution
with cleverer scheduling without any fast signaling over the X2
connection and without any centralized network element.




Low loaded cell: Only slow changes in
frequency domains
High loaded cell: Robust inter-cell
interference avoidance based on CQI reports
nsn.com
Page 10
3. Smart Scheduler with explicit
multi-cell coordination
Further performance improvements can be obtained by coordinating
resource allocation in adjacent base stations. The network architecture
options for supporting multi-cell scheduling are shown in Figure 9.
a) Distributed RAN with X2+ and
non-ideal backhaul
X2+
eNB#1
Coordinated
scheduling
(inter-eNB)
Fast local
scheduling
eNB#1
Coordinated
scheduling
(inter-eNB)
Fast local
scheduling
eNB#1
Fast local
scheduling
X3
eNB#N
Fast local
scheduling
Coordinated scheduling
...
b) Distributed RAN with slow
centralized scheduling and
non-ideal backhaul
Super-eNB (baseband pool)
Common packet scheduling
Direct ber with multi-Gbps
...
c) Centralized RAN with fast
centralized scheduling
and dark ber connection
Fig. 9. Network architecture options for explicit multi-cell scheduling
3.1 Distributed RAN with X2 and
non-ideal backhaul
Todays LTE architecture (99% of deployments) is shown in Figure 9a
using non-ideal backhaul with microwave radio, IP connected fber or
copper based transport. The multi-cell scheduling needs to coordinate
the resource usage in adjacent base stations over non-ideal backhaul
while still fully utilizing FSS gains in fast scheduling. The coordination
between cells of diferent base stations will utilize the X2 interface.
Each scheduler that requests coordination from its neighboring
base stations to aid a user at the cell edge can still take into account
FSS gains for that user, thus FSS gains can be fully preserved while
adding the gains from multi-cell coordination. The evolution from
fully distributed architecture to multi-cell coordination over X2 is
a straightforward software upgrade no new network elements
or interfaces are needed. Note that fast local coordination can be
implemented between the cells in one base station without any inter-
base station coordination.
nsn.com
Page 11
3.2 Distributed RAN with slow centralized
scheduling and non-ideal backhaul
Another architecture alternative is shown in Figure 9b with a new
centralized network element for coordinating the distributed schedulers.
A new interface between base stations and the centralized scheduler is
required. Involving an additional interface and information exchange to
an additional entity has a negative impact on the responsiveness of this
architecture. The distributed base stations still run the fast scheduling
while the centralized element can only set scheduling limitations to
minimize the interference. The performance gain of the centralized
element is similar to the coordination over the X2 interface.
Coordinated multi-cell scheduling and muting over non-ideal backhaul
was studied in 3GPP under the title Enhanced Coordinated Multipoint
(eCoMP) during 2013. The conclusion taken in December 2013 was that
the gains for inter-site macro-macro scenario are below 5% in the best
case over intra-site and less if the backhaul latency increases (several
tens of ms). The further focus of eCoMP will be in the HetNet scenarios
between macro cells and small cells.
3.3 Centralized RAN (C-RAN)
The fnal multi-cell architecture shown in Figure 9c is centralized
scheduling in the baseband pool. This is the architecture for a network
with ideal transport. The baseband pool requires a low latency direct
dark fber connection between the RF heads and the baseband pool.
The baseband pool is also referred to as Centralized Radio Access
Network (C-RAN).
C-RAN is like a super-sized base station. C-RAN enables the most
advanced multi-cell coordination because all the functionalities are in
the same location: link adaptation, power control, fast FSS and multi-cell
coordination. C-RAN architecture also enables Joint Transmission and
Joint Reception Coordinated Multipoint (CoMP) between diferent sites
while intra-site CoMP can be implemented also in the distributed
RAN architecture.
CoMP functionality is defned in 3GPP Release 11 but uplink CoMP can
be implemented also with legacy Release 8 UEs while the downlink CoMP
requires Release 11 UEs. Uplink CoMP gives more gain, while downlink
CoMP gains are limited. An excellent use case for C-RAN is to boost
capacity in stadiums and other mass event locations. These events
tend to be uplink limited because many people want to send pictures
from the event. The UE transmission is received by a single cell in the
traditional solution while the same UE transmission can be received
by multiple cells and combined in the baseband module. The inter-cell
interference turns into a constructive signal. The solution is illustrated in
Figure 10. The installation of fber between baseband modules and RF is
relatively simple in these event areas.
nsn.com
Page 12
NSN Flexi Multiradio base station enables CoMP by providing fast
interconnections between the baseband modules. NSN C-RAN has
been validated in commercial networks in large stadiums and the
practical gains exceed 100%.
3.4 Enhanced Inter-Cell Interference Control
(eICIC) with co-channel small cells
Small cells are an attractive solution for boosting hot spot capacity
and coverage. The interference management needs to be considered
when the small cells are deployed on the same frequency as the
macro cells.
3GPP Release 10 brings a solution for managing the interference
in the time domain. The solution is called enhanced Inter-Cell
Interference Coordination (eICIC) and is shown in Figure 11. The macro
cell leaves some empty sub-frames called Almost Blank Subframes
(ABS). During these sub-frames, the small cell can serve UEs that
would otherwise receive too much co-channel interference from the
macro cell.
Fig. 10. Centralized RAN for boosting mass event capacity
nsn.com
Page 13
The beneft of eICIC comes when several small cells can beneft from
macro cell empty subframes. eICIC performance is further boosted in
Release 11 by using UE interference cancellation for the minimization
of inter-cell interference, which is known as further enhanced ICIC
(feICIC). Optimized eICIC requires that the number of ABS frames and
the handover parameters are adjusted dynamically according to the
instantaneous trafc conditions and UE locations. The semi-static
solution is a slow approach for modifying the feICIC parameters over
several seconds. The fast feICIC adaptation uses quick adaptation for
the number of ABS sub-frames to reallocate resources between macro
cells and small cells depending on the instantaneous requirements.
NSNs unique algorithm is based on the fast adaptation of ABS-
blanking and cell range extension for maximum beneft from small
cell deployments. The throughput gains are shown in Figure
12: dynamic eICIC can nearly double the user throughputs in
heterogeneous networks.
Fig. 11. Time domain interference management with enhanced ICIC
Fig. 12. Throughput gain from enhanced ICIC




= Sub-frame with normal transmission
= Almost blank sub-frame (ABS)
Pico cell can reuse same frequency
as macro when UE is closer to pico
Pico cell can serve also such UEs that
receive stronger macro cell signal
Sub-frame (1
ms)
Macro
Pico
Baseline
w/o felCIC:
0%
Semi-static
felCIC
70%
Fast
felCIC
90%
nsn.com
Page 14
4. Further evolution of LTE scheduling
3GPP is working with Inter-site carrier aggregation in Release 12.
The feature allows the UE to receive data simultaneously from the
macro cell and from the small cell. The two cells do not need any fber
backhaul, although wireless backhaul with some delay is fne. The X2
interface is used between the macro cell and small cell for scheduler
coordination. The macro cell and the small cell can share the same
frequency or the small cell can use a dedicated frequency. The feature
is illustrated in Figure 11.
Inter-site carrier aggregation uses Dual Connectivity where the UE has
simultaneous radio connection to both macro and to small cell. That
brings benefts in terms of reliable mobility.
3GPP is also working on a solution where UEs can cancel the inter-cell
interference by obtaining assistance information from the network.
This feature is called Network Assisted Interference Cancellation and
Suppression (NAICS) and it is part of Release 12. If UEs can cancel
interference, it may be more efcient to use all resources in co-
channel cells instead of muting resources. The multi-cell scheduling
and muting algorithms need to be designed in such a fexible way that
they can beneft from the future advanced UE capabilities.
Fig. 13. Inter-site carrier aggregation in Release 12
Inter-site Carrier Aggregation
and Dual Connectivity
nsn.com
Page 15
5. Summary
While LTE has been highly standardized by 3GPP, the network
algorithms including packet scheduling are not standardized. The
packet scheduling in LTE has the freedom to control the resource
allocation in the time and in the frequency domain.
Smart Scheduler can push cell edge data rates by more than 100% in
the presence of inter-cell interference compared to baseline wideband
scheduling, and improve the cell capacity by more than +20%. The
main component of Smart Scheduler is frequency selective scheduling
that avoids the fading and interference in the frequency domain
combined with Quality of Service diferentiation and intra-frequency
load balancing. NSNs innovation Interference Shaping increases the
cell edge throughput further by up to 100% when the cell loading is
unbalanced.
Additional cell edge gains can be obtained by multi-cell scheduling.
Multi-cell scheduling is a simple software upgrade to distributed base
stations. Scheduling information is shared between base stations
over the X2 interface. The detailed standardization of multi-cell
coordination is considered in 3GPP Release 12.
The most advanced multi-cell coordination can be obtained with
baseband pooling in Centralized RAN. The baseband pool deployment
assumes direct fber connection between baseband and RF sites.
Centralized RAN provides the biggest benefts in uplink capacity, which
is most useful in high capacity events. The efciency of small cell
deployment can be boosted by using dynamic eICIC confguration to
manage the interference between macro cells and small cells.
6. Abbreviations
3GPP Third Generation Partnership Project
BTS Base Station
CoMP Coordinated Multipoint
CQI Channel Quality Information
C-RAN Centralized Radio Access Network
eCoMP Enhanced CoMP
eICIC Enhanced Inter-Cell Interference Coordination
FDD Frequency Division Duplex
FSS Frequency Selective Scheduling
LTE Long Term Evolution
OFDMA Orthogonal Frequency Division M
QoS Quality of Service
RRH Remote Radio Head
SC-FDMA Single Carrier Frequency Division Multiple Access
SRS Sounding Reference Symbols
UE User Equipment
Nokia Solutions and Networks
P.O. Box 1
FI-02022
Finland
Visiting address:
Karaportti 3, ESPOO, Finland
Switchboard +358 71 400 4000
Product code C401-00961-WP-201402-1-EN
2014 Nokia Solutions and Networks. All rights reserved.
Public
NSN is a trademark of Nokia Solutions and Networks. Nokia is a registered
trademark of Nokia Corporation. Other product names mentioned in this
document may be trademarks of their respective owners, and they are
mentioned for identifcation purposes only.
nsn.com