You are on page 1of 25

Author:

HARP Consortium
Document number D6.1-001



Theme ICT

FP7-ICT-2011-8 Future Networks


Funding scheme

Collaborative project
Project Title

High capacity network Architecture with Remote radio heads & Parasitic
antenna arrays
Acronym

HARP
Project No
318489

DELIVERABLE D6.1
Aggregation Network Definition
Work package 6
Leading partner: AIT
Dissemination level: PU

Delivery date: 30-04-2014



DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
1 of 25

Author: HARP Consortium


Document number D6.1-001

Contributors

Partner

Contributing authors

UEDIN
AIT
EUR
IMPERIAL
NTNU
DTU
RADIOCOMP
ALBLF
FT


K. Georgakilas



H. Christiansen
A. Checko
L. Roullet












DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
2 of 25

Author: HARP Consortium


Document number D6.1-001

Table of Contents
1.0 Introduction ........................................................................................................... 7
2.0 Overall Aggregation Network Structure .................................................................. 8
2.1 C-RANs with packet-based front haul ..................................................................... 10
2.2 Other options (OTN) ................................................................................................ 11
3.0 Aggregation Network Architecture Characteristics ............................................... 13
3.1 Benefits ................................................................................................................... 13
3.1.1 Efficient use of transport resources ................................................................. 13
3.1.2 Optimal multiplexing gain ................................................................................. 14
3.2 Challenges ............................................................................................................... 17
3.2.1 Synchronization IEEE 1588 and CPRI .............................................................. 17
3.2.2 Controlling special antennas ............................................................................. 19
3.2.3 Other Challenges .............................................................................................. 20
4.0 Network Planning Aspects ................................................................................... 20
4.1 Aggregation Network Optimization Problems Definition ....................................... 21
4.1.1 Network Dimensioning ..................................................................................... 22
4.1.2 BBU Pool Placement ......................................................................................... 22
4.1.3 Assignment of Cells to BBUs ............................................................................. 23
DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
3 of 25

Author: HARP Consortium


Document number D6.1-001




4.2 Network Optimization Methodology ...................................................................... 23
5.0 Summary ............................................................................................................. 24
6.0 References ........................................................................................................... 24















DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
4 of 25

Author: HARP Consortium


Document number D6.1-001

List of Figures
Figure 1: A traditional aggregation network, which aggregates traffic from a number of
access networks. In this case a C-RAN based access network is shown. ............................ 8
Figure 2: The HARP network, complete with beamforming antennas, RRH aggregation
network, CRAN and the traditional backhaul. .................................................................... 9
Figure 3: C-RAN with Ethernet based fronthaul ............................................................... 11
Figure 4: C-RAN with OTN based fronthaul ...................................................................... 12
Figure 5: Two fundamentally different ways of connecting the RRHs to the BBU pool. a)
direct connection (basically circuit switching) and b) packet switching. ......................... 14
Figure 6: Daily load on base stations varies depending on base station location [1] ....... 15
Figure 7: Statistical multiplexing gain in C-RAN architecture for mobile networks .......... 16
Figure 8: CPRI over Ethernet modeling setup ................................................................... 18
Figure 9: Communication between RRHs and the BBU-pool. a) the normal C-RAN case b)
Adding special (e.g., beamforming) antennas to the RRHs. c) Packet-based fronthaul. .. 20





DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
5 of 25

Author: HARP Consortium


Document number D6.1-001

Acronyms
BBU
CPRI
C-RAN
GPS
ILP
OTN
PTP
QoS
RRH
SAA
WDM
LTE
LTE-A

BaseBand Unit
Common Public Radio Interface
Cloud Radio Access Network
Global Positioning System
Integer Linear Program
Optical Transport Network
Precision Time Protocol
Quality of Service
Remote Radio Head
Sample Average Approximation
Wavelength Division Multiplexing
Long Term Evolution
Long Term Evolution Advanced









DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
6 of 25

Author: HARP Consortium


Document number D6.1-001

1.0 Introduction
This deliverable deals with the definition of the aggregation network of the HARP
project. The goal is to establish the terminology and explain the focus of the project at
this stage regarding the aggregation network. Moreover, it aims to give a clear picture
of the aggregation network architecture and to act as a basis for the forthcoming
deliverable D6.2 that deals with the optimization of the network.
A significant point in the projects duration has been reached where the partners have
agreed to focus mostly on the fronthaul part of the network, that is the interconnection
between BBUs and RRHs, the RRH aggregation network. The classical backhaul
aggregation network remains under the scope of the project but not as the main focus.
The overall aggregation network architecture based on a packet-based solution is
presented, along with the respective benefits and challenges. Methods on how these
will be demonstrated and optimized in future deliverables are also provided. In the
context of optimization, the document finally illustrates the projects vision regarding
network planning through the most important problems that are going to be considered
for the RRH aggregation network.







DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
7 of 25

Author: HARP Consortium


Document number D6.1-001

2.0 Overall Aggregation Network Structure


The aggregation network is normally considered to aggregate traffic from a number of
access networks as depicted in Figure 1. The figure shows two C-RAN based LTE access
networks. It shows how in the fronthaul fibers are used to interconnect the Baseband
pool or baseband Unit (BBU) pool to the Remote radio heads (RRHs). Direct fibers are
normally used as high bitrates are required to transport the IQ samples between the
BBU pools and the RRHs. Moreover, the figure also depicts the beamforming antennas
attached to the RRHs in the HARP case.
Legend

RF

Direct fiber
connection
Logical
connection

RF

Access
network
BBU
Base
Base
Base
Base
pool
band
band
band
band

RF

S1
R
F

X2

RF

Aggregation
network

RF

EPC
MME

S1
BBU
Base
Base
Base
Base
pool
band
band
band
band

RF
R
F

S5
SGW

Fronthaul

Application server

Internet

S11
PGW

Application server

Backhaul

Figure 1: A traditional aggregation network, which aggregates traffic from a number of access networks.
In this case a C-RAN based access network is shown.

The choice of technology as well as dimensioning for the aggregation network depicted
in Figure 1 might not be straightforward but is a standard planning task for a mobile
operator. It is not different from planning a network without C-RAN but with an
identical number of cells and will not be considered any further in this deliverable.

DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
8 of 25

Author: HARP Consortium


Document number D6.1-001




However, when we refer to the aggregation network in this deliverable we mean the
RRH aggregation network, i.e., aggregation of traffic from multiple RRHs (or cells) to the
BBU pool. This is shown in Figure 2, i.e., the traditional aggregation network is part of
the backhaul, whereas the RRH aggregation network that we consider here is part of the
fronthaul.
Legend

RF

Logical
connection

RF

Access
network
RF

RRH Aggregation
network

BBU
Base
Base
Base
Base
pool
band
band
band
band

S1
R
F

X2

RF

Aggregation
network

RF

RF

RRH Aggregation
network
R
F

EPC
MME

S1
BBU
Base
Base
Base
Base
pool
band
band
band
band

S5
SGW

Fronthaul

Application server

Internet

S11
PGW

Application server

Backhaul

Figure 2: The HARP network, complete with beamforming antennas, RRH aggregation network, CRAN
and the traditional backhaul.

The benefit of introducing an RRH aggregation network is that we can eliminate a


number of drawbacks of the traditional C-RAN architecture using direct fiber
connections, namely:

Inflexible use of fronthaul transport resources.


In the traditional C-RAN architecture, point-to-point high-capacity fibers are
deployed to all RRHs. Obviously, cells are not loaded equally around the clock,
but due to the fixed allocation of fibers, capacity is always allocated and cannot
be shared among cells or with other access networks in the same area.

DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
9 of 25

Author: HARP Consortium


Document number D6.1-001

Fixed BBU cell assignment


Another consequence of the direct fiber connections is that a given cell cannot
be moved from one BBU pool to another. This is problematic as the multiplexing
gain of a BBU pool (which is one of the C-RAN advantages) depends on the traffic
characteristics on the cells sharing a pool. Hence, if the traffic characteristic of a
cell changes, the multiplexing gain is potentially reduced.

The purpose of the RRH aggregation network is to get rid of these drawbacks. Section 3
studies the pros and cons of these alternative approaches in more detail.
Now, with the RRH aggregation network we have basically two technology options (in
addition to the direct fiber connections):
1. An Ethernet based RRH aggregation network, which is the most interesting and
2. An OTN based aggregation network, which is mainly treated here for
completeness
2.1 C-RANs with packet-based front haul
To build a packet-based fronthaul the most obvious choice of technology is Ethernet. It
is relatively cheap and equipment is readily available. An example of an Ethernet based
fronthaul is shown in Figure 3. The figure shows how RRHs serving the individual cells
are connected to in this case two BBU pools.

DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
10 of 25

Author: HARP Consortium


Document number D6.1-001

Ethernet

BBU pools
RRH aggregation network

Fronthaul with beamforming antennas and packet based RRH aggregation network


Figure 3: C-RAN with Ethernet based fronthaul

2.2 Other options (OTN)


Another option for a fronthaul network technology would be an OTN based network.
This is shown in Figure 4.

DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
11 of 25

Author: HARP Consortium


Document number D6.1-001

OTN

BBU pools
RRH aggregation network

Fronthaul with beamforming antennas and packet based RRH aggregation network


Figure 4: C-RAN with OTN based fronthaul

DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
12 of 25

Author: HARP Consortium


Document number D6.1-001

3.0 Aggregation Network Architecture Characteristics


Packet-based fronthaul clearly has some advantages in terms of capacity flexibility.
However, this flexibility comes at a cost, mainly in terms of the requirement for
synchronization. This section examines in more depths the pros and cons of packet-
based fronthaul networks.
3.1 Benefits
There are mainly two benefits that you get when using a C-RAN with packet-based
fronthaul. Both are related to multiplexing, but at different levels.

Efficient use of fronthaul transport resources due to sharing of capacity.


Efficient use of BBU-pool resources (in the case with more BBU-pools) due to a
flexible mapping of cells to BBU-pools.

3.1.1 Efficient use of transport resources


It is well known that packet-based transport offers an efficient use of resources. This
efficiency comes from the fact that in packet-based systems, transmission capacity is
used only when packets are actually transmitted. Therefore, capacity can be shared
among a number of users if we assume that they are not active simultaneously. This is
called statistical multiplexing.
In Figure 5 this is illustrated for a C-RAN based network, where we want to interconnect
the RRHs to the BBU pool. Case a) shows the traditional C-RAN approach with direct
fiber connections. In this case each fiber must have a capacity which is sufficiently large
to transport every cells peak traffic. Moreover the capacities of the fibers are statically
assigned to the RRHs and can thus not be shared with other technologies (e.g., fixed
broadband users in the same area). Case b) shows the packet-based approach. Here
there is some capacity available in the fronthaul that we can flexibly assign to RRHs. The
resources can also be allocated to other use, e.g., by other technologies.
DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
13 of 25

Author: HARP Consortium


Document number D6.1-001

b)

a)

RF

RF

BBU
pool

RF

RF

BBU
pool


Figure 5: Two fundamentally different ways of connecting the RRHs to the BBU pool. a) direct connection
(basically circuit switching) and b) packet switching.

3.1.2 Optimal multiplexing gain


A C-RAN addresses the so-called tidal effect, meaning that base station utilization
changes throughout the day depending on the type of area it serves. Base stations
located in office areas will experience highest traffic load during working hours, while in
the evening they will remain underutilized.

DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
14 of 25

Author: HARP Consortium


Document number D6.1-001




Of,ice base station

Residential base station

50
Load

40
30
20
10
0
0

12

18

24

Time (h)


Figure 6: Daily load on base stations varies depending on base station location [1]

Conversely, base stations located in residential areas are underutilized during the day,
while they need to maintain the capacity to serve users coming home in the evening.
Figure 6 illustrates the fluctuations in the base station load throughout the day. In
traditional RANs baseband capacity is statically assigned to the cell, meaning that the
resources are allocated regardless of the users movements throughout the day. In a C-
RAN BBU resources can be dynamically allocated, thereby benefiting from statistical
multiplexing gain adapting to traffic fluctuations. Figure 7 illustrates that aggregated
traffic in a C-RAN is less bursty throughout the day than traffic profiles of each of the
cells.

DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
15 of 25

Author: HARP Consortium


Document number D6.1-001

RRH 2

RRH 2
...

RRH 1

...

RRH 1

RRH n

RRH n

BBU 1

Aggregated
Traffic (h)

BBU 2
BBU n

= n
Mobile
Backhaul Network

Cloud

= ?
BBU Pool

24 h

Mobile
Backhaul Network

Figure 7: Statistical multiplexing gain in C-RAN architecture for mobile networks

We define a statistical multiplexing gain as presented in Equation 1. We expect the


value to be approximately 2, as statistically a user is at least in two locations during a
day: home and work. We have done an initial modeling of a multiplexing gain, which
gave an approximate of the upper bound value 4 [2]. The modeling only covered daily
traffic fluctuations for chosen application definitions.
N

MultiplexingGain =

PeakThroughput
n =1

RAN

( n)

PeakThroughputC RAN

Equation 1: Statistical multiplexing gain for N base stations

Statistical multiplexing gain will be influenced by daily traffic distribution as well as


traffic (application) definitions. Therefore it will be interesting not only to estimate its
value, but also to see how it is influenced by aforementioned factors. We expect the
value to change depending on the traffic mix, therefore in HAPR task 6.2 we intend to
check:

DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
16 of 25

Author: HARP Consortium


Document number D6.1-001

Exemplary value for statistical multiplexing gain including mobile fronthaul


protocol stack.

How the statistical multiplexing gain depends on percentage of office and


residential cells aggregated in one BBU Pool. OTN or Ethernet based fronthaul
network will enable to explore such a dependency flexibly connecting cells of
different types to different BBU Pools.

3.2 Challenges
Although it might seem that packet-based fronthaul is the Holy Grail in mobile
networks, there are a few challenges.
3.2.1 Synchronization IEEE 1588 and CPRI
Ethernet is an easy to install, cost-effective and mature technology. It supports
multiplexing, which is considered to be cheaper than multiplexing over Wavelength-
division multiplexing (WDM). It is widely deployed; therefore an existing infrastructure
can be reused for C-RAN. However, non-Synchronous Ethernet introduces
synchronization errors that may not meet tight requirements of mobile networks. In
order for an Ethernet switch to be cheap it uses a crystal-based oscillator that can
introduce up to 100 ppm error, while the frequency accuracy requirements for LTE
standard are within 50 ppb. Only 16 ppb can be introduced by mobile backhaul, as
typically 2/3 of inaccuracies are introduced within a base station.
IEEE 1588 or GPS are typically used to assure synchronization in packet-based mobile
backhaul network. GPS signals are hard to receive in urban canyons, where city buildings
may limit the visibility of GPS satellites for small cells. For political reasons, operators
may not want to use the system operated by one country to secure synchronization in
their networks. Last, but not least, GPS signal is susceptible to jamming. IEEE
standardized Precision Time Protocol (PTP) in 1588 standard in order to reply to those
challenges; the latest version of the standard from 2008 is called IEEE 1588-2008. It is
designed to synchronize devices by exchanging time stamps. In HARP task 6.2 we will be
DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
17 of 25

Author: HARP Consortium


Document number D6.1-001




working on answering the research question how IEEE 1588 performs in the Ethernet
environment where time stamps accuracy will be affected by 100 ppm clock
inaccuracy.
1588 Master

1588 Slave

Carrier
Eth

CPRI
RRH

CPRI2Eth

Carrier
Eth

...
Ethernet
switch

Ethernet
switch

CPRI
CPRI2Eth
BBU

Figure 8: CPRI over Ethernet modeling setup

We will model the setup as presented in Figure 8, where (starting from the left) BBU will
be sending CPRI traffic. We will propose a CPRI to Ethernet gateway mapping CPRI to
Ethernet, then there is going to be variable amount of Ethernet switches and finally a
gateway mapping Ethernet traffic back to CPRI and a RRH at the end of communication
channel. 1588 will be working towards achieving synchronization between master clock
located in BBU and slave clock located in RRH. We intend to check:
-

what is the frequency error that achieved in the setup with one Ethernet switch
introducing 100 ppm, if it satisfies 16 ppb then:

what is the maximum number of Ethernet switches that can be present on a link
between BBU and RRH before the 16 ppb requirement is exceeded.

Variable network traffic can influence the performance of 1588. 1588 packets will be
having variable delays passing queues of variable sizes in the Ethernet switches.
Therefore in our modelling we will include:
-

variable background traffic, of both constant and varying nature,

DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
18 of 25

Author: HARP Consortium


Document number D6.1-001

usage of both standard and jumbo Ethernet frames.

If needed, we will work on proposing changes to 1588 that will enable the protocol to
satisfy the requirement for CPRI over Ethernet transmission.
3.2.2 Controlling special antennas
In HARP, special antennas are being used. These antennas might require special control
information, which must be interchanged between the RRHs and the BBU pool. This is
illustrated in Figure 9, which shows a number of cases:
a) The normal C-RAN case in which CPRI is commonly used.
b) Adding special (e.g., beamforming) antennas to the RRHs requires control
information to be interchanged. We call this CPRI++, i.e., an extension of the
CPRI protocol such that the antenna control information can be carried along
with the IQ samples.
c) In the case of a packet-based fronthaul the same information as in in case b)
must be interchanged. However, the packet-based transport might mandate
special treatment.

DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
19 of 25

Author: HARP Consortium


Document number D6.1-001




a)

a)
RF

RF

CP

CP
RI+
+

RI
BBU
pool

CPRI

RF

RF

CPRI++

BBU
pool

c)
RF

P-C
P

RF

RI+
+
BBU
pool

P-CPRI++


Figure 9: Communication between RRHs and the BBU-pool. a) the normal C-RAN case b) Adding special
(e.g., beamforming) antennas to the RRHs. c) Packet-based fronthaul.

The required protocol changes / extensions will be further studied in T6.3.


3.2.3 Other Challenges
There might be other challenges related to using a packet-based fronthaul or at least to
get the full benefit hereof. One example is to adjust the CPRI rate according to the cell
activity.

4.0 Network Planning Aspects


After the presentation of the performance metrics which are of interest in HARP [3] and
the definition of the HARP aggregation network architecture in this deliverable, we aim
DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
20 of 25

Author: HARP Consortium


Document number D6.1-001




to develop network planning models that will provide significant insight on the
advantages of the RRH aggregation network and will also help to further optimize the
design and operation of the C-RAN based network architecture.
In this context, we present here a set of problems that demonstrate great interest for
HARP and specifically for the RRH aggregation network. These problems will be used as
the basis for the network planning models and algorithms that will be developed in Task
6.2 Aggregation Network Optimization during the forthcoming months of the project
duration.
4.1 Aggregation Network Optimization Problems Definition
The network design/planning problems for the optimization of the aggregation network
are targeted at scenarios/use cases II and III, as these are defined in [4]. Scenario II,
named Multiple RRH deals with a service zone with different service areas where a
single macro BS (as part of the BBU pool) serves multiple RRHS. Scenario III Dense
Environment represents typical use cases in urban environments.
As already defined in [3], the performance metrics of interest in HARP are divided in the
following classes: data rate, interference, quality of service, complexity and adaptability.
The aggregation network planning problems introduced here mainly target data rate,
QoS and adaptability metrics and aim to directly or indirectly optimize them to increase
the performance of the network and thus the end user experience. This will be done
through the selection of the parameters that affect the aforementioned metrics, such as
topology, resilience, transmission capacity, baseband processing capacity, BBU pool
location/placement and coverage. Based on these parameters, we define in the
following sections a set of problems that will be further modeled and investigated in
D6.2. These problems vary in terms of the objective function to be optimized, the
parameters to be evaluated and the constraints that describe the details of each
scenario. In every case, we aim to use the defined scenarios (II and III) as references
where each problem can be applied.
DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
21 of 25

Author: HARP Consortium


Document number D6.1-001




4.1.1 Network Dimensioning
Through network dimensioning problems we can compute the minimum network
capacity required for a given traffic input. In such problems, we usually start with a
static traffic matrix (a traffic instance) that represents a worst case scenario in terms of
the total traffic load that needs to be exchanged between two nodes. We then
formulate the objective function either as a direct representation of the number of links
(and even their paths in a green-field problem) and the respective capacity, or as a cost
function in the case where we have specific inputs regarding e.g. cost per Ethernet port.
Then, for the specific traffic matrix, we compute the optimal network capacity that is
needed to handle the traffic, taking into account all the constraints that might be
present.
In more elaborate models, we apply similar optimization techniques for multiple traffic
periods and given a traffic pattern similar to the one presented in Section 3.1.2. In this
case, we obtain a finer detail of the capacity needed at each network link for different
time periods. This can be highly interesting since it gives us the ability to efficiently plan
the capacity sharing for different time periods and to avoid over-provisioning or
blocking.
Although we focus on the network part, similar models can be developed that aim to
jointly dimension the network and the BBU computing resources. The objective can also
be tied either to cost or to power consumption of the network and BBU equipment. In
the latter case, efficient planning can give us the ability to shut down unused parts of
the optical network and/or the BBU pool and achieve further savings in terms of power
consumption. Similar models for different networks have already been demonstrated
among others in [5].
4.1.2 BBU Pool Placement
Starting with the assumption that the BBU pool is already placed and we can only decide
on the dimensioning of the infrastructure (network and BBU), an interesting problem is
also the consideration of BBU pool placement. In this case, we aim to decide again
DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
22 of 25

Author: HARP Consortium


Document number D6.1-001




based on traffic characteristics and services requirements such as delay, where on the
RRH aggregation network we should place the available BBU pools. This kind of problem
can consider different constraints, such as the aforementioned service requirements,
the budget, and of course the requirements imposed by the protocols such as CPRI,
where the maximum distance between the RRH and the BBU pool is defined at 40 km.
4.1.3 Assignment of Cells to BBUs
As already mentioned, the choice of a packet-based fronthaul RRH aggregation network
offers among others the ability to assign cells (RRHs) to more than one BBU pools. In
such problems, we consider a set of interconnected BUU pools, each one serving a set of
RRHs. Using appropriate models, we can plan the assignment of RRHs to BBU pools
according to traffic characteristics and user movement and thus optimize the network in
terms of the adaptability network characteristic.
4.2 Network Optimization Methodology
All problems are going to be modeled as optimization problems using exact methods
such as Integer Linear Programming (ILP) and stochastic optimization methods such as
the Sample Average Approximation [6] method for two-stage stochastic linear
problems. In their general form, the problems consist of an objective function to be
minimized and several constraints that describe the details of each problem. According
to each specific problems requirements, we will model deterministic models through
the most usual ILP formulations (e.g. for dimensioning). Additionally, more elaborate
formulations such as those obtained by applying the SAA method will be developed in
order to capture events of stochastic nature, such as user mobility. Similar problems
have been presented in [7] and [8].
These formulation fall in the class of offline problems, since due to complexity they
cannot be solved and provide solutions in a dynamic environment, but are rather used
for planning purposes, as the title implies. Although these are going to be the main
optimization techniques, we also aim to investigate the formulation of online decision
DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
23 of 25

Author: HARP Consortium


Document number D6.1-001




making problems to model more dynamic scenarios, such as the RHH-BBU assignment
according to changing parameters related to user mobility and service requirements.

5.0 Summary
This deliverable has presented the basic model of the RRH aggregation network that
deals with the fronthaul network of the HARP project C-RAN based network
architecture. The overview of the aggregation network and the benefits of a packet-
based fronthaul have been presented. Also, the main benefits and challenges have been
discussed, mainly focusing on the multiplexing gain and synchronization issues. The
deliverable concluded with the definition of important problems to be addressed
through network optimization in the upcoming deliverable D6.2 and the respective
methods to be used.

6.0 References
[1] C-RAN The Road Towards Green RAN, China Mobile Research Institute, Tech. Rep.,
October 2011.
[2] A. Checko, H. Christiansen, and M. S. Berger, Evaluation of energy and cost savings
in mobile Cloud-RAN, in Proceedings of OPNETWORK Conference, 2013
[3] Requirements, metrics and network definition D3.1 (M8)
[4] Y1 Scientific Report D1.3 (M12)
[5] Georgakilas, K.N.; Tzanakaki, A.; Anastasopoulos, M.; Pedersen, J.M., "Converged
optical network and data center virtual infrastructure planning," Optical
Communications and Networking, IEEE/OSA Journal of , vol.4, no.9, pp.681,691,
Sept. 2012

DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
24 of 25

Author: HARP Consortium


Document number D6.1-001




[6] B. Verweij et al., The sample average approximation method applied to stochastic
routing problems: A computational study, Computational Optimization and
Applications, vol. 24, pp. 289333, 2003.
[7] Georgakilas, K.N.; Anastasopoulos, M.P.; Tzanakaki, A.; Zervas, G.; Simeonidou, D.,
"Planning of converged optical wireless network and DC infrastructures in support of
mobile Cloud services," Optical Communication (ECOC 2013), 39th European
Conference and Exhibition on , vol., no., pp.1,3, 22-26 Sept. 2013
[8] Georgakilas K.N; Anastasopoulos, M.P.; Tzanakaki, A., Stochastic Optimization for
Deployment of Correlated Cloud Services over Optical Networks, IEEE
Communications Letters, to appear.

DOCUMENT NUMBER

REV

D.6.1-001

1.0.0

TITLE
D6.1 Aggregation network definition

The information contained herein is considered proprietary and confidential


All rights reserved

PAGE
25 of 25

You might also like