Beruflich Dokumente
Kultur Dokumente
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Dimensioning Guideline
Dimensioning Mobile Access, MBB CS Network Engineering
For internal use only, approved, version 1.0, 09.05.2013
Please always check the latest version of this document under the following link: here
2/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Revision History
Version
1.0
Date
Change Notes
Authors
Name
Department
Dominik Dulas
Szymon Listwan
3/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Table of contents
1.1
1.1.1
1.1.2
1.1.3
1.1.4
1.1.5
1.1.6
1.2
1.2.1
1.2.2
1.3
1.3.1
1.3.2
1.3.3
1.3.4
1.3.5
2.1
2.1.1
2.1.2
2.2
2.2.1
2.2.2
2.2.3
2.2.4
2.2.5
2.2.6
2.2.7
4/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
References
[LTE_Transp_SFS]
[LTE_Capacity+TM_SFS]
[LTE_Perf_SFS]
[LTE_TM]
[LTE_Radio_Dim]
[LTE_EPC_Dim]
[MD1]
M. H. A. Davis, J. M. Howl, A Markovian Analysis of the FiniteBuffer M/D/1 Queue, Proceedings: Mathematical, Physical and
Engineering Sciences, Volume 453, Issue 1964, pp. 1947-1962
[MGR-PS]
[TS29.281]
[TS36.413]
3GPP TS 36.413, Evolved Universal Terrestrial Radio Access (EUTRA); S1 Application Protocol (S1AP)
[TS36.423]
[LTE_Transp_+_QoS]
5/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
1.
1.1
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
LTE Access Transport dimensioning covers interfaces at the eNB side (last mile) as
well as at the network side (front mile).
Note:
Evolved UTRAN
S1_MME
eNB
X2_C
X2_U
MME
eNB
O&M i/f
S1_U
6/63
SAE-GW
O&M i/f
O&M
Overview of the LTE Access interfaces for U-Plane, C-Plane and M-Plane, together with
the involved protocol stacks can be seen in Figure 2.
7/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
LTE U-Plane
eNB 2
eNB 1
SAE-GW
X2-U
S1-U
GTP-U
GTP-U
GTP-U
UDP
UDP
UDP
GTP-U
IP
IP
IP
IP
IPSec (*)
IPSec (*)
IPSec (*)
IPSec (*)
L1/L2
L1/L2
L1/L2
L1/L2
UDP
LTE C-Plane
eNB 2
eNB 1
MME
S1-AP
S1-MME
SCTP
X2-AP
SCTP
IP
IP
IPSec (*)
IPSec (*)
L1/L2
L1/L2
eNB 2
eNB 1
X2-C
SCTP
S1-AP
MME
X2-AP
SCTP
IP
IP
IPSec (*)
IPSec (*)
L1/L2
L1/L2
LTE M-Plane
eNB 2
eNB 1
NetAct
O&M i/f
O&M i/f
Mgmt. Appl.
Mgmt. Appl.
UDP/TCP
UDP/TCP
Mgmt. Appl.
IP
IP
IP
IPSec (*)
IPSec (*)
IPSec (*)
L1/L2
L1/L2
L1/L2
UDP/TCP
Figure 2 U-Plane (UP), C-Plane (CP) and M-Plane (MP) protocol stacks in LTE Access.
8/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
1.1.1
1.1.2
1.1.3
1.1.4
1.1.5
1.1.6
1.2
9/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
S1, X2
1.2.1
BTS transport configuration and the supported capacity determine the LTE Access
dimensioning process, as described in Sec. 2.2.
10/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
1.2.2
Value
Comments
3 (*) (**)
6 (***)
Users scheduled in
200 (*) the time domain by
the RRC scheduler;
250 (**) (***) subset of Active
users.
Number of GbE
interfaces at the eNB
1000 Mbps that can be used
towards the Core is
limited to 1.
(*) for RL10, (**) for RL20 / RL15, (***) for RL30 / RL25 / RL40
1.3
11/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
RL20/RL15
RL30/RL25
RL40
RL50/RL35
Note: Only features impacting the LTE Access Dimensioning are shown.
Figure 4 RL10, RL20 / RL15, RL30 / RL25, RL40, and RL50 / RL35 transport features
affecting LTE Access dimensioning.
1.3.1
RL10 features
1.3.1.1
12/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
E-UTRAN
UE
EPC
eNB
S-GW
Internet
P-GW
Peer
Entity
End-to-end Service
EPS Bearer
E-RAB
Radio Bearer
Radio
External Bearer
S5/S8 Bearer
S1 Bearer
S1
S5/S8
Gi
LTE bearers are split into 2 types: Guaranteed Bit Rate (GBR) bearers and non-GBR
bearers. Each bearer type is allocated different parameters in terms of supported bit
rates, QoS Class identifier (QCI) and Allocation and Retention Priority (ARP). Another
split of LTE bearers is between default and dedicated bearers:
Default LTE bearer: first bearer being established for UE at connection setup.
Dedicated LTE bearers: bearers established for UE in addition to the default
bearer.
Current system limitations in RL10 are the following:
Only nGBR bearers are supported (supported QCIs: QCI5 QCI9)
One nGBR bearer per UE (a default one, no dedicated bearer support)
Up to four realtime or two non-realtime services per bearer (either/or)
These limitations affect the traffic model definition in terms of number and type of
supported services per LTE user.
1.3.1.2
13/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
number of services and users supported per cell. No CAC on S1 means no blockingbased performance measures are considered in the LTE Access dimensioning (like
Erlang-B and/or Erlang-C).
1.3.1.3
QCI
Conversational Voice
Conversational Video
DSCP
DiffServ
PHB
Ethernet
p-bits
46
EF
36
AF42
Not supported
in RL10
GBR
Non-conversational Video
26
AF31
46
EF
IMS signaling
34
AF41
18
AF21
28
AF32
10
AF11
BE
non-GBR
Figure 6 QoS differentiation and mapping options in the LTE Transport supported in
RL10.
14/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
1.3.2
1.3.2.1
QCI
DSCP
DiffServ
PHB
Ethernet
p-bits
Supported in RL20
Conversational Voice
46
EF
Conversational Video
36
AF42
Non-conversational Video
26
AF31
Not supported
46
EF
in RL20
IMS signaling
34
AF41
18
AF21
28
AF32
10
AF11
BE
with LTE010
GBR
non-GBR
Figure 7 Service differentiation options supported in RL20 / RL15 for LTE Access
Transport.
15/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
1.3.2.3
1.3.2.4
1.3.2.5
1.3.2.6
16/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Table 2 LTE458 parameters impacting LTE Access dimensioning and their maximum
values.
Radio channel bandwidth
Parameter
3 MHz
5 MHz
10 MHz
15 MHz
20 MHz
120
480
600
720
840
40
192
240
240
250
(*) Active users: RRC_connected users with a data bearer established; traffic demand per eNB is
calculated with respect to these users
(**) Actively scheduled users: Active users scheduled by the RRM scheduler to be served in the time
domain; Air I/F throughput limits are verified with respect to these users.
1.3.3
1.3.3.1
1.3.3.2
1.3.3.3
17/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
some settings it may achieve throughputs which should be accounted into overall link
capacity calculation. After some analysis only three scenarios, which have significant
impact on dimensioning, were specified. They can be selected by a user and taken into
account during link dimensioning. These are:
2 TWAMP sessions per BTS, low TWAMP sending rate
results with ~7 [kbps] additional throughput
2 TWAMP sessions per BTS, high TWAMP sending rate
results with ~64 [kbps] additional throughput
Max TWAMP bandwidth configuration (single TWAMP session per BTS with highest
rate and biggest packet size)
results with ~120 [kbps] additional throughput
1.3.3.4
1.3.4
RL40 features
1.3.4.1
1.3.4.2
18/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
1.3.5
1.3.5.1
19/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
2.
Note that these are logical interfaces which share the physical transport interface at the
eNB. Specification of the transport capacity requirements for these interfaces is of the
utmost importance if the operator does not have its own transport infrastructure and
has to use leased lines. eNB transport capacity components subject to dimensioning
are shown in Figure 8.
In the following sections the LTE Access dimensioning workflow is described. Section
2.1 provides a definition of input and output parameters used in the dimensioning. In
Section 2.2 a detailed dimensioning methodology for LTE Access is described.
20/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
2.1
21/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
LTE Access
dimensioning inputs
for the eNB along with the capacity split per logical interfaces. Detailed description of
the output parameters is provided in Sec. 2.1.2.
Features configuration
Feature selection
Feature parametrization
Services
Subscriber profile
definition
Service set definition
Traffic demand
Transport configuration
Number of sites
QoS Class configuration
Scheduler weights
Area configuration
Front Mile
Area type
Software release
Site layout
Area selection
Number of eNBs
Load Distribution Factor
U plane dimensioning
LTE Access
dimensioning outputs
C plane dimensioning
(S1_MME & X2_C)
Dimensioning Output
Transport capacity per logical interface
S-Plane (optional)
Total transport capacity
2.1.1
Area configuration
Features configuration
Services
Traffic demand
Transport configuration
Aggregation point inputs
Area configuration
This group of inputs specifies the area configuration. It consist of such parameters as
the duplex mode, area type, operating band, bandwidth, channel bandwidth, and
software release. Note, that not all of them impact directly the access network
dimensioning. In this section, only parameters directly impacting the access
dimensioning will be covered. They should be specified in the first place as they
determine the overall dimensioning process.
22/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
2.1.1.2
Default value
Comment
RL40
Features configuration
Features configuration step allows the feature set configuration. It is divided into three
sections:
Services
Services are used for dimensioning based on the user traffic demand only. In this step
user defines the subscriber profile with the service sets. For each service it is possible
to set the basic parameters describing it.
Table 4 LTE Access dimensioning inputs: Traffic model inputs LTE User Profile.
Service parameters
Default value
Name
Internet Static
Application
Bearer
QCI 9
Phase
Enabled
Packet size
Service dependant
Comment
23/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
2.1.1.4
2%
15 [ms]
Traffic demand
Traffic demand is used for dimensioning based on the user traffic demand only. Traffic
demand specifies the user traffic requirement. For each subscriber profile and service
set, it is possible to define parameters separately, which enables high flexibility of traffic
model definition. Parameters used in this step are divided into four subsections:
General
U-plane
C-plane
Smart applications
TM parameters defined at
Default value
End User/application
level
N/A
Comment
Indicates whether the traffic demand
is specified:
per subscriber
per area
Indicates whether the traffic demand is
specified at:
Network level
End User/application level
Defines number of subscribers per
area in case of traffic demand
specified per subscriber.
24/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Default value
Dependent on
selected service
UL/DL
Comment
Service name
Direction indication
BHCA
Dependent on
selected service
Call/session duration
Dependent on
selected service
Dependent on
selected service
Traffic demand
Dependent on
selected service
Default value
Comment
0,1
1,1
50%
15%
4,5
1,8
25/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Default value
Comment
90 [s]
Percentage of incoming
messages
50%
150 [B]
90 [s]
0,5
26/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
2.1.1.5
Transport configuration
Transport configuration is used for dimensioning based on the user traffic demand only.
This section consist of transport network configuration such as VLAN tag, QCI mapping
to Per Hop Behavior (PHB) queue and PHB weights. Parameters are described in Table
9.
Table 9 LTE Access dimensioning inputs: Transport configuration.
Transport configuration
parameters
Number of sites
Default value
N/A
N/A
2.1.1.6
VLAN tag
Comment
Enabled
QCI
Dependent
PHB
dependent
PHB
dependent
27/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Number of sites
2.1.1.7
Default value
N/A
Comment
Defines which areas are included into
aggregation point calculation
False
Default value
Comment
Area
N/A
Number of sites
N/A
2.1.2
28/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Comment
2.2
2.2.1
General concept
There are two general approaches proposed for the transport capacity dimensioning:
-
Network planner is free to use any of the above options knowing the availability of input
data and potential advantages and drawbacks of each option, as summarized in Sec.
2.2.2.3. Here, a general dimensioning workflow is described for both options and
illustrated in
Figure 9.
Note:
29/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
2.2.1.1
Specification of the traffic demand: First, the traffic offered to the LTE Access
network has to be specified. Parameters required to appropriately model the
offered U-Plane and C-Plane traffic are described in the LTE Traffic Model and
Signaling sections. The TM parameters values need to be set by the operator
or, alternatively, taken from the LTE reference traffic model specified in
[LTE_TM].
Optional: Front mile: Selection of aggregation node (with its interface transport
capacity) and specification of number of eNBs to be aggregated.
2.2.1.2
30/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
LTE Access dimensioning steps for both options are presented in Figure 10.
LEGEND:
Intermediate results
QoS settings
Resulting eNB
transport capacity
Area configuration
Radio-related
inputs
Output
DIMENSIONING BASED
ON THE RADIO I/F THROUGHPUT
Input
2.2.2
31/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
2.2.2.1
Service type
[RT or NRT]
QCI
[GBR or non-GBR]
Dimensioning
formula
QoS target
GBR
MDErlang
blocking probability
Non-GBR
M/D/1
single-link delay
Non-GBR
Ext M/G/R-PS
single-link delay
Real-time
Non-real-time
For real-time services there are two possibilities for dimensioning formula selection. If
the service is real-time, the further check is necessary. For services assigned to GBR
bearers, the MDErlang formula is used. Note that calculation is performed once for all
GBR services and no QoS target in seconds is applied except blocking probability.
Real-time services assigned to non-GBR bearers are calculated with M/D/1 formula with
single-link delay as QoS target (each separately). For all non-real-time services
extended M/G/R-PS with single-link delay QoS target is applied.
Note:
32/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
{ S1 _& _ X 2 _ Capacity _ RT
Service
Service
The RT capacity per service is calculated using either the MDErlang formula (for all RT
GBR services, Equation 3) or the M/D/1 formula [MD1] (for RT non-GBR services,
Equation 4):
Equation 3
Equation 4
where:
packet_delay Service [ms] is the max tolerable transport delay of IP packets on the
last mile link. It is the assumed QoS target for dimensioning the RT traffic,
specified separately for each service class. Default value (for each service class)
is 15 ms.
packet_size
Service
33/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
PHB_weight Service [#] is the weight of the DiffServ scheduler queue that serves
the QoS class of the service; it determines the bandwidth available to that QoS
class during the congestion (i.e. all queues fully occupied with traffic).
S1_U_OH Service [%] is the S1_U transport overhead per service.
Note: the offered user traffic load should be compared against the average cell
capacity. This is to check if the offered traffic can be accommodated within the
supported radio capacity.
Note:
The application e2e delay per service is calculated using the M/G/R-PS formula [MGRPS]:
Equation 6
34/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Equation 7
Uu_OH Service [%] denotes the protocol overhead on the Radio interface
(including RLC and PDCP layers),
Note: the offered user traffic load should be compared against the average cell
capacity. This is to check if the offered traffic can be accommodated within the
supported radio capacity.
The application e2e delay per service is calculated for the dimensioned link capacity
and for ideal, non congested link. Using these two delays, the Transport Network delay
can be estimated, which is the QoS requirement for the dimensioning. Finally, NRT
capacity which fulfills QoS requirement for each service can be calculated.
Equation 8
file _ sizeService
where:
35/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Dimensioning the U-Plane bandwidth for NRT traffic using M/G/R-PS is illustrated in
Figure 11.
Calculating user traffic due to inter-eNB handovers
U-Plane traffic of an eNB due to inter-eNB handovers is calculated with the following
formula:
Equation 9
DPDCP [ kbits ] 3
3
2
10 ( 1,3452 Uu _ util 1,6896 Uu _ util Uu _ util 0,1024 ) ( Uu _ util 0,3 )
where:
Equation 11
Uu _ util [# ]
where:
gross_traffic_load_per_cell [kbps] is the total user traffic in a radio cell with radio
overhead.
36/63
Note:
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
If calculated X2_U capacity is zero, but the X2 load is greater than zero, it
is assumed that X2 traffic is accommodated in S1_U capacity.
General dimensioning concept of the U-Plane bandwidth traffic (based on the user
traffic demand) is graphically illustrated in Figure 11.
Service = RT
YES
Blocking_probability
Service =
GBR
GBRservice
YES
MD-Erlang
(for all RT
GBR services)
NO
Load RT GBR
S1_(&_X2)_traffic_load Service [kbps]
packet_delay Service [ms]
packet_size Service [bytes]
S1_U_OH Service [%]
M/D/1
S1_(&_X2)_Capacity_
RT_GBR [kbps]
Max
PHB_weight Service
NO
per service
ave_cell_throughput [kbps]
Uu_OH Service [%]
Weighted
M/G/1-PS
QCI_weight
S1_(&_X2)_traffic_load per cell [kbps]
S1_(&_X2)_Capacity_
RTnGBR [kbps]
S1_(&_X2)_Capacity_
NRT+RTnGBR+
RT_GBR_load [kbps]
Sum
r_peak Service
Max from all
NRT services
S1_(&_X2)_Capacity_
NRT [kbps]
M/G/R-PS
per PHB
TN Delay
estimation
PHB_weight Service
file_size Service [kbytes]
TN delay <
Target Packet
delay
S1_(&_X2)_Capacity [kbps]
2.2.2.2
All Average
All Average / Single Peak
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
All Average
With this option the total U_Plane transport capacity at eNB shall support the
aggregated average capacity of all cells. The calculated U_Plane transport capacity
should be:
Equation 12
Cell average
Cell peak
37/63
Figure 12 Dimensioning based on the Radio i/f throughput the All Average option.
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
eNB _ U _ PlaneTransport _ Capacity Radio Based [ kbps ] MAX { peak _ cell _ throughput ;
# _ of _ cells _ per _ eNB avg _ cell _ throughput } ( 1 Uu _ OH ) ( 1 S1 _ U _ OH )
where:
Cell average
The All Average / Single Peak option is graphically presented in Figure 13. This is the
recommended option for Radio-based dimensioning (often 3G operators use it to
dimension the HSPA traffic).
Cell peak
38/63
Figure 13 Dimensioning based on the Radio i/f throughput the All Average / Single
Peak option.
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Cell average
Cell peak
39/63
Figure 14 Dimensioning based on the Radio i/f throughput the Single Peak plus All-butone-Average option.
All Peak
With this approach the U-Plane transport capacity at eNB shall support the aggregated
peak capacity of all cells. As the load per eNB will hardly achieve peak values in all cells
at the same time, this approach will lead to over-dimensioning, thus usually extra costs.
The calculated U-Plane transport capacity should be:
Equation 15
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Cell peak
40/63
Figure 15 Dimensioning based on the Radio i/f throughput the All Peak option.
Figure 16 shows how the U-Plane traffic of logical S1_U and X2_U interfaces is
distributed within the eNB and how it is mapped onto the available transport capacity.
Radio side
(eUTRAN)
Transport side
(ePC)
eNB
S1_UL
eNB
X2_DL_out
UL
eNB_Transp_Cap (UL)
S1_UL
X2_DL_out
X2_DL_out
S1_DL
S1_DL
S1_DL - X2_DL_out
S1_DL - X2_DL_out
DL
eNB_Transp_Cap (DL)
X2_DL_in
X2_DL_in
a)
b)
Figure 16 U-Plane traffic distribution within the eNB: a) logical distribution, b) mapping to
the transport capacity.
41/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Equation 16
where:
X2_U_DL_out [kbps] is the outgoing X2 traffic, for which the eNB is the source
eNB,
X2_U_DL_in [kbps] is the incoming X2 traffic, for which the eNB is the target
eNB
X 2 _ U _ DL _ out[kbps] X 2 _ U _ DL _ in[kbps]
Assume as well that the total X2_U_DL traffic is some percentage of the average S1_U
DL traffic:
Equation 19
42/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Equation 22
2.2.2.3
43/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Table 14 Dimensioning based on the user traffic demand vs. dimensioning based on the
Radio interface throughput comparative analysis.
PROS
Dimensioning based
on User traffic demand
Dimensioning based
on Radio I/F
throughput
CONS
2.2.3
2.2.3.1
General concept
Similarly to the U-Plane traffic dimensioning, also the C-Plane traffic is dimensioned
according to the two proposed options:
-
Both dimensioning options are described in detail below. General concept of the
dimensioning is shown in the Figure 17.
44/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
LEGEND:
Input
List of considered
events
Intermediate results
Output
Message size
Overhead size
Attach frequency
BHCA
Number of handovers etc.
Frequency of an event
C-Plane traffic as a
percentage of the U-Plane
Figure 17 General concept of control plane traffic load calculations in LTE access
network.
2.2.3.2
45/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
1. List of events causing the control plane traffic. Each event is described by a detailed
message flow and size of control messages, including the overhead (SCTP + IP +
(IPSec) + Eth). These parameters are used for calculating the traffic load per event
as a sum of lengths of all messages sent on each interface and direction.
Considered signaling events are:
Attach
Detach
Handover
Tracking Area Update
State transition
Paging
2. Frequency of each event per subscriber (Fevent [1/s]) is calculated based on the
traffic model. For every event the formula is different and is presented in this
chapter.
3. Traffic load per event resulting from multiplying the traffic load generated by the
event (Levent [bits]) with its frequency (Fevent [1/s]):
Equation 24
T bps # UE Tevent
event
where:
T [bps] is the total traffic load generated by the selected events on S1_MME/X2_C
interface,
Signaling events
Following subchapters explain a set of signaling events used for calculation of the
control plane load. Each subchapter contains following information:
Event load - is a total amount of data sent in each interface and direction without
transport overhead.
Event frequency this section will describe the formula and parameters used for
event frequency calculation.
46/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
[bytes]
S1 (DL)
S1 (UL)
S1 (DL)
S1 (UL)
Attach
396
594
Detach
82
108
Attach/Detach
Event frequency:
It is assumed that frequency of Attach event is equal to Detach event and the Attach to
LTE_Active state is equal to Attach to LTE_Idle state. Default value can be found in
NSN Traffic Model.
Handover (Access Stratum signaling)
For C-plane calculations intra-eNB handover is not taken into calculations due to the
fact that no traffic is generated either in S1_MME or in X2_C interfaces. Only inter-eNB
handover is considered.
Table 16 Event load: Handover.
Signaling load
Inter eNB
handover
[# of messages]
[bytes]
[# of messages]
[bytes]
S1 (DL)
S1 (UL)
S1 (DL)
S1 (UL)
X2 (DL)
X2 (UL)
X2 (DL)
72
88
320
Event frequency:
The handover frequency per single call is being specified by user.
[bytes]
[# of messages]
S1 (DL)
S1 (UL)
S1 (DL)
S1 (UL)
60
90
Event frequency:
The handover frequency per single subscriber is being specified by user.
X2 (
13
47/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
[# of messages]
[bytes]
S1 (DL)
S1 (UL)
S1 (DL)
S1 (UL)
120
132
Event load is different according to the initiator. It is assumed that the event is triggered
by UE and Network equally.
Event frequency:
Equation 26
1
Idle _ to _ Active _ Load BHCAU plane
s
LTE_Active to LTE_Idle
Table 19 Event load: LTE_Active to LTE_Idle.
Signaling load
[bytes]
[# of messages]
LTE_Active to
LTE_Idle
S1 (DL)
S1 (UL)
S1 (DL)
S1 (UL)
40
78
Event frequency:
Assuming constant share of active subscribers, the amount of LTE_Idle to LTE_Active
state changes is equal to the reverse event.
Impact of smartphones on number of state transitions.
Smartphones load the network in new way. Their always-on applications rely on keepalive messages to keep the connectivity open. On the other hand, to preserve the
battery, the fast dormancy features have been introduced by UE vendors. With that
feature, shortly after data transmission is over, the smartphone UE goes to the idle
mode. messages sent by application cause reverse state change to active mode. These
additional events are accounted in c-plane dimensioning based on the equation below:
48/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Equation 27
0
1
sm _ msg 3600
s time _ msg
where:
Equation 28
0
1
3600
sm _ k a _ msg
s time _ keep alive _ msg
where:
Equation 29
1
sm _ events sm _ msg sm _ k a _ msg
s
where:
Note 1:
If the keep-alive_interval parameter is smaller than the RRC_release_timer, no
additional events are considered as the RRC_release_timer will never expire (Figure
18):
49/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
time between
messages
RRC
connected
call /
session
RRC idle
RRC release
timer
Note 2:
If the keep-alive_interval parameter is greater than the RRC_release_timer, the
additional events are added to the total number of state transitions. Only the Idle users
are considered, as users who actively exchange data with network cannot go to idle
mode (RRC_release_timer will not expire) (Figure 19):
50/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
time between
messages
RRC
connected
call /
session
RRC idle
RRC release
timer
RRC release
timer
State transitions
active->idle / idle->active
[# of messages]
[bytes]
S1 (DL)
S1 (UL)
S1 (DL)
S1 (UL)
60
Event frequency:
Equation 30
1
Paging retry _ factor TAcp TAsize MTCSF BHCAU Plane
s
where:
51/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Default value given by NSN default traffic model: retry_factor = 1,1 [#]
TACP [#] is number of Tracking Areas under a network paging control point
(MME, eNB).
Default value given by NSN default traffic model: TACP = 1 [#]
2.2.3.3
BHCAU-plane [#] is a sum of BHCA of all applications per subscriber. The same
parameter as used for state transitions calculations.
Equation 34
52/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Equation 35
S1_MME_DL [kbps] =
2 CP _ Share[%]
eNB _ Transport _ Capacity _ DLALL _ AVG
2 HO _ Share[%]
Equation 36
HO _ Share[%]
eNB _ Transp _ Cap _ ULALL _ AVG 2 HO _ Share[%] eNB _ Transp _ Cap _ DLALL _ AVG
Signaling traffic on the X2 interface due inter-eNB handovers for downlink is calculated
as a percentage of the downlink X2 U-Plane traffic:
Equation 37
53/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
2.2.4
2.2.5
S-plane dimensioning
S-plane interface is in use with Timing over Packet (ToP) standard. For dimensioning,
very straightforward approach is adopted. Having the rate of messages and their
lengths with transport overhead, the required interface capacity can be derived. Below
one can find calculation details.
2.2.5.1
BW _ ToP[ kbps ] PTP _ sync _ msg _ size Trs _ OH PTP _ sync _ msg _ rate
Exemplary calculation:
The values used below are defaults. For some configurations they can
vary.PTP_announce msg_size:
544 [bits]
PTP_sync_msg_size:
352 [bits]
PTP_announce_msg_rate:
0,5 [1/s]
PTP_sync_msg_rate:
16 [1/s]
Trs_OH (Eth+IP+UDP):
336+160+64 = 560 [bits]
Equation 43
Synchronous Ethernet
Although Synchronous Ethernet is not in use with S-plane interface, it may be used
instead of the ToP standard for synchronization purposes. Therefore appropriate
calculation is shown below.
54/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
BW _ SyncE [ kbps ] ( SSM _ PDU _ size SSM _ Hdr _ size ) SSM _ msg _ rate
where:
Exemplary calculation:
The values used below are default ones. For some configurations they can vary.
SSM_PDU_size:
SSM_msg_rate:
SSM_Hdr_size:
6104 [bits]
5 [1/s]
224 [bits]
Equation 45
2.2.6
55/63
Radio side
( eUTRAN)
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Transport side
( ePC)
eNB
eNB
S1_U_UL
X2_C_out_DL
X2_U_out
S1_ MME_UL
X2_C_out_UL
X2_U_out
S1_U_UL
S1_ MME_UL
X2_U_out
S1_U_ DL - X2_U_out
X2_C_out_DL
FTM
(FTIB or
FTLB)
X2_C_in_UL
UL
S1_U_DL
S1_U_DL
S1_ MME_DL
S1_U_ DL - X2_U_out
S1_ MME_DL
X2_U_in
X2_C_out_UL
X2_C_in_DL
X2_C_in_UL
X2_C_in_DL
a)
DL
X2_U_in
b)
Figure 20 Distribution of the eNB traffic between UL and DL parts of the eNB transport
card: a) logical flow distribution, b) allocation to the eNB transport interface.
Based on Figure 20, the transport capacities to be set for UL and DL parts of the eNB
transport card can be figured out.
2.2.7
C C1 C2
then the difference (C1 + C2) - C indicates the benefit resulting from traffic aggregation.
56/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
The dimensioning algorithm finds the lowest front mile link capacity that is still able to
provide the required level of service to all aggregated eNBs. Calculation methodology
proposed here applies exactly the same models as for last mile dimensioning, extended
with one step to calculate the multiplexing gain. M/D/1/MDErlang models are used for
Real-Time services and M/G/R-PS for non-Real-Time; necessary modification and
adjustment of input parameters is shown below. The description focuses on the
differences comparing to the last mile dimensioning.
Note:
In order to obtain the statistical multiplexing gain both, last and front mile,
links capacities are required. Due to the fact, that the QoS target may be
different for last and front mile, the referenced last mile result, used for
multiplexing gain calculation, is re-calculated with front mile delay target.
Aggregated from all eNBs RT capacity per service is calculated using either the
MDErlang formula (for all RT GBR services, Equation 48) or the M/D/1 formula [MD1]
(for RT non-GRB services, Equation 49)
Equation 48
where:
S1_&_X2_traffic_load
and X2.interfaces.
Service
where:
eNB ,Service
57/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
packet_delay Service [ms] is the max tolerable transport delay of IP packets on the
front mile link. It is the assumed QoS target for dimensioning the RT traffic,
specified separately for each service class. Default value (for each service class)
is 15 ms.
packet_size
Service
eNB ,Service
eNB
The same PHB scheduling weights are applied for all eNBs as well as for
aggregation point.
S1_&_X2_traffic_load
eNB,Service[kbps]
eNB
M/D/1
When looking at last and front mile links dimensioning, two major differences can be
found:
no comparison of offered user traffic load against the average radio cell
capacity. This step is already performed during LM link calculation, therefore no
need to repeat this.
input parameter: traffic load per service is a sum of traffic from all users served
by aggregated eNBs.
58/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
As the same model is applied for front mile dimensioning, the dimensioning process is
very similar to the one used in 2.2.2.1. Red marking indicates the modification
comparing to the Figure 11.
Total_U-Plane_traffic_per_cell [kbps]
ave_cell_throughput [kbps]
Uu_OH Service [%]
QCI_weight
Weighted
M/G/1-PS
S1_&_X2_traffic_load_per_cell [kbps]
r_peak Service
S1-U_OH Service [%]
S1_&_X2_traffic_load
eNB
[kbps]
M/G/R-PS
PHB_weight Service
TN Delay
estimation
S1_&_X2_Capacity_NRTService[kbps]
Figure 22 Front Mile S1_U dimensioning with NRT traffic using M/G/R-PS.
no comparison of offered user traffic load against the average radio cell
capacity. This step is already performed during LM link calculation, therefore no
need to repeat this.
input parameter: traffic load per PHB is a sum of traffics from all users served by
aggregation point:
Equation 50
PHB ( Service )
where:
o k number of aggregated eNBs
aggregated eNBs (traffic sources) can have different traffic behavior (e.g.
ave_cell_throughput). For each eNB i, the appropriate set of parameters is being
used for M/G/R-PS and the resulting front mile bandwidth is weighted by the
load of individual eNB:
Equation 51
k
FM _ BWNRT FM _ BWNRT (i )*
i 1
Load _ eNB(i)
k
Load _ eNB( j)
j 1
59/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
where:
FM_BWNRT [kbps] is a resulting front mile capacity of NRT services,
k [#] is a number of aggregated eNBs,
2.2.7.1
The multiplexing gain calculation is a very last step of front mile dimensioning. Having
the last and front mile link capacities, one can calculate the statistical multiplexing gain
resulting from merging the traffic from several sources:
Equation 52
Aggregated _ BW
Multiplexi ng _ gain [%] 1
Bw _ 1 .. Bw _ n
where:
Aggregated_BW [kbps] is the resulting front mile link capacity,
Bw_... [kbps] are the transport capacities of eNB last mile links.
The resulting multiplexing gain value indicates how much the traffic from several
sources can be squeezed and what saving can be achieved on front mile link.
60/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
2.2.7.2
One can see in the Figure 24, the even traffic distribution results in high traffic demand
during busy hour at aggregation point. This solution would satisfy link requirements for
extreme cases; however, when considering the larger amount of eNBs, such situation is
highly improbable.
61/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
When considering more realistic approach of uneven traffic load distribution in time
across the network, one can observe (Figure 25) that the load at aggregation point is
quite evenly spread out. The reason behind this is that the busy hours of multiple eNBs
occur at different time of a day:
Figure 26 Difference between even and uneven traffic load distribution at aggregation
point.
Equation 53
62/63
LDF
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
A
B
where:
A [kbps] traffic load at aggregation point for even load distribution case,
B [kbps] traffic load at aggregation point for uneven load distribution case,
LDF is applied for link capacity estimation at the aggregation point and is used in
dimensioning process for accounting the impact of traffic distribution over the time. As
the LDF factor indicates, how the traffic load from particular eNBs can be decreased, it
should be applied before front mile calculations and multiplexing gain calculations.
Therefore, for dimensioning cases with Load Distribution Factor applied, the input
parameter: traffic load per PHB queue presented in Equation 50 is re-calculated:
Equation 54
PHB ( Service )
eNB
i 1
LDF
63/63
Dimensioning Guideline
For internal use only
APPROVED, 09.05.2013
Abbreviations
AMBR
BTS
DRB
eNB
EPC
ISD
LDF
LTE
MBR
MDErlang
MME
MTU
O&M
PDCP
PDN GW
PHB
QCI
QoS
RRC
RTP
SAE
SAE-GW
SCTP
SYNC-E
TCP
ToP
UDP
WFQ