User Plane and Control Plane Separation

Framework for Home Base Stations

 Zhaojun Li  Mick Wilson

(Manuscript received June 30, 2009)

Home base stations (HBSs), which are considered to be useful extensions to

mobile operators’ networks, are a promising solution for many issues in mobile
communication networks, such as coverage, capacity, and cost. They deliver a great
user experience for services in the home or in office buildings using customers’
broadband access for backhaul. However, connecting to the mobile operators’
networks via the Internet brings uncertain delay and hence unreliability into the HBS
access architecture, especially the impact on the handover between an HBS and the
surrounding base stations. In this paper, we investigate the handover management
performance in a typical HBS deployment scenario by analysing the handover
procedure durations. A user plane and control plane separation framework is
proposed to support uncoordinated HBS integration with the mobile network
through unmanaged, generic Internet protocol networks.

1. Introduction Access Network). A typical HeNB deployment

A home base station (HBS), sometimes also scenario involving E-UTRAN is illustrated in
called a femtocell, is considered to be a promising Figure 1. The HeNB interconnects with the
solution for many issues in mobile communication LTE Evolved Packet Core (EPC) over a fixed-line
networks, such as coverage, capacity, and cost. broadband access network. Similar to the macro
HBSs deliver a great user experience for services evolved node Bs (eNodeBs or eNBs; hereinafter,
in homes or in office buildings, at very low cost, we use eNBs)1),2), the HeNBs are also connected via
using customers’ broadband access (such as the S1 interface to the EPC. More specifically, the
via digital subscriber line or cable television) HeNB is connected to the Mobility Management
for backhaul. Intensive discussion of HBSs is Entity (MME) via the S1-MME interface, which
underway in standardisation bodies such as the provides the control functions for Idle mode user
3rd Generation Partnership Project (3GPP) for equipment (UE) reachability and Active mode
both the 3G HBS (called Home Node B [HNB]) UE handover support. Furthermore, the HeNB
and the LTE HBS (called Home evolved node B is connected to the Serving Gateway (S-GW) via
[HeNB]), where 3G stands for third generation the S1-U interface. In addition, the eNBs (both
and LTE stands for Long Term Evolution. In macro eNBs and HeNBs) may be interconnected
this paper, we concentrate on HBSs for LTE with each other via the X2 interface.
networks, i.e., HeNBs. In a typical HeNB deployment scenario,
The 3GPP’s work item on LTE is extra delay (due to the Internet link) is
E-UTRAN (Evolved UMTS [Universal Mobile introduced into the links between the MME
Telecommunications System] Terrestrial Radio or S-GW and HeNBs and between HeNBs and

FUJITSU Sci. Tech. J., Vol. 46, No. 1, pp. 79–86 (January 2010) 79
Z. Li et al.: User Plane and Control Plane Separation Framework for Home Base Stations

MME
EPC
S-GW
MN NM
GW S1-U S1
MME
Internet Resource
S1-MME reservation
X2
Fixed line Source Target
X2 eNB eNB
GW
NU UN
E-UTRAN
- eNB
HeNB
Phone UE
TV
Computer Synchronisation

MN, NM, NU, UN: refer to Table 1.

Figure 1 Figure 2
Typical HeNB deployment scenario. Configuration for intra-LTE handover mechanism.

eNBs. Connecting to the operators’ networks Table 1

CP delay budget in E-UTRAN.
via the Internet introduces uncertain delays and
Delay
hence unreliability into the HNB/HeNB access Abbr. Description
budget
architecture. MN
Transmission from MME to eNB (7 ms) and
9 ms
processing in eNB (2 ms)
In the work described in this paper,
One-way transmission from eNB to MME
we investigated the handover management NM
(7 ms) and processing in MME (2 ms)
9 ms

performance in a typical HNB/HeNB deployment NU

One-way transmission from eNB to UE
4 ms
(2 ms) and processing in UE (2 ms)
scenario by analysing the handover procedure
One-way transmission from UE to eNB
durations. A user plane (UP) and control plane UN
(2 ms) and processing in eNB (2 ms)
4 ms

(CP) separation framework is proposed to RR Resource reservation in eNB 5 ms

support uncoordinated HNB/HeNB integration Sync Synchronisation time at UE 20 ms
with the mobile network through unmanaged,
generic Internet protocol (IP) networks. The rest
of paper is organised as follows. The related CP 2.1 Delay budget
delay budget and the message flow charts for the 2.1.1 Delay budget in E-UTRAN (macro eNBs)
handover procedures, as well as the handover The configuration for an intra-LTE handover
delays for different scenarios are analysed in mechanism in shown in Figure 2. The intra-LTE
Section 2. The proposed UP/CP separation handover is a UE-assisted network-controlled
framework is introduced in Section 3. The handover. Here, part of the handover command
conclusions and main benefits of the proposed comes from the target eNB and is transparently
framework are summarised in Section 4. forwarded to the UE by the source eNB, the
source eNB passes all necessary information to
2. Handover duration analysis the target eNB to prepare the handover, the UE
To investigate the handover management synchronises with the target cell, and the target
performance, we analysed the following handover eNB sends a PATH SWITCH message to the
scenarios in terms of the S1 handover procedure MME to inform it that the UE has changed cell.
duration: We investigated the CP delay budget in
• LTE macro -> LTE HeNB E-UTRAN. The applied one-way transmission
• LTE HeNB -> LTE macro times and the associated processing times are
• LTE HeNB -> LTE HeNB summarised in Table 1. We assumed that the

80 FUJITSU Sci. Tech. J., Vol. 46, No. 1 (January 2010)

 Z. Li et al.: User Plane and Control Plane Separation Framework for Home Base Stations

processing times in the UE, eNB, MME, and 2.2 Intra-LTE S1 handover
S-GW are 2 ms each. The one-way transmission The intra-LTE S1 handover procedure
delay for the downlink from eNB to UE and where MME is involved is defined in ref. 2). As
uplink from UE to eNB were assumed to be 2 ms shown in Figure 3, during the S1 handover
each. We also assumed that the synchronisation procedure, the handover preparation phase
time at the UE is 20 ms. is initiated by the Source eNB by sending a
Note that in this study, we aimed to analyse
the handover procedures on the CP. Therefore, Table 2
the values were chosen to be higher than the Delay budget in HeNB scenarios.
Abbr.
UP latency values found in Reference 1). All the Description Delay budget

conclusions made later on are also valid for lower MH

Transmission from MME to HeNB
(Dinternet ms) and processing in (Dinternet + 2) ms
values. HeNB (2 ms)
One-way transmission from HeNB
HM to MME (Dinternet ms) and processing (Dinternet + 2) ms
2.1.2 Delay budget in HeNB scenarios in MME (2 ms)
In a typical HeNB deployment scenario, One-way transmission from HeNB
extra delay (due to the Internet link) is introduced HU to UE (2 ms) and processing in UE 4 ms
(2 ms)
into the links between the MME and HeNBs, and
One-way transmission from UE to
between HeNBs and eNBs. The delay budget in UH HeNB (2 ms) and processing in 4 ms
HeNB scenarios is given in Table 2, assuming HeNB (2 ms)

that the average one-way Internet delay is Dinternet RR Resource reservation in HeNB 5 ms
Sync Synchronisation time at UE 20 ms
(ms).

UE Source eNB Target eNB MME S-GW

0. Area restriction provided

1. Measurement control
UL allocation Legend
L3 signalling
2. Measurement report
3. Handover L1/L2 signalling
decision 4. Handover required
5. Handover request
6. Admission control
7. Handover request Ack.
DL allocation 8. Handover command

9. RRC connection config 10. SN status transfer

Detach from old cell
Synchronize to new cell
Synchronisation

UL allocation

11. RRC connection config complete 12. Path switch request 13. UP update request

14. Switch
downlink path
17. UE context release 16. Path switch response 15. UP update response
DL: Downlink
18. Release
resource RRC: Radio resource control
SN: Sequence number
UL: Uplink
Figure 3
Handover procedure with MME involvement (S1 handover).

FUJITSU Sci. Tech. J., Vol. 46, No. 1 (January 2010) 81

Z. Li et al.: User Plane and Control Plane Separation Framework for Home Base Stations

HANDOVER REQUIRED message to the MME Case 2: DS1 = NM + MN + 2HM + MH + NU + UH

to initiate the handover via the S1 interface, and + RR + Sync
it is finished upon reception of the HANDOVER Case 3: DS1 = HM + MH + 2NM +MN + HU + UN
COMMAND message in the Source eNB. The + RR + Sync
handover execution phase mainly includes the Case 4: DS1 = 3NM + 2MN + NU + UN + RR +
procedures where the UE detaches from the Sync
Source eNB and attaches to the Target eNB. Here, HM, MH, HU, UH, RR, Sync, NM, MN,
The handover procedure is completed by “Path NU, and UN are defined in Tables 1 and 2.
Switch” procedures that allow the core network
to update the UP paths with the S-GW. 2.2.2 Numerical results
The effect of Internet delay on the handover
2.2.1 Handover procedure duration procedure duration was investigated. The
In this study, we examined four handover durations for four handover cases are shown in
cases to investigate the S1 handover procedure Figure 4.
duration (Table 3). In case 4, which is treated as With increasing Internet delay, the duration
a reference case for comparison, handover takes for S1 handover between macro eNBs remained
place between two macro eNBs. unchanged, while the duration increased in all
We also assumed that the S1 handover the other three handover cases involving HeNB.
procedure duration starts when the Source eNB In general, S1 handover duration is greater when
sends out “Handover Required” and ends when a UE is handed over to an HeNB than to a macro
the MME receives “Path Switch Request”. eNB. The worst case occurs when the handover
S1 handover duration is between HeNBs.
Case 1: DS1 = 3HM + 2MH + HU + UH + RR +
Sync 3. UP/CP separation framework
In the light of the above observations, we
Table 3 can see that in general the performance of LTE
Handover cases. HeNBs in terms of latency is not reliable. It is
Target eNB
Source eNB HeNB Macro eNB particularly critical for control signalling that
HeNB case 1 case 2 maintains the communication between HeNBs
Macro eNB case 3 case 4 and the operator’s network. To tackle this issue,

HeNB --> HeNB Macro eNB --> HeNB

HeNB --> Macro eNB Macro eNB --> Macro eNB
600
S1 handover duration (ms)

500

400

300

200

100

0
10 20 30 40 50 60 70 80 90 100

Internet delay (ms)

Figure 4
S1 handover procedure durations.

82 FUJITSU Sci. Tech. J., Vol. 46, No. 1 (January 2010)

 Z. Li et al.: User Plane and Control Plane Separation Framework for Home Base Stations

we propose a UP/CP separation framework for for the proposed architecture is the route
the LTE HeNB. split between the HeNB’s CP and UP. In this
architecture, we propose that the HeNB provides
3.1 Overview the S1 CP functionalities via the relay eNB (e.g.,
The LTE HeNB functional architecture by microwave transmission). As shown in the
is shown in Figure 5. The key consideration figure, the S1-MME interface between HeNB and
MME is relayed through an eNB, while the S1-U
interface between HeNB and S-GW is set up
eNB S1-MME through the public network (e.g., the Internet).
HeNB MME
(relay)
The UP protocol architecture for S1-U3) is
shown in Figure 6 (a), where a GTP-U tunnel
(GPRS [general packet radio service] tunnelling
Transit IP S1-U
access network S-GW
protocol for UP) is set up between HeNB and
S-GW.4) For security reasons, a secure IPSec
Figure 5
(IP security) tunnel is set up between the HeNB
LTE HeNB architecture with UP and CP separation. and the core network for mutual authentication,

Uu S1-U

Application

PDCP PDCP GTP-U GTP-U

UDP UDP

RLC RLC Remote IP IP

IPSec IPSec

MAC MAC Transport IP Transport IP Transport IP

PHY PHY Access layers Access layers Access layers

UE LTE HeNB Generic IP network S-GW

(a) UP protocol architecture

Uu S1-MME

NAS NAS

RRC RRC S1-AP S1-AP

PDCP PDCP SCTP SCTP

RLC RLC IP IP

MAC MAC MAC MAC

Relay
PHY PHY PHY PHY

UE LTE HeNB eNB MME

(b) CP protocol architecture

MAC: Media access control S1AP: S1 application part

NAS: Non-access stratum SCTP: Stream control transmission protocol
PDCP: Packet data convergence protocol UDP: User datagram protocol
PHY: Physical layer Uu: Interface between UE and LTE HeNB
RLC: Radio link control

Figure 6
LTE HeNB UP/CP protocol architecture.

FUJITSU Sci. Tech. J., Vol. 46, No. 1 (January 2010) 83

Z. Li et al.: User Plane and Control Plane Separation Framework for Home Base Stations

encryption, and data integrity. and that the one-way transmission delay between
The CP protocol architecture for S1-MME5) the HeNB and the MME via the relaying eNB is
is shown in Figure 6 (b), where a macro eNB 10 ms.
relays all the signalling messages between the We compared the S1 handover procedure
HeNB and an S-GW. The method by which the duration in the proposed framework with that in a
HeNB links to a relaying macro eNB is outside typical deployment scenario. The results, shown
the scope of this paper. in Figure 7, compare S1 handover durations for
the worst case “handover between HeNBs” for a
3.2 Performance evaluation typical HeNB architecture and the proposed CP/
The performance of the proposed framework UP separation framework.
was evaluated by analysing the S1 handover In general, compared with the existing
procedure duration for the handover cases listed mechanism, the proposed CP/UP separation
in Table 3. The delay budget in the proposed framework demonstrates an obvious advantage
framework is summarised in Table 4. We in terms of S1 handover procedure duration.
assumed that the one-way transmission delay With increasing Internet delay, the S1 handover
between the HeNB and its relaying eNB is 5 ms duration increases in the typical HeNB
deployment scenario whereas it remains the same
in the proposed framework. In general, the S1
Table 4 handover duration is greater in a typical HeNB
Delay budget in CP/UP separation framework.
deployment than in the proposed framework.
Delay
Abbr. Description
budget With increasing Internet delay, the gap between
MH`
Transmission from MME to HeNB via eNB
12 ms
these two widens significantly.
(10 ms) and processing in HeNB (2 ms)
One-way transmission from HeNB to MME
HM` via eNB (10 ms) and processing in MME 12 ms 4. Conclusion
(2 ms) In this paper, we discussed the S1 handover
HU
One-way transmission from HeNB to UE
4 ms procedure duration in different HeNB handover
(2 ms) and processing in UE (2 ms)
One-way transmission from UE to HeNB
scenarios. In a typical HeNB deployment
UH 4 ms
(2 ms) and processing in HeNB (2 ms) scenario, HeNB interconnects with the EPC
RR Resource reservation in HeNB 5 ms over a fixed-line broadband access network and
Sync Synchronisation time at UE 20 ms through the Internet. Extra delay (due to the

Existing Proposed
600
Handover duration (ms)

500
400
300
200
100
0
10 20 30 40 50 60 70 80 90 100

Internet delay (ms)

Figure 7
Comparison of S1 handover durations between HeNBs based on existing and
proposed approaches.

84 FUJITSU Sci. Tech. J., Vol. 46, No. 1 (January 2010)

 Z. Li et al.: User Plane and Control Plane Separation Framework for Home Base Stations

Internet link) is introduced into the links between References

S-GW/MME and HeNB, and between HeNB and 1) 3GPP TSG RAN, TR R3.018 (v1.0.0): Evolved
UTRA and UTRAN; Radio Access Architecture
eNB/HeNB. and Interfaces (Release 7).
The numerical results show that with 2) 3GPP TSG RAN, TS 36.300 V8.8.0: E-UTRA
and E-UTRAN Overall description; Stage 2
increasing Internet delay, the S1 handover (Release 8).
duration increases in handover cases that 3) 3GPP TSG RAN, TS 36.401 V8.5.0. Evolved
Universal Terrestrial Radio Access Network
involve an HeNB. In the case of S1 handover, (E-UTRAN); Architecture description (Release 8).
the handover duration is greater for handover to 4) 3GPP TSG SSA, TS 23.401 V8.2.0: General
Packet Radio Service (GPRS) enhancements
an HeNB than to a macro eNB. The worst case for Evolved Universal Terrestrial Radio Access
(longest duration) occurs when the handover is Network (E-UTRAN) access Service description;
Stage 2 (Release 8).
between HeNBs. 5) 3GPP TSG RAN, TS 36.413 V8.5.0. Evolved
To guarantee reliable communication Universal Terrestrial Radio Access (E-UTRA); S1
Application Protocol (S1AP) (Release 8).
in terms of the control signalling exchange
between HeNBs and MMEs, a UP/CP separation
framework for the HeNBs was proposed. The
performance evaluation results show that this
framework demonstrates an obvious advantage
in terms of S1 handover procedure latency
compared with the existing mechanism.
In general, the proposed architecture
provides the following benefits for supporting
uncoordinated HeNB integration into the
mobile network through unmanaged, generic IP
networks:
- Reliable CP signalling delivery via a macro
eNB relay. This is particularly important for
femtocell enterprise deployment scenarios,
where reliability is crucial for business
customers.
- Flexible UP data delivery via a public
network (e.g., the Internet) at very low cost.
Using public transportation for UP data
delivery supports various application
services in homes, on campuses, or in offices
at very low cost. It also has great potential
to create a totally new set of services in a
“connected home”, “connected classroom”, or
“connected office”.
- The proposed framework does not require the
introduction of a new node into an operator’s
network, so the effect on existing standards
is minimal. This is important in order to
protect the investments of operators.

FUJITSU Sci. Tech. J., Vol. 46, No. 1 (January 2010) 85

Z. Li et al.: User Plane and Control Plane Separation Framework for Home Base Stations

Zhaojun Li Mick Wilson

Fujitsu Laboratories of Europe Ltd. Fujitsu Laboratories of Europe Ltd.
Dr. Li received an M.Sc. degree in Mr. Wilson received a B.Sc. degree in
Computer Engineering from Huazhong Computer Science from Staffordshire
University of Science and Technology, University in 1976. He joined FLE in
China in 1996 and a Ph.D. degree in 1997 and is currently a manager within
Telecommunications from the University the Network Systems Research division
of Surrey, UK in 2005. Before joining focusing on wireless broadband access.
FLE in 2007, she had several years of Before joining Fujitsu, he was with
industrial and academic experience. Plessey Office Systems and in 1987
As a senior engineer in Motorola, she joined Siemens Roke Manor Research
participated in several projects, including CDMA 2000 BTS working on distributed systems and intelligent networks and led
development in the USA (1999–2000) and GSM system resource a number of European collaborative projects. He represents
management function development in the UK (2001–2002). As Fujitsu in 3GPP standardization meetings focusing on radio
a research fellow (2003–2007) at the Centre for Communication network access technologies, previously for UMTS and LTE
Systems Research, University of Surrey, UK, she participated and currently for LTE Advanced. In addition, he serves as an
in several European research projects and UK-funded industrial chairman of a major UK research activity (Mobile-
projects. She is the author of several patents and research VCE) investigating flexible networking. He also represents FLE
publications in international journals and conferences. Her main on the board of the Digital Communications Knowledge Transfer
research interests include mobility management (and resource Network, a UK-government-funded organisation looking at
management), location management, quality of service, 3GPP Communications R&D strategy for Digital Britain.
mobile systems, multimedia communications, heterogeneous
communications, and IP networks. She is a Senior Researcher
at Fujitsu Laboratories of Europe (FLE).

86 FUJITSU Sci. Tech. J., Vol. 46, No. 1 (January 2010)