Anda di halaman 1dari 3

Wireless Backhaul for LTE - Requirements, Challenges and Options

G Kalimani Venkatesan, Kishor Kulkarni, Wipro Technologies, Bangalore 560100 Email: {kalimani.venkatesan.kishor.kulkarni}@wipro.com

support this inter-Base station communication requirement.


Abstract-Next Generation Broadband Wireless Technologies such as 3GPP Long Term Evolution (LTE) and WiMax offer significantly higher data rates and require suitably higher capacity backhaul networks. While some service providers have started rolling out WiMax, 3GPP LTE is expected to be standardized during 2009, and many service providers are planning to offer LTE services by 2010-2012. Apart from significantly higher speeds, LTE Base stations (eNBs) require logical full mesh connectivity due to the flat all-IP architecture. This paper explores the Wireless backhaul network infrastructure options for addressing the LTE bandwidth and connectivity challenges. The paper details the backhaul requirements for 3GPP LTE as specified in the LTE specifications. The paper analyses different architectures for the Backhaul Access and Aggregation networks. The paper proposes a logical topology model for the Aggregation network, examines its realization via Carrier Ethernet Transport and IP/MPLS, and identifies technology gaps in realizing the logical topology model. The paper concludes that while MPLS satisfies the requirements better today, given the time available for LTE evolution, native Carrier Ethernet Transport could emerge as a strong candidate for future deployments.

MME/&GW

~
i\
I

MME/5-GW

~
II

1\ 1\

{J

en )F: "\

i
I

~
\

~
I

,/" en :(J~)
I

,/i
I

i
:

"

eNS ' \ \ \

~~O/~
eNS

/ /

X2

E-UTRAN

eNS

Fig.I. LTE Backhaul- Connectivity requirement

A summary of the key requirements for LTE Backhaul is given below 1. 2. 3. 4. Higher data rates: From 100/50 Mbps DL/UL up to 1 Gbps DL between eNB and MME/SAE-GW. Flat Architecture: MMEs/ SAE-GW and eNB shall be connected in a many-to-many relationship. Low latency: Providing lower user and control plane latency [e.g. less than 5ms in ideal conditions]. Connectivity: Point-to-point (P2P), point-to-multipoint (P2MP, Multicast) and multipoint-to-multipoint (MP2MP, Broadcast) traffic between various network entities. Guaranteed QoS, multi CoS: Handle different IP traffic classes - Real-time, Non-Real time, and Mission critical. High Availability, Resiliency, Fault and Performance management

I. LTE ARCHITECTURE AND BACKHAUL REQUIREMENTS

GPP Long Term Evolution (LTE) [1] is a significant evolutionary step for UMTS in terms of capacity and architecture. LTE aims to offer a minimum of 100/50 Mbps DL/UL (Downlink/Uplink, 1 sector, 20MHz spectrum) and up to 1 Gbps DL (3 sectors, DL/UL of 300/150 Mbps per sector) per LTE Base Station (eNB). This is significant increase over 3G with peak data rates around 30 Mbps. Consequently, the existing 3G Wireless Backhaul Access network capacities need to be significantly increased to cater to LTE backhaul. The Radio Access Network (RAN) connectivity has been evolving towards a flat, all-IP architecture with reduced number of Network elements for lower latency. As seen in Fig.l. LTE defines direct connectivity ("SI") between the LTE eNB and the LTE Gateways (Mobility Management Entity-MME) without any Radio Network Controller. Hence, LTE eNB carries more intelligence for functionalities like Radio Resource Management towards distributed decision making, which requires inter-eNB logical connectivity ("X2"). Again, existing Backhaul networks do not naturally

5. 6.

Additional intrinsic requirements on the Backhaul network such as (i) Scalability (in Bandwidth provisioning for incremental rollouts, in Cell Sites), (ii) Logistical (Access Network Reach, Ease/Speed of new deployment), (iii) Migration / Upgrade paths (from existing 2G/3G UMTS/CDMA backhaul networks, as well as scaling for the future), (iv) Service continuity (support 2G/3G services and Network elements over the backhaul, TDM/ATM emulation, and Timing/Synchronization transport), all at Lower Capex/Opex and at service levels equal to or better than existing TDM based network infrastructures.

Authorized licensed use limited to: Universidad de chile. Downloaded on October 8, 2009 at 00:28 from IEEE Xplore. Restrictions apply.

II. LTEBACKHAULARCHITECTURALOPTIONS
A. Access Network Technology Options

Bandwidth, Cell-site Reach, available Upgrade paths are key criteria for evaluating the Access options. Broadband Access technologies such as (i) VDSL2 (with bonding for capacity, co-located DSLAMs at Cell Sites), (ii) PON (GPON, G EPON; Upgrades via with lOG GPON/10G GEPON, WDM PON), (iii) Ethernet over Fiber (WDM - 1 to 10 Gbps, with Carrier Ethernet Metro rings), (iv) Microwave Radios: PDHlSDH Radios or native Ethernet Framed Radios with Switched Ethernet functionality and (v) NG SONET/SDH (investment utilization, till transformation to Carrier Grade PSNs) can be considered for the LTE Access Network.
B. Aggregation Network - Technology Options

Synchronization via Packet based Timing, suit this model. - MP2MP model avoids having N tunnels from each eNB to every other eNB (N2 tunnels), and requires a single tunnel at a eNB, leaving the interconnection complexity to the network. When mapped to MEF services [2][3] (via IP/MPLS or native Carrier Ethernet Transport), P2P maps to Ethernet Virtual Private Line (EVPL) service, P2MP maps to Ethernet Virtual Private Tree (EVP-Tree) service, and MP2MP maps to Ethernet Virtual Private LAN (EVP-LAN) service. The following sections evaluate native Carrier Ethernet Transport and MPLS for realizing the logical connectivity models.
1) Carrier Ethernet Transport Realizing the connectivity models is through a mix of Provider Bridging (PB) [4]/ Backbone Bridging (PBB) [5]/ Backbone Bridging with Traffic Engineering (PBB-TE) [6] technologies, and the options are listed below in the order of suitability. Carrier Ethernet Transport evolution and challenges have been evaluated in [12].
TABLE I
REALIZING CONNECTIVITY MODELS WITH PB / PBB / PBB-TE

Depending on the user traffic and network control traffic patterns, and associated CoS requirements, appropriate connectivity models can be adopted for the Aggregation network. As in the following figure, the LTE eNB ~ Network Controller connectivity can be modeled as a set of P2P or P2MP tunnels, or a mix of both. The Inter-eNB connectivity, for Network control traffic such as handoffs, would fit a MP2MP model.

Connectivity P2P (EVP-Line)

P2MP (EVP-Tree) MP2MP (EVP-LAN)

Options [* - Not standardized] - PBB-TE with NMS or GMPLS, involving CAC, pinned paths and sub-50 msec protection via CCM. - PBB with PLSB as control plane* - PB with STPIIVL, or VLAN tunneling or cross connect. - PBB with PLSB as control plane * - PB with STPIIVL, or VLAN tunneling or cross connect (P2MP with more than one SVLAN) - PBB with PLSB as control plane * - PB / PBB with IVL + P2P PBB-TE (like VPLS with EoMPLS Pseudowires)

Fig.2. LTE Backhaul- Logical Connectivity with P2P, P2MP and MP2MP

- P2P model requires N (number of eNBs) traffic engineered tunnels. Multiple traffic classes can be supported within one tunnel via CoS identifiers. Connection oriented P2P tunnels with admission control (CAC), and Path protection (by an Active/Standby P2P tunnel pair) is ideal for meeting stringent QoS requirements. - P2MP model, which allows for bidirectional eNB ~ NC connectivity can also be used to carryall traffic classes. One P2MP tunnel can be provisioned across all eNBs, or one tunnel per set of eNBs, or one tunnel per service class. The illustration depicts two P2MP tunnels, one per service class. Multicast applications, such as Multicast Video and Network

For P2P transport, PBB-TE offers a Connection oriented model. PBB-TE addresses network/node scalability, avoids xSTP limitations and learning based forwarding with a configured or a GMPLS control plane based approach, provides resilience with preconfigured protection path and sub-50 msec switchover. However, currently, PBB-TE is limited to P2P. P2MP or MP2MP support can be derived via PLSB (Provider Link State Bridging). Future enhancements to PBB-TE for native P2MP support can be via GMPLS [7]. Also, PBB-TE today does not support Pseudowires natively [8], which is a key requirement for 3G services continuity. Instead of PBB-TE, PB and PBB can be configured to offer P2P, P2MP and MP2MP connectivity. A P2P VLAN tunnel with PB can be constructed by configuring each eNB in a unique S-VLAN, and the Network Controller attaching to multiple S-VLANs. This would impose the known VLAN scaling issues (4094) for eNB. With PBB, a P2P VLAN can be constructed without scaling limitations, by mapping the 1SID for each eNB as a unique service instance. However, with PB and PBB, it would be more common to use a Shared/Service based EVC model, with an S-VLAN shared across a set of eNBs for a specific Traffic class.

Authorized licensed use limited to: Universidad de chile. Downloaded on October 8, 2009 at 00:28 from IEEE Xplore. Restrictions apply.

3 such static Pseudowires to Dynamic P2P tunnels.


2) MPLS Being the technology of choice in Mobile core network, MPLS can logically be extended into the Aggregation. Also, MPLS is more than a mere transport, as the many services that have spawned off it (Pseudowires, L3 VPN, L2 VPN) have become indispensable. Pseudowires are indispensable to migrating the 2G TDM and 3G ATM elements to a Packet backhaul, and MPLS is the only defined PSN today for carrying Pseudowires. The table below lists the possible MPLS options and availabilities in the order of suitability.
TABLE II
REALIZING CONNECTIVITY MODELS WITH MPLS

3) Hybrid: Carrier Ethernet & MPLS


TABLE III CARRIER ETHERNET TRANSPORT AND IPIMPLS COMPARISON Carrier Ethernet Transport IPIMPLS Strengths - Possible Lower Capex/Opex. - Proven - Good for P2P with PBB-TE - Pseudowires - P2PIP2MP with TE - Stron~er 0 AM toolkit Weakness - P2MP with xSTP, no TE/QoS - Perceived complexity, - Not proven higher Capex and Opex.

Connectivity P2P (EVP-Line)

P2MP (EVP-Tree) MP2MP (EVP-LAN)

Options [* - Not standardized] - EoMPLS Pseudowires or native IP over MPLS - CAC with RSVP-TE or NMS provisioned - Protection with FRR and PW redundancy [8]* - Multi-segment PW with PW-stitching for scalability[7] * - Dry-Martini for simplification at Access * - Multicast MPLS - RSVP-TE, LDP with P2MP extensions, which doesn't require IP Multicast Routing. - FRR for P2MP MPLS-TE * - VPLS - Scaling with H-VPLS;VPLS with 802. lah/802. lad [9]*

MPLS provides P2P Traffic Engineered tunnels, setup by RSVP-TE. These P2P tunnels can be protected by fast-reroute (FRR), which can offer 50 msec restoration times. End-to-end Pseudowire path protection can also be provisioned via PW Redundancy. MPLS support for P2MP has been standardized, with P2MP extensions for RSVP-TE and LDP. With one MPLS P2MP tunnel (from NC to eNB) and multiple P2P tunnels (from eNB to NC), an EVP-TREE topology can be realized. This can be used for both Unicast and Multicast traffic. MP2MP support to communicate across eNBs can be via VPLS, or H-VPLS with 802.lad (PB) or 802.lah (PBB). - MPLS P2P solution would require tunnels as the number of eNBs (O[N]). To reduce the scalability burden, Multisegment Pseudowires with PW-stitching can be used. Each eNB/CSG gets connected via a PW tunnel till the Mobile Aggregation Gateway (Access segment), where many such PW are "stitched" together to provide a single tunnel towards the NC (Aggregation segment). This would ease the number of tunnels at the Aggregation. - The Access segment Pseudowires can also be statically configured, reducing the complexity. At the Access nodes, running a full IP routing and MPLS signaling functionality not needed. It is possible to simplify this by having static configuration for routes, MPLS labels and Pseudowire label, and the data plane encapsulation alone for Pseudowires. Again, using PW stitching, Aggregation gateway can connect

The table above compares Carrier Ethernet and IP/MPLS for the LTE backhaul. Rather than adopting single technology approaches, solutions can take advantage of each technology's strengths for specific application. An example is a Hybrid solution for providing MP2MP inter eNB connectivity. Assuming the Aggregation is PB/PBB based, the P2MP/MP2MP approach with the multicast PLSB support is only a future option. Instead, it is possible to emulate the VPLS model by having PB clouds (connecting sets of eNBs) connecting via PBB-TE tunnels to each other. Or, a first level access/aggregation network could provide PBB-TE based VPLS, interconnecting over an MPLS Aggregation Network. III. CONCLUSION This paper has analyzed the LTE backhaul requirements at Access and Aggregation. The LTE connectivity model goes beyond the existing 'single tunnel aggregation' model and requires traffic engineered P2P, P2MP and MP2MP connectivity. Between native Carrier Ethernet and MPLS, MPLS is seen as a better option in the current assessment to satisfy LTE backhaul requirements. However, innovations in Carrier Ethernet transport, and time factor for LTE deployment could bring Carrier Ethernet into contention.
REFERENCES E-UTRAN Architecture description, 3GPP TS 36.401, 3GPP specifications [online].:http://www.3 gpp.orgj [2] Ethernet Services Attributes Phase 2, MEF 10.1.1 specification, June 2008, Metro Ethernet Forum [online]. http://www.metroethemetforum.org [3] Mobile Backhaul Implementation Agreement, MEF D00065_ 003, June 2008, Metro Ethernet Forum [online]. http://www.metroethernetforum.org [4] Provider Bridges, IEEE 802.lad, draft 6, March 2008 [online] Available: http://ieee802.orgj [5] Provider Backbone Bridges, IEEE 802.lah draft, Jan 2008,[online] [6] Provider Backbone Bridging with Traffic Engineering, IEEE 802.1 Qay, March 2008, [online] [7] Don Fedyk, et aI., "GMPLS control of Ethernet PBB-TE", draft-fedykgmpls-ethernet-pbb-te-02, IETF draft [online]..http://www.ietf.orgj [8] D. Allan, et aI., "Carrying PWE3 Pseudo Wires over Provider Backbone Transport", draft-allan-pw-o-pbt-03, IETF [online]. http://www.ietf.orgj [9] Bocci, et aI., "An Architecture for Multi-Segment Pseudo Wire Emulation Edge-to-Edge" draft-ietf-pwe3-ms-pw-arch, IETF [online] [10] Muley, et aI, "Pseudowire Redundancy", draft-muley-pwe3-pwredundancy, IETF draft [online]. Available: http://www.ietf.orgj [II] A. Sajassi, et aI., "VPLS Interoperability with Provider Backbone Bridges", draft-sajassi-12vpn-vpls-pbb-interop-Ol, IETF draft [online]. [12] Kishor Kulkarni, G Kalimani Venkatesan, "Carrier Ethernet - State ofArt and Challenges Ahead"; IEEE ANTS Conference, Dec 2007. [I]

Authorized licensed use limited to: Universidad de chile. Downloaded on October 8, 2009 at 00:28 from IEEE Xplore. Restrictions apply.

Anda mungkin juga menyukai