Beruflich Dokumente
Kultur Dokumente
Cellular Backhaul
The Business Case for PacketBased Radio Access Networks
By Gaby Junowicz Director, Product Strategy and Business Development RAD Data Communications
Abstract
The expected bandwidth upsurge associated with the rollout of new mobile broadband services and NG mobile standards, shifts the transport burden to the RAN backhaul the connections between the base stations and their controllers. While in the past a small number of E1/T1 links were sufficient to service 2G and some 3G traffic, the bandwidth requirements of HSPA, 3GPP's UTRAN LTE, and Mobile WiMax are of a different scale altogether, leading cellular operators to reevaluate their backhaul strategy. In addition, operators are challenged with the need to simultaneously support multiple generations and technologies. They need to ensure service continuity for legacy 2G and 3G traffic, as well as to guarantee service quality for emerging data-intensive and rich-media applications. Moreover, the related (and substantial) costs they are about to incur cannot be passed on to the mobile users, who expect more bandwidth and higher quality, at lower service rates. The key to successfully control the growth in mobile traffic is in decoupling bandwidth from cost, by optimizing transport capacity and migrating to costeffective backhaul technologies as Carrier Ethernet, MPLS and DSL. This paper examines the implications of the IP backhaul evolution on operations and services, reviews the various tactics operators can employ in realizing a balanced transition, and provides tools to assess their financial benefits.
Contents
The Bandwidth Surge ................................................................................ 2 Decoupling Capacity and Cost .................................................................... 4 Reducing Costs with Packet Transport ........................................................ 5 Compelling Market Drivers .................................................................... 5 Microwave Backhaul ............................................................................. 7 Cellular Backhaul Performance Requirements .............................................. 8 Backhaul Network Performance............................................................. 8 Transport Elements Performance .......................................................... 8 Pseudowires ........................................................................................ 9 Alternatives for Backhaul Cost Reduction ................................................. 10 Aggregation and Overbooking ............................................................. 10 RAN Gateways ................................................................................... 13 Aggregation over Ethernet .................................................................. 14 Leveraging Wholesale Services ............................................................ 15 HSDPA Offload........................................................................................ 17 Conclusion .............................................................................................. 20
Figure 1: The evolution of mobile services In addition to the exponential increase in bandwidth requirements per user, subscriber numbers are also climbing Infonetics, for example, predicts 3 billion cellular end-users by 2009, up from 2 billion in 2005. To accommodate the continuously expanding subscriber base and new service rollouts, mobile backhaul connections are expected to grow 69 percent, from just under 2 million in 2005 to 3.3 million in 2009.
With base station coverage expanding to support the new, bandwidth-guzzling services, the transport burden shifts to the RAN backhaul networks, where traffic is transmitted between the base stations/Node Bs and the base station controllers or radio network controllers. As most base station connections have traditionally consisted of only a few E1/T1 links, it seems fairly safe to assume that the expected bandwidth explosion will create a bottleneck in cellular backhaul. In the past, one or two E1/T1 links were sufficient to support 2G voice and SMS data traffic. Initial 2.5/3G launches involved applications with bandwidth consumption of up to 348 kbps, a rate that was still addressable by E1 circuit addition and 2G/3G traffic aggregation. HSDPA requirements, however, are of a different scale altogether, with a single connection offering theoretical peak throughput of 14.4 Mbps per user, increasing the number of required connections anywhere from eight to 16 E1/T1s for a single 3G cellular site. As a result, cellular operators anticipate tripling backhaul capacity at urban cell sites over the next few years. According to Heavy Reading 2007 reports, the average forecasted backhaul capacity for a typical urban cell site at the end of 2010 is 25 Mbps and even 40 Mbps in traffic-intensive business districts (see figure 2).
The expected bandwidth upsurge does not stop here, as NG mobile standards and services such as 3GPP's UTRAN Long Term Evolution (LTE), Mobile WiMax and 3GPP2's Ultra Mobile Broadband (UMB), all pursue data connectivity rates beyond 100 Mbps, and potentially even 200 Mbps, per cell site. This outlook mandates re-examination of backhaul strategy by cellular providers, as no economies of scale can be implemented in achieving the required RAN capacity by using legacy leased lines or PDH microwave. Simply put, reaching the desired bandwidth by adding E1 or T1 links to cell sites is not economically feasible. To complicate matters even further, cellular operators and service providers are challenged with the need to simultaneously support multiple generations and technologies. They need to ensure service continuity for 2G voice and prepare for a reliable and economically sound service launch of 3G and beyond, while following the mobile telecom industry roadmap towards the convergence of fixed and mobile networks (FMC).
These cost pressures compel mobile operators to seek ways to dramatically reduce their cost per bit in order to meet profitability targets, at the same time creating an opportunity for transport providers to pursue new revenue generators and offer alternative backhaul services. The key to successfully control the growth in transport capacity lies in separating bandwidth from cost, by using two parallel routes: capacity optimization with efficient traffic management and backhaul migration to cost-effective alternative technologies such as Carrier Ethernet, MPLS and DSL.
Without a doubt, however, the primary driver for migrating to Ethernet backhaul is its costeffectiveness. While E1/T1 price levels are directly proportioned to the increase in bandwidth, Ethernet connections are typically offered in x10 increments (i.e. 10 Mbps, 100 Mbps, 1,000 Mbps, etc.) and are charged accordingly. Addition of bandwidth within these increments does not require a connection upgrade and has no effect on pricing. Figure 3 illustrates these differences.
Ethernet backhaul therefore allows substantial cost-per-bit savings. A simple cost comparison between a 10 Mbps Ethernet service and the equivalent E1 leased lines in urban areas shows annual savings, in transport costs alone, of over 17 million for every million subscribers:
Leased Line (E1) Monthly backhaul costs per 3G Node B Monthly backhaul costs per 1 million subscribers Annual backhaul costs per 1 million subscribers Annual savings per 1 million subscribers 1,250 x 2,860 Node Bs 2 = 3,575,000 3,575,000 x 12 = 42,900,000 5 x 250 = 1,250 1
As can be seen in figure 3, service pricing turns favorable to Ethernet at 3-4 E1s. However, as Ethernet service offerings begin at 10Mbps, the comparison was based on a 5 E1equivalent According to RAD estimates and based on operator feedback, a typical 2G BTS serves an average of 1,500
subscribers. Also, there are three times more 3G base stations than 2G BTSs per covered area, with 3G coverage rate at app 70%. Based on these parameters, a typical Node B serves (1,500/3) x 0.7 = 350
subscribers. Therefore, 1 million subscribers within the same geographical area require 1,000,000/350 = 2,860 Node Bs 6 2008 RAD Data Communications Ltd.
Although the above cost model is based on E1 connections, similar conclusions can be drawn for the North American market. As can be seen in table 1, a 10 Mbps Ethernet service provides five times the bandwidth of a single E1, for three times the cost. It should be noted that these figures are based on current Ethernet pricing, which is still in its early, pre-mass deployment phase. In contrast, E1/T1 prices have already been considerably discounted, especially for large operators. As incumbents and third party service providers begin extending Carrier Ethernet backhaul services to cell sites, and as early adopters among cellular operators complete their self-built Ethernet networks, we can expect price, coverage and performance to become increasingly competitive, bringing the aboveindicated cost reductions to even more significant levels. These principles apply to Last Mile segments as well as to Metro networks.
Microwave Backhaul
The above comparison refers to copper links, which make 89% of cell site connections in North America. The bandwidth bottleneck concern, however, is also relevant to TDM microwave networks, serving almost 70% of European cell sites and 40% of the cell sites in Asia, primarily in India. Here too, point-to-point Ethernet microwave offers better cost per Mbps, an advantage that becomes even more pronounced with the decline in license fees and equipment pricing that is associated with the growing use of the technology:
As can be seen in table 2, Ethernet radios introduce significant savings in CapEx when upgrading backhaul capacity to 16 E1 links or more, providing three times the bandwidth for a 30% cost increase. These principles also apply to T1 connections and prices.
Backward- and future-compatibility with all generations and types of mobile traffic, including GSM, CDMA2000, Mobile WiMAX and HSPA, as well as the ability to convert legacy traffic to packets Comprehensive network management and end-to-end performance monitoring, including OAM for automatic fault isolation Small footprint access/backhaul equipment for radio sites and BSC/RNC locations It is also important to note, that one of the main concerns for communications providers in moving to packet-based RAN is their need to amortize their existing TDM and ATM installedbase. Transport elements have therefore a major role in successfully leveraging such equipment in delivering Next-Gen services. This long list of requirements necessitates a sophisticated technological solution that is tailored to packet-based environments precisely the premise on which pseudowire emulation edge-to-edge was developed, and later expanded to pseudowire emulation endto-end.
Pseudowires
Pseudowire technology is a widely accepted enabler for migrating cellular backhaul to packet-switched transport, as it allows the delivery of legacy and HSPA traffic over nondeterministic, statistical networks. Pseudowire emulation edge-to-edge (PWE3) carries leased lines and ATM services over packet media by creating virtual tunnels across Ethernet, IP or MPLS networks, to allow seamless connection of GSM and UMTS network elements. These transport tunnels enable transparent delivery of real-time traffic across packet networks without distortion, avoiding the need to translate signaling data and ensuring that clocking and synchronization criteria are met.
Pseudowire solutions are particularly suited to cellular backhaul, not only because they enable transparent delivery of TDM and ATM circuits, but they also apply to any transport, including DSL, metro Ethernet networks, MPLS, and Ethernet microwave. The comprehensive set of standards defined by IETF, ITU, MFA, and MEF facilitates easy implementation of the technology and creates the foundation for future interoperability.
10
The number of R99 E1 lines at the aggregation ingress will be N x n(E1 R99), where N is the number of aggregated cell sites, and n(E1 R99) is the number of E1 lines required for voice and real-time traffic in each cell site . Similarly, the number of HSDPA E1 lines at aggregation ingress will be N x n(E1 HSDPA), where N is the number of aggregated cell sites, and n(E1 HSDPA) is the number of E1 lines required for data traffic in each cell site. Given the different bandwidth requirements for these services, n(E1 R99) n(E1 HSDPA). The number of egress E1 links, after applying statistical multiplexing to the ingress traffic, can be extracted from the equation:
As shown in table 5, statistical multiplexing can reduce the required capacity at aggregation egress to half of the ingress bandwidth.
11
Data: # E1 per Node B - 3G R99: # E1 per Node B - HSDPA: Total E1s per Node B: Voice (R99) occupancy: HSDPA occupancy: # Node Bs per aggregation point: Calculations: # E1 R99 Ingress # E1 HSDPA Ingress Total E1s at Ingress # E1 R99 Egress (with overbooking) # E1 HSDPA Egress (with overbooking) Total E1s at Egress Total Savings per aggregation point (%) 3x1=3 3 x 4 = 12 15 3 x 0.6 = 1.8 12 x 0.4 = 4.8 7 53% 1 4 5 60% 40% 3
Such savings translate to millions of Euros/Dollars in reduced backhaul expenses annually. Table 6 summarizes the economics of extending aggregation and overbooking capabilities to the access, for one million subscribers. Although referring to European E1 links and prices, these principles apply to T1 connections as well.
12
Regular Operation # 3G/3.5G Node Bs (UMTS/HSDPA) per 1 Million Subscribers # Hub sites (@ 3 Node Bs/Hub Site) # E1s Required per Hub Site Total # of E1s Required per 1 Million Subscribers Annual Cost (@ 250/E1/month) Annual Savings per 1 Million Subscribers 2,860 3 950 15 14,250 42,750,000
RAN Gateways
Mobile operators planning on migrating their services to packet backhaul can do so by either deploying IP-ready base stations or by connecting existing TDM and ATM base stations and controllers to multiservice pseudowire RAN gateways. These two routes are not necessarily mutually exclusive, however, they involve several aspects that require consideration: Cell site equipment with native Ethernet interfaces simplifies network architecture but raises issues of availability, compatibility and cost. For example, initial massive deployments of IP-ready base stations are not expected before 2009 or even 2010, with packet technology supported only in new product releases. This may coincide with many operators individual migration schedules; others may wish to initiate new service rollouts and backhaul cost reductions as soon as possible, as well as avoid the need to phase out existing infrastructure, by deploying multiservice gateways to connect their legacy cell sites to new transport networks. Another issue is the transport of real-time TDM traffic. Pseudowire-enabled multiservice gateways implemented at cell sites and at aggregation sites allow direct backhaul of 2G TDM traffic and 3G ATM traffic over the same packet transport.
Delay-sensitive service requirements can also be addressed by starting the PW session right off the Node B, with the attendant upgrade considerations. Provided they are fully equipped to meet the various requirements for packet transport elements, including support capabilities for multi-generation colocation, pseudowire multiservice RAN gateways are instrumental in minimizing the economic impact of the massive bandwidth expansions at hand, allowing even substantial savings in capital and operating expenses.
Even an aggressive discount schedule for leased E1 lines would still position their cost an order of magnitude above other alternatives. As can be seen in table 7, a 1,000 Mbps (1 Gbps) Ethernet service, already priced in some areas below 30,000 per year, enables 650% more throughput than an STM-1 line, for a 30% cost increase. It should be noted that these larger throughput alternatives require fiber access to the aggregation/hub site. Still further cost reductions can be realized by replacing traditional leased lines with low cost wholesale DSL services, using cell site gateways to extend Ethernet services over widely available infrastructure.
14
Connection
Available Bandwidth
T1 E1 ADSL2+ Wholesale 4
The cost of wholesale xDSL services for operators is typically twice as high as the cost of residential services, due to enhanced QoS attributes Common downstream bit-rate at 6 kilofeet/1.8 km, 26AWG 15
A variety of DSL flavors present a choice of bandwidth and symmetry alternatives to meet various service requirements. As an asymmetric service, ADSL2+ offers high downstream bit-rate but only 1-2 Mbps upstream, rendering it unsuitable for mobile voice but ideal for mobile broadband data services. In addition, a dramatic drop in line speed at long distances dictates that cell sites are located within a 2 km reach of the local DSLAMs, in order to fully benefit from this service. SHDSL, on the other hand, is offered at higher price points but provides longer reach, better line quality and supports symmetric services including voice applications. SHDSLs relatively low throughput of 4 Mbps has been significantly enhanced to mid-band Ethernet rates with copper bonding protocols such as ATM (IMA), M-pair G.SHDSL or EFM, which enable higher bandwidth/longer reach and dynamic rate-adaptive capabilities.
VDSL2 is a relatively new symmetrical service, offering theoretical rates of up to 100 Mbps, but is limited to copper lengths of up to a few hundreds of meters. VDSL2 will become necessary in short loop applications as higher speed broadband services become available.
Table 9 summarizes the available DSL services and their use in cellular backhaul:
SHDSL Traffic Bandwidth Symmetrical Low (Mid-band through bonding) Reach Cell sites/ Services Long - 2G/3G Voice - 3G/HSDPA Data + Voice
16
Figure 6: DSL services for cellular backhaul As can be seen in figure 6, multiservice cell site gateways incorporating built-in DSL support eliminate the need for additional DSL modems, and enable operators to benefit from economical xDSL transport without unnecessary investments.
HSDPA Offload
The upshot of the IP RAN migration process is multi-generation traffic delivery over a unified, packet switched backhaul end-to-end, as illustrated in figure 7:
17
As an interim solution, many operators are opting for a hybrid approach that allows them to maximize existing assets. This approach, known as HSDPA Offload, involves the use of two traffic-dedicated transport paths: Real-time voice and video are carried over existing SDH/SONET or ATM backbones, while UMTS and HSDPA data traffic is delivered over a PSN, using pseudowire encapsulation. HSDPA Offload offers a gradual transition, in which operators and service providers can begin capitalizing on packet transports economical benefits while continuing to utilize legacy infrastructure and ensuring stringent QoS for delay-sensitive services.
Hybrid transport is especially cost-effective in dense urban areas, where xDSL lines can be used to carry the data traffic to the packet network, allowing a number of cell sites to share common DSLAM elements.
18
Table 10 compares the cost of transporting 20 Mbps from the aggregation point, for TDMonly backhaul and a hybrid E1/ADSL2+: Backhaul # E1 links # ADSL2+ links Total E1 cost/month @ 250/month () Total ADSL2+ cost/month @ 60/month () Total monthly cost () Total annual cost for 60,000 6 cell sites ( millions) Savings 2,500 2,500 600 TDM only 10 TDM/16 Mbps ADSL2+ 2 1 500 60 560 134
466,000,000 (78%) Table 10: Cost savings using a hybrid E1 and DSL backhaul
A cost comparison between T1-only transport and T1/ADSL2+ hybrid draws similar conclusions, as illustrated in table 11.
Backhaul # T1 links # ADSL2+ links Total T1 cost/month @ $250/month ($) Total ADSL2+ cost/month @ $60/month ($) Total monthly cost ($) Total annual cost for 60,000 cell sites ($ millions) Savings -
$586,000,000 (75%) Table 11: Cost savings using a hybrid T1 and DSL backhaul
Different DSL flavors can be used to transport various services, subject to available infrastructure, accessible wholesale services, local legislation, and other country-specific conditions. For example, ADSL2+ can be used for HSDPA and VDSL2 for HSDPA/HSUPA, while G.SHDSL and SDH/ATM connections are used for high-priority business traffic. All these alternatives, however, require a cell site gateway to offload the data traffic and encapsulate it in packets using pseudowire tunnels and connections.
Conclusion
The rise in traffic volumes associated with emerging mobile broadband services requires cellular operators to re-evaluate their backhaul operations. With market conditions expected to curtail ARPU growth and current RAN infrastructure dictating a linear correlation between traffic increase and transport expenditures, operators are pursuing innovative solutions to lower the cost per bit of their WAN connections. This paper establishes the economic justification for using packet-based networks as flexible, viable alternatives to legacy cellular transport and as means to decouple bandwidth requirements from transmission costs. In addition to greatly improving price-performance results, IP-RANs are compatible with the industrys roadmap, featuring intrinsic support for future developments, such as fixed and mobile convergence and the new 3GPP standards.
20
This paper also demonstrates how operators can realize a balanced migration to an all-IP backhaul and provides tools to assess the financial benefits of pseudowire-enabled multiservice RAN gateways, statistical multiplexing and aggregation techniques, wholesale xDSL services, and HSDPA offload methodology.
With vast field experience and expertise in radio access networks, RAD enables operators to substantially increase throughput, while considerably reducing CapEx and OpEx. Leveraging cost effective transport technologies and services, RADs extensive RAN gateway portfolio supports voice and data service delivery over PSN, by incorporating standardized pseudowire capabilities with high performance clock recovery mechanisms, as well as carrier-grade, SLA-based QoS/CoS guarantees.
21
www.rad.com
International Headquarters RAD Data Communications Ltd. 24 Raoul Wallenberg St. Tel Aviv 69719 Israel Tel: 972-3-6458181 Fax: 972-3-6498250 E-mail: market@rad.com www.rad.com
North America Headquarters RAD Data Communications Inc. 900 Corporate Drive Mahwah, NJ 07430 USA Tel: (201) 529-1100, Toll free: 1-800-444-7234 Fax: (201) 529-5777 E-mail: market@radusa.com www.radusa.com
The RAD name and logo are registered trademarks of RAD Data Communications Ltd. 2008 RAD Data Communications Ltd. All rights reserved. Subject to change without notice. Version 0/2008