It’s a growing concern that as HSPA networks roll out and, more importantly, as data intensive service start to be used across those networks, the existing backhaul infrastructure will become too expensive to run, with bits costing more to transport end to end across the network than operators can charge for them. What solutions are out there to help operators as they face up to this issue?
Today, many operators’ backhaul requirements from a single cell site often amount to only one E1; in some cases this even includes both 2G service and 3G service (R99 version).
However, operators are marketing HSDPA with aggressive performance metrics in response to expected threats such as WiMAX.
” height=”<% height %>” align=”right” alt=”HSDPA backhaul – Meeting the backhaul challenge” class=”articleimage” />
These marketing efforts should be viewed against a backdrop where ARPU and customer churn continue to challenge mobile
operators’ business models. In order to meet customer expectations and provide a high quality user experience, it is critical that HSDPA furnish the bandwidths advertised – even as the number of subscribers grows. Thus, operators acknowledge that HSDPA threatens to increase bandwidth requirements in the RAN by 2x, 4x, or more. To achieve these capacities while simultaneously keeping the cost of HSDPA economically viable, operators openly admit the cost per bit absolutely must be reduced compared to today’s leased-line E1 backhaul.
HSDPA Offload Using ADSL
To meet the requirement for increased bandwidth with lower cost per bit, the logical first step is to offload the HSDPA traffic onto a packet network. HSDPA represents the data
portion of a “best-effort” mobile service: the service is asymmetric in nature, user traffic tends to have bursty characteristics, and expectations are satisfied with non-continuous, brief utilization of the network bandwidth. This is quite different from the nature and expectations for voice traffic, which tends to have a more continuous rate, and where a continuous, reliable connection is essential – even if the traffic is formed in packets or cells. Thus, separating the HSDPA traffic from the voice traffic, continuing to use leaseline E1s to backhaul the voice and using a low-cost packet transport to backhaul the bursty HSDPA traffic makes sense.
DSL is a logical choice for the packet transport. It is a mature technology that has been widely deployed for residential and business data traffic. DSL enjoys the economies of scale of a mainstream data technology, with low-cost DSL modems readily available using the ubiquitous Ethernet interface to connect with users’ data traffic. xDSL furnishes the advantages inherent to packet transport, including statistical sharing of bandwidth among multiple users that is ideally matched to HSDPA needs.
In particular, ADSL2 and ADSL2+ provide the speed and asymmetry that match HSDPA while retaining the cost and availability advantages. Plus, Carrier Ethernet is emerging as the next-generation technology for backhaul of ADSL2 traffic. A new generation of Ethernet DSLAMs has come to market that enables ADSL2 and ADSL2+ backhaul over Carrier Ethernet.
Pseudo-Wire for HSDPA and ADSL
As much as ADSL is the first logical step for offload of HSDPA traffic, Pseudo-Wire technology is the logical solution for matching HSDPA traffic to ADSL backhaul. In its broadest sense, Pseudo-Wire includes service emulation for transport of frame-based and cellbased services over MPLS, IP, and Ethernet networks. Because HSDPA uses ATM cells and ADSL modems use an Ethernet connection, Pseudo-Wire is a perfect interconnect technology.
A typical Pseudo-Wire deployment in the RAN using both access devices and gateways would deploy Pseudo-Wire Access Devices, available with 1, 2, 4, or 8 E1 ports, at the cell sites where they connect to the existing NodeB equipment. The access devices offer E1 interfaces that can be configured to perform ATM service emulation – recognising the native ATM cells used for HSDPA service and even terminating the IMA protocol from the NodeB. The result – HSDPA service from multiple E1s is aggregated onto a single, efficient Pseudo-Wire and delivered to the ADSL modem on a single Ethernet connection.
The ADSL traffic is backhauled to the mobile operator’s RNC location using either an ATM network or, increasingly so, a Carrier Ethernet metro network. A Pseudo-Wire Gateway is deployed at the RNC, where it terminates the ATM Pseudo-Wire services and functions as a gateway to the RNC. Notice that the Pseudo-Wire gateway aggregates and delivers consolidated ATM flows to the RNC. This functionality, taken with the IMA capability in the Pseudo-Wire access devices, obviates the need to pass the HSDPA traffic through a legacy ATM switch at the RNC location.
The gateway furnishes STM-1 interfaces (VC4) to deliver the consolidated ATM flows directly to the RNC.
Beyond ADSL
While ADSL is a logical technology for offload of HSDPA traffic, many see it as only a first step. A number of questions remain unanswered regarding the overall capacity of ADSL to handle the expected quantity of HSDPA subscribers, both in the aggregation network and the last-kilometer link to the NodeB site. ADSL2 transport in the last kilometer can typically furnish 8 Mb/s of downstream bandwidth and 2 Mb/s of upstream bandwidth. Given typical oversubscription ratios, these last-kilometer link speeds are expected to support growth of HSDPA services for some time. However, in the aggregation network, existing DSL networks (both the ATM-based and the Ethernet-based) have already been architected with significant oversubscription ratios for fixed line
ADSL services. These oversubscription ratios are often as high as 50x for residential consumers and 20x for business customers. Overlaying mobile services on top will surely stress the capacity of these DSL networks with yet another layer of oversubscription.
In addition, HSDPA is itself only a first step. While HSDPA is well matched to the downstream rates of ADSL2, new generations of services, including HSUPA and beyond, are already on the horizon. However, it appears certain that a new generation of backhaul beyond ADSL is needed to handle these services.
New Transport Technologies
Several new technologies have emerged that can meet the challenge of next-generation backhaul beyond ADSL, including VDSL, copper bonding using Ethernet in the First Mile (EFM – based on IEEE 802.3ah), and Carrier Ethernet over optical fiber.
To support this Pseudo-Wire Access Devices could be deployed at cell sites using VDSL, EFM, and optical Ethernet for backhaul transport.
At each cell site, the access devices connect to both NodeB and BTS equipment, enabling backhaul of GSM, R99, and HSDPA traffic over any of these transport technologies.
Each access device offers not only E1 interfaces for ATM service emulation, but also E1 interfaces that can be configured for circuit emulation service (CES) to support the Abis interface for 2G services. In addition, the devices provide customer facing Ethernet interfaces for future service needs, such as WiFi or WiMAX hot spots, as well as connectivity needs specified for the R5 NodeB.
While VDSL, EFM, and optical Ethernet are all supported for last-kilometer transport to the cell site, typically each of these technologies would funnel into a switched Ethernet or IP/MPLS network. VDSL would use an Ethernet DSLAM, while EFM may use stackable EFM termination nodes, and optical Ethernet may use a carrier-grade Ethernet aggregation multiplexer. Traffic hand-off from the Carrier Ethernet network to the Gateway at the RNC/BSC is typically performed using Gigabit Ethernet.
The Gateway offers not only ATM STM-1 connectivity, but also channelized E1 and channelized STM-1 ports for interfacing to legacy BSC equipment. In GSM applications, the CES Pseudo-Wires (that originated as E1s with Abis at the BTS) are transported across the packet network and then
consolidated in the gateway into TDM bit streams on STM-1 or E1 interfaces to connect with the BSC.
So multiple generations of mobile services could be carried over a single converged RAN. But also notice that both voice and data are carried over this packet-based RAN. In fact, there is no E1 or SDH transport used at any of the cell sites.
With no SDH in the backhaul network to furnish synchronization, the Pseudo-Wire solution must supply very accurate and reliable clocking at the BTS and the NodeB. Accurate, jitter-free clock sources are critically important for both GSM and UMTS services. Axerra’s Pseudo-Wire solution, for example, obtains clocking from the BSC or RNC and extends the timing across the Carrier Ethernet packet network. The result is that multiple generations of service can aggregated onto a single Carrier Ethernet network.
*The above is extracted and amended from a white paper available from Axerra Networks. www.axerra.com