By Robert Best - Award Solutions

Summary: The deployment of LTE networks will be a disruptive event in wireless network evolution, what we refer to as a Black Swan.  Traditional backhaul solutions will need to be reconsidered.   Though we have always had one way of handling wireless backhaul, LTE will dictate a need to look at alternatives.  Those alternatives will likely involve the use of MPLS and Ethernet for backhaul.  Though some have considered MPLS to be a core network technology, LTE is likely to push MPLS deep into the wireless access network.

I am reading a book by Nassim Taleb called The Black Swan; The Impact of the Highly Improbable.  I was reminded of this when I watched a PBS program last night called "The Ascent of Money", and moderated by Harvard Professor, Niall Ferguson, during which he mentioned the Black Swan concept.  It turns out that, prior to the discovery of Australia; the Western world had believed that all Swans were white.  After all no one had ever seen a Black Swan.   Among other new and interesting things that came from the discovery of Australia what the fact that some swans are actually black.  This was a disruptive event that caused us to have to change the way we considered and explained swans. 

The basic premise of the book as well as the TV program is that our lives are fairly routine, and we develop the belief that what happens to us routinely will continue to happen; "because that is the way things have always been".  The fact is, if we look over our life time, the things that really define us and how we live, are actually a small set of major unpredictable events.  These events can cause us to redirect our lives in a totally different direction, where yesterday's rules no longer apply.   Taleb offers as an example of a Black Swan, the events of September 11, 2001.  The events of that day were totally unexpected, but the effects of that day have changed how we all look at life.  Ferguson describes as a Black Swan, the unexpected events that led to the current financial crisis, but again have impact on all of our lives.   Can you think of any such events in your life? 

So where does LTE come in to this?  LTE is a 4G wireless access technology along with mobile WiMAX.  One thing that LTE (and mobile WiMAX) offers is a disruptive increase on bandwidth supported across the air.  While early wireless technologies enables the subscriber to transmit data at the rate of 9.6 kbps, LTE claims a theoretical throughput rate of 300 Mbps; an increase factor of an astonishing 30,000.  LTE is also claiming to be capable of carrying a vast array of service types, each with its own requirements regarding network treatment.  Everyone knows that traffic received across the air must be backhauled to some network node using TDM-based T1 carriers; it's always been done that way. 

If the throughput offered across the air interface is increased, then obviously we will need more T1s to backhaul this added traffic.  The question is, how many T1s should be deployed?  Sprint (XHOM) faced this question in September of 2008 when they finally launched the first 4G network in the US (a mobile WiMAX) in their Baltimore-Washington market.  (http://www.icellphonedeals.com/how-much-is-sprint-xhom-wimax-service-price/).  More recently (January 2009) the XHOM partner Clearwire has solidified the rollout of mobile WiMAX by announcing the commercial deployment in Portland Oregon.  4G is apparently gaining a foot hold (http://www.businessweek.com/technology/content/jan2009/tc2009017_733573.htm?chan=top+news_top+news+index+-+temp_news+%2B+analysis).  

Addressing the backhaul question, the preliminary mobile WiMAX architecture described backhaul in terms of multiple T1s from each base station.  Though their initial architecture expressed the backhaul requirements in terms of multiples of T1s, It soon became apparent that 4G may be one of those Black Swans.   That is, possibly the 4G networks are so disruptive in terms of offered bandwidth and requirements to support multiple media sources, that the old T1 solutions no longer apply.   We will see later that this is fully recognized by the XHOM people and their plans include significant change in the way they handle backhaul. 

Bases on gut feel, and support by what Sprint has published regarding their backhaul plans, this white paper considers 4G technologies to be just such a Black Swan event.  We therefore need to consider new solutions for backhaul in the 4G environment.  The proposed new backhaul approach will ultimately phase out T1s as a backhaul solution and at the same time extend the reach of MPLS possibly to the base station, to support Quality of Service and Traffic Engineering.  Though the most recognized use for MPLS is today's network is in the core, I have been telling my MPLS students that MPLS will be reaching into the access network with the introduction on 4G services.  Below I will support this claim.

Before discussing the claim that MPLS will eventually reach the wireless base station, we should consider the question of timing.  If I am right, but it won't happen for another 5 years, when there is no urgency.  However if it is just around the corner, then we need to begin planning for it now. 

If we believe what the carriers are saying today, the dominate 4G technology will ultimately be LTE and not mobile WiMAX.  However the word on the street is that LTE will not be ready until 2010 or 2011.  Could the successful launch of mobile WiMAX instill a sense of urgency for deploying LTE?  The answer may be YES.  Verizon Wireless (as well as AT&T and other carriers) are planning to eventually roll over to LTE.  In a recent article, and because of the deployment of mobile WiMAX by XHOM, VzW has expressed more on an urgency to deploy LTE.  In an article entitled "Verizon accelerating LTE, putting further pressure on Clearwire, WiMAX", VzW has stated the intent to escalate the deployment on LTE in their network.  Verizon Communications Chief Technology Officer *** Lynch said that VzW will have its first 4G networks rolled out by the end of 2009! Lynch said the first commercial networks would likely go live by the end of 2009. "We expect that LTE will actually be in service somewhere here in the US [in 2009]" (http://telephonyonline.com/wireless/news/verizon-accelerating-lte-1210/).  Along with the deployment of LTE comes the need to decide how to handle backhaul. I have not seen anything to tell me how VzW will address the backhaul, but they will undoubtedly be faced with the same Black Swan event faced by XHOM.  In other words, the need for a solution will likely be sooner rather than later.

So how is Sprint/ XHOM planning to their backhaul?  A hint can be seen in a paper presented by Dr. Esmail Dinan of Sprint at the MPLS 2007 conference in July of 2007.  The presentation is entitled, "MPLS Backhaul for Mobile WiMAX and LTE" (http://www.isocore.com/mpls2007/cd/Presentations/323%20Esmael%20Dinan.pdf).  In the presentation, Dr. Dinan described how XHOM planned to migrate from today's nxT1 solution to an MPLS solution of tomorrow.  The approach describes an Aggregation network between the Access network and the Core network.  (i.e. three networks participate in the solution; the Access network from the BTS to some aggregation node; the Aggregation network from the aggregation node to a gateway node; then the Core network beyond the gateway node.   The XHOM solution has traffic received across the air being received by a base station in the Access network.  Legacy traffic (i.e. CDMA) would be backhauled to an aggregation node (at the edge of the Aggregation network) using the traditional T1 solution.  Traffic received by WiMAX base stations, would be backhauled as Ethernet to the same aggregation node.  Aggregated traffic would then be transported as IP over Ethernet over MPLS through the Aggregation network to a switching node at the gateway site.  This switching node would separate the legacy traffic and the WiMAX traffic delivering the legacy traffic to a CMDA BSC, and the WiMAX traffic to the gateway node.   How the traffic is to be handled in the core network was not considered.   The Chart below was comes from Dr. Dinan's presentation.

Dr. Dinan goes on to describe how MPLS would be used in the solution, primarily by deploying Layer 2 Virtual Private Networks (L2VPN) and Pseudowires.  He later presents his vision of the ultimate solution which takes MPLS to the Base Station, elimination the need for Ethernet altogether.  His claim is that this solution will result in simpler network architecture and enable an effective QoS environment. 

So what is this L2VPN and why is it important to MPLS and to LTE backhaul?  L2VPN enables the service provider to transport any type of layer 2 protocol (e.g. ATM, Ethernet or even any proprietary L2 protocol) encapsulated inside an MPLS header without requiring the "customer" (e.g. base station) to change protocols to meet the backhaul requirements.   When the traffic is delivered to the "other end", the Layer 2 frame is retrieved in the same form as it was constructed, and delivered to the appropriate network device for processing.  That is, encapsulating packets at the aggregation point on the base station side and transporting them to the disaggregation point on the gateway side is just like sending a bit-stream across a carrying "wire" (i.e. a "Pseudowire").  The intermediate network is totally unaware of the traffic being carried, nor does it need to be.  In XHOM's case, legacy traffic from CDMA 1x may be one bit stream, while WiMAX traffic may be another bit stream.  The Pseudowire between the aggregation point and the disaggregation point carries both streams concurrently without the need to distinguish between them.  The receiving disaggregation node (the Ceterus UTX 8212 in XHOM's case) recovers the initial layer 2 frame and forwards it to a CDMA BSC if it originated as CDMA traffic, and to the ASN-GW if it originated as WiMAX traffic.

Is there any other indication of interest to use MPLS to backhaul 4G (e.g. LTE) traffic?  Recognizing the emerging need for a new way to manage backhaul in the 4G environment, IP/MPLS Forum (formerly known as the MFA Forum) has published proposed technical  specifications the describe a framework and requirements for the use of MPLS in mobile wireless access and aggregation networks (http://www.ipmplsforum.org/tech/IPMPLSForum20.0.0.pdf).  According to this paper, the term "Mobile Backhaul" represents the first segment of transport network used for carrying traffic from the Base Station.  Mobile access networks are described as either "centralized mobile networks" or "flat mobile networks".  Centralized mobile network are characterized by mobile access networks where base stations home to some controlling node (e.g. the RNC) forming basically a star architecture.  In Flat mobile networks, the Base stations are able to communicate directly with other without the need for this controlling node.  Traffic is backhauled to a gateway node (e.g. ASN-GW) when it needs to be transported outside the local networking environment.  We find centralized mobile networks in 3G and earlier technology, while LTE and WiMAX networks (i.e. 4G) fall into the category of flat mobile networks.

As is the case with the XHOM network, IP/MPLS Forum defines the wireless carrier's network as being composed of three sub-networks; the "access network" (BTS to some aggregation node); the "aggregation network" (aggregation node to a Radio Controller - e.g. RNC in 3) for centralized wireless networks; or to a gateway node (e.g. Serving Gateway - SGW for LTE) for flat wireless networks.  The third sub-network is represented by the legacy packet switched core network.  The paper describes seven different MPLS backhaul cases for centralized mobile networks and seven different MPLS backhaul cases for flat wireless networks.  Since we are focusing on LTE, we will leave it up to the reader to further examine centralized wireless network cases. 

A characteristic that distinguishes flat mobile networks from centralized wireless networks is that flat mobile networks are always IP based.  IP/MPLS Forum describes four different flat mobile networks; LTE, mobile WiMAX, HSPA+ flat, and UMB).  We will only discuss LTE and Mobile WiMAX here. Our interest is primarily in LTE so we will not describe the other categories.

For flat wireless networks, IP/MPLS Forum describes four L2VPN-based cases and three L3VPN-based cases.  The chart below depicts the L2VPN-based cases.  Note that these cases describe various proposed ways that Ethernet and Virtual Private LAN Service (VPLS) can interwork to provide a backhaul solution.

 

L2VPNs can be either point-to-point (known also as Virtual Private Wire Service - VPWS), or point-to-multipoint (known as Virtual Private LAN Service - VPLS).  The offered L2VPN solutions use VPLS hence represent point-to-multipoint networks.  Intuitively VPLS networks behave like a giant Ethernet switch.  L2VPNs are characterized by having Customer Edge (CE) routers (routers outside the MPLS network but directly connecting to an MPLS PE router), peering with each other.  The implication is that the CE routers will exchange routing information directly with each other, and the MPLS network is never made aware of the external routes.   The top two cases describe Ethernet being backhauled through the access network, then transported via MPLS through the aggregation network.  The two are differentiated by whether the PE (MPLS edge router) is in the access network or in the aggregation network.  In case 3, MPLS is extended all the way to the Cell Site gateway (CSG).   Not seen in the chart is the fact that the same MPLS backhaul solution selected for transporting traffic to the gateway must also be used for BS to BS transport.

The fourth L2VPN solution involves "H-VPLS".  H-VPLS partitions the network into several edge domains that are interconnected using an MPLS core. The considerations for the edge devices are now simplified, as they need only learn of their local n-PE devices and therefore do not need large routing table support.  In the solution, individual MPLS-based Pseudowire spokes would connect the CSG nodes to the PE router at the edge of the aggregation network, and the aggregation network would function as this H-VPLS network.

In L3VPNs, the CE router and the PE router peer and (external) routes are advertised from the CE to the PE.  Therefore the MPLS network actually participates in the overall routing process.  L3VPNs provide a degree of scalability not available in L2VPNs, and enable the MPLS provider to more effectively participate in QoS and congestion management issues.   As with the L2VPN cases, L3VPN can serve as only part of the backhaul solution of may become the total solution.  When the L3VPN solutions reaches all the way to the BTS side access network, each CSG is treated as a separate L2VPN "customer" represented by its own logical routing table (VRF - Virtual Routing Forwarding instance)   

/Dr. Robert Best