Backhaul Pains

Backhaul, which is the connection between a radio node and the core network, is providing mobile-wireless operators possible with the biggest headache ever (apart from keeping a healthy revenue growth in mature markets ;-) ... it can be difficult to come by in the right quantities and can be rather costly with conventional transmission cost-structures ... Backhaul is expected to have delayed the Sprint WiMAX rollout of their Xohm branded wireless internet service. A Sprint representative is supposed to have said: “You need a lot of backhaul capacity to do what’s required for WiMax.” (see forexample WiMax.com blog)

What's a lot?

Well ... looking at the expected WiMAX speed per Base Station (BS) of up-to 50 Mbps (i.e., 12 - 24x typical backhaul supporting voice demand), it is clear that finding suitable and low-cost bachaul solutions might be challenging. Conventional leased lines would be grossly un-economical at least if priced conventionally; VDSL and Fiber-to-the-Premises (FTTP) infrastructure that could support (economically?) such bandwidth demand is not widely deployed yet.

Is this a Sprint issue only? Nope! .... Sprint cannot be the only mobile-wireless operator with this problem - for UMTS/HSPA mobile operators the story should be pretty much the same (unless an operator has a good and modern microwave backhaul network supporting the BS speed).

Backhaul Pains - Scalability Issues
The backhaul connection can be either via a Leased Line (LL) or a Microwave (MW) radio link. Sometimes a MW link can be leased as well and might even be called a leased line.

With microwave (MW) links one can easily deliver multiples of 2.048 Mbps (i.e., 10 - 100 Mbps) on the same connection for relative low capital cost (€500 - €1,000 per 2.048 Mbps) and low operational expense. However planning and deployment experience and spectrum is required.

In many markets network operators have been using conventional (fixed) leased lines, leased from incumbent fixed-line providers. The pricing model is typically based on an upfront installation fee (might be capitalized) and a re-occurring monthly lease. On a yearly basis this operational expense can be in the order of €5,000 per 2.048 Mbps, i.e., 5x to 10 x the amount of a MW connection. Some price-models trade-off the 1-off installation fee with a lower lease cost.

Voice was the Good for Backhaul; Before looking at the broadband wireless data bandwidth demand its worth noticing that in the good old Voice days (i.e., GSM, IS95, ..) 1x to 2x 2.048 Mbps was more than sufficient to support most demands on a radio base station (BS).

Mobile-Wireless Broadband data enablers are the Bad and quickly becoming the Very Ugly for Backhaul; With the deployment of High Speed Packet Access (HSPA) on-top of UMTS and with WiMAX (a la Sprint) a BS can easily provide between 7.2 to 14.4 Mbps or higher per sector depending on available bandwidth. With 3 sectors per BS the total supplied data capacity could (in theory ... ) be in excess of 21 Mbps per radio Base Station.

From the perspective of backhaul connectivity one would need at least an equivalent bandwidth of 10x 2.048 Mbps connections. Assuming such backhaul lease bandwidth is available in the first instance, with conventional leased line pricing structure, such capacity would be very expensive, i.e., €50,000 per backhaul connection per year. Thus, for 1,000 radio nodes an operator would pay on an annual basis 50 million Euro (Opex directly hitting the EBITDA). This operational expense could be 8 times more than a voice-based operational leased-line expense.

Now that's alot!

Looking a little ahead (i.e., next couple of years) our UMTS and WiMAX based mobile networks will undergo the so-called Long-Term Evolution (LTE; FDD and TDD based) with expected radio node downlink (i.e., base station to user equipment) capacity between 173 Mbps and 326 Mbps depending on antenna system and available bandwidth (i.e., minimum 20 Mhz spectrum per sector). Thus over a 3-sectored BS (theoretical) speeds in excess of 520 Mbps might be dreamed of (i.e., 253x 2.048 Mbps - and this is HUGE!:-). Alas across a practical real-life deployed base station (on average) no more than 1/3 of the theoretical speed should be expected.

"Houston we have a problem" ... should be ringing in any CFO / CTO's ears - a. Financially near-future developments could significantly strain the Technology Opex budgets and b. Technically providing cost-efficient backhaul capacity that can sustain the promised land.

A lot of that above possible cost can and should be avoided; looking at possible remedies we have several options;

1. High capacity microwave backhaul can prevent the severe increase in leased line cost; provided spectrum and expertise is available. Financially microwave deployment has the advantage of being mainly capital-investment driven with resulting little additional operational expense per connection. It is expected that microwave solutions will be available in the next couple of years which can provide connection capacity of 100 Mbps and above.

Microwave backhaul solutions are clearly economical. However, it is doubtful that LTE speed requirements can be met even with most efficient microwave backhaul solutions?

2. Move to different leased line (LL) pricing mechanisms such as flat pricing (eat all you can for x-Euro). Changing the LL pricing structure is not sufficient. At the same time providers of leased-line infrastructure will be "forced" (i.e., by economics and bandwidth demand) to move to new types of leased bandwidth solutions and architectures in order to sustain the radio network capabilities; ADSL is expected to develop from 8(DL)/1(UL) Mbps to 25(DL)/3.5(UL) Mbps with ADSL2+; VDSL (UL/DL symmetric) from ca. 100 Mbps to 250 Mbps with VDSL2 (ITU-T G.993.2 standard).

Clearly a VDSL2-based infrastructure could support today's HSPA/WiMAX requirements, as well as the initial bandwidth requirements of LTE. Although VDSL2-based networks are being deployed around Europe (and the world) it is not not widely available.

Another promising mean of supporting the radio-access bandwidth requirements is Fiber to the Premises (FTTP), such as for example offered by Verizon in certain areas of USA (Verizon FiOS Service). With Gigabit Passive Optical Network (GPON, ITU-T G.984 standard) maximum speeds of 2,400 Mbps (DL) and 1,200 Mbps (UL) can be expected. If available FTTP to the base station would be ideal - provided that the connection is priced no higher than a standard 2.048 Mbps leased line to day (i.e., €5,000 benchmark). Note that for a mobile operator it could be acceptable to pay a large 1-off installation fee which could partly finance the FTTP connection to the base station.

Cost & Pricing Expectations
It is in general accepted by industry analysts that broadband wireless services are not going to add much to mobile operators total service revenue growth. In optimistic revenue scenarios data revenue compensates for stagnating/falling voice revenues. EBITDA margins will (actually are!) under pressure and the operational expenses will be violently scrutinized.

Thus, mobile operators deploying UMTS/HSPA, WiMAX and eventually (in the short-term) LTE cannot afford to have its absolute Opex increase. Therefore, if a mobile-wireless operator has a certain backhaul Opex, it would try to keep it at the existing level or reduce it over time (to mitigate possible revenue decline).

For the backhaul leased-capacity providers this is sort of bad news (or good? as it forces them to become economically more efficient) .... as they would have to finance their new fixed higher-bandwidth infrastructures (i.e., VDSL or FTTP) with little additional revenue from the mobile-wireless operators.

Economically it is not clear whether mobile-wireless cost-structure expectations will meet the leased-capacity providers total-cost of deploying networks supporting the mobile-wireless bandwidth demand.

However, for the provider of leased fixed-bandwith, providing VDSL2 and/or FTTP to the residential market should finance their deployment model.

With more than 90% of all data traffic being consumed in-house/in-door and with VDSL2/Fiber-to-the-Home (FTTH) solutions being readily available to the Homes (in urban environments at least) of business as well as residential customers, will mobile-wireless LTE base stations be loaded to the extend that very-high capacity (i.e., beyond 50 Mbps) backhaul connections would be needed?

Google Openness - Your Click make Google Tick

In an open letter to the chairman of the USA Federal Communications Commission (FCC) Kevin J. Martin, Google pleas for openness in the recently ended (i..e, March 2008) auction for the 700 MHz;

1. Open Applications - users can gain access to and use any applications, services or content.
2. Open Devices - any device on any network.
3. Open Wholesale Services - Service Providers and Virtual Network Operators should get wholesale access to the 700 MHz based network(s) on reasonably non-discriminatory commercial terms.
4. Open Network Access - service provides and virtual network operators should be allowed to interconnect with the 700 MHz wireless network(s).

The two first points of Open Applications and Open Devices are in principle independent of the 700 Mhz auction, although they can of course be made mandatory in the particular auction requirements and ..... so they were.

The Open Wholesale Service point makes the mind bugle (well at least mine) while figuring out funny wholesale models that would be non-discriminatory to both the wireless operator (having invested in spectrum and network) and Googles-and-alike (GAAs ... whomever other than Google that might be?). The biggest question for a network operator providing wholesale to GAAs is likely going to be how to get a piece of the Google advertisement revenue pie.

Open Applications
Within devices capabilities and network possibilities this does not sound like mission impossible. Obviously, if a wireless network is interconnected to the web (i.e., 4th requirement) services and application available in general to a device (pc, laptop, etc.) connected to the fixed internet would also be available to the mobile device.

However, there are particular services and applications that wireless operators might want to traffic control and manage. Particular in the case of having only an 11 MHz bandwidth available (i.e., USA nationwide C-band @ 700MHz) on the air-interface, heavy peer-2-peerk applications and streaming might result in severe congestion and loss of service quality. Thus, the ability to control and manage the Quality of Service per application / content category will be necessary in order to avoid that few heavy users jeopardize the service quality for the majority of average wireless users.

The wireless operator however should have no problems in complying with Google Point 1.

Open Devices - Open Networks
This requirement might appear harmless and not worth worrying about. In principle a user with a subscription and who pays the access price can have access to any network his device is capable of communicating with (not exactly true in most mobile standards, such as GSM and UMTS/HSPA, of today). Basically WiFi hotspot access comes closest to such a business principle and if the customer does not care about the mobile operator services and content this would suffice.

The mobile business model is (even technically) not build on principles of free access between wireless / mobile networks. A mobile subscription (post-paid or pre-paid) is associated with customer acquisition cost, often subsidising the user terminal. The subscription requires the customer to keep paying for a period of time to pay-back the upfront customer investment done by the mobile operator.

Allowing a business model were customers can freely roam/move across networks might require a different financing mechanism (or none) of the consumers device and access rights. As a service provider with whole-sale agreements with several wireless network operators could enable this for their customer base. For a traditional mobile operator such a model would not be very attractive unless national roaming is invoked due to lack of coverage in a given area.

The Google proposal is from a business model very interesting (altans likely disruptive) although would also require some rethinking of current mobile AAA (i.e., Authentication, Authorization, and Accounting) architecture.

Furthermore, one might fear that by moving to the proposed Google model, that few internet-based businesses would end-up "owning" the customer-data (g-search, gmail, g-chat, g-blog, g-msisdn, g-device, etc..), while the customer-data ownership is currently spread out across several mobile and fixed telecommunication business. The legacy mobile / wireless operator becomes a bit carrier paid by those few internet-based businesses.

Open Wholesale Services
For the Google business this is a really fun one to think about. How would that work for an entity as Google?, for which close to 100% of revenues comes from advertising (i.e., 2007-earnings shows that 98.91% of their $16.594 billion from advertising).

The value for Google going wireless is clearly from opening up a new channel for advertising. The growth potential entering the mobile channel is potential enormeous, with mobile penetration approaching 100% and even far beyond in many European markets (closer to 120%+).

Normal telecommunication wholesale models are based on volumetric usage (i.e., Minutes or Bytes). However Google would hardly trigger any direct volumetric usage with exception of the volume it takes to download google.com. Alas there might be a considerable traffic stream arising from YouTupe and some from gmail usage. Even following an advertising link will generate traffic although not necessarily generating much additional volumetric usage. Of course the question is how to distinguish between Google generated traffic and non-Google traffic?

Furthermore, the price per advertisement click that Google earns could be significantly different (i.e., higher) from the cost of the click according with a standard volumetric wholesale model. Furthermore, a different click might have different values but still generate the same volume and associated cost.

Maybe the wireless operator should not care too much how Google earns its money as long as the traffic generated by providing access to happy Googlers and GAAs are recovered by a healthy margin and does not jeopardize the quality of other customers.

Open Network Access
Yeah this sort of make sense .... without this the first three points become rather academic. There is no essential technical barriers for interconnect.

Winner of the 700-MHz Auction is ... Google !

The United States has recently ended (March 2008) the auction 5 blocks (see details below) of the analog TV spectrum band of 700-MHz. More specifically the band between 698 - 763 MHz (UL) and 728 - 793 MHZ (DL), with a total bandwidth of 2x28 MHz. In addition a single band 1x6 MHz in 722 - 728 MHz range was likewise auctioned. The analog TV band is expected to be completely vacated by Q1 2009.

The USA 700 MHz auction result was an impressive total of $19.12 billion, spend buying the following spectrum blocks: A (2x6 MHz), B (2x6 MHz), C (2x11 MHz) and E (1x6 MHz) blocks. The D (2x5 Mhz) block did not reach the minimum level. A total of 52 MHz (i.e, 2x23 + 1x6 MHz) bandwidth was auctioned off.

Looking with European eyes on the available spectrum allocated per block it is not very impressive. The 700 MHz frequency is clearly very economical for radio network coverage deployment in particular compared the high-frequency AWS spectrum used by T-Mobile, Verizon and Sprint. However, the 6 to 11 MHz (UL/DL) is not very impressive from a capacity sustainance perspective. It is quiet likely that this spectrum would be exhausted and rapidly leading to a significant additional financial commitment to cell splits / capacity extensions.

This $19.12 billion for 52 MHz translates to $1.22 per MHz spectrum per Population @ 700 MHz.

This should be compared to following historical auctions
* $0.56/MHz/Pop @ 1,700 MHz in 2006 US AWS auction
* $0.15/MHz/Pop (USA Auction 22 @ 1999) to $4.74/MHz/Pop (NYC, Verizon).
* $1.23/MHz/Pop Canadian 2000 PCS1900 Auction of 40MHz.
* $5.94/MHz/Pop UK UMTS auction (2001) in UK auctioning a total of 2x60 MHz FDD spectrum (TDD not considered).
* $7.84/MHz/Pop German UMTS auction in 2001 (2x60 MHz FDD, TDD not considered).

(Note: the excesses of the European UMTS auctions clearly illustrates a different time and place).

What is particular interesting is that Verizon "knocked-out" Google by paying $4.74 billion for the nationwide C-block of 2x11 MHz. "Beating" Google's offer of $4.6 billion.

However, Google does not appear too sadened of the outcome and .... why should they! Google has to a great extend influenced the spectrum conditions allowing for open access (although it remains to be seen what this really means) to the C spectrum block; The USA Federal Communications Commission (FCC) has proposed to apply "open access" requirements for devices and applications on a the nation wide spectrum block C (2x11 MHz).

Clearly Google should be regarded as the winner of the 700 MHz auction. They have avoided committing a huge amount of cash for the spectrum and on-top having to deploy even more cash to build and operate a wireless network (i.e., which is really their core business anyway).

Googling the Business Case
Google was willing to put down $4.6 billion for the 2x11 MHz @ 700 MHz. Let's stop up an ask how their business case possible could have looked like.

At 700 MHz, with not too ambitious bandwidth per user requirements, Google might achieve a typical cell range between 2.5 and 4 km (Uplink limited, i.e., user equipment connection to base station). Although in "broadcast/downlink" mode, the cell range could be significantly larger (and downlink is all you really need for advertisement and broadcast;-).

Assume Google's ambition was top-100 cities and 1-2% of the USA surface area they would need at least 30 thousand nodes. Financially (all included) this would likely result in $3 to $5 billion network capital expense (Capex) and a technology driven annual operational expense (Opex) of $300 to $500 million (in steady-state). On top of the spectrum price.

Using above rough technology indicators Google (if driven by sound financial principles) must have had a positive business case for a cash-out of minimum $8 billion over 10 years, incl. spectrum and discounted with WACC of 8% (all in all being very generous) and annual Technology Opex of minimum $300 million. On top of this comes customer acquisition, sales & marketing, building a wireless business operations (obviously they might choose to outsource all that jazz).

... and then dont forget the customer device that needs to be developed for the 700 MHz band (note GSM 750 MHz falls inside the C-band). Typically takes between 3 to 5 years to get a critical customer mass and then only if the market is stimulated.

It would appear to be a better business proporsition to let somebody else pay for spectrum, infrastructure, operation, etc... and just do what Google does best ... selling advertisments and deliver search results ... for mobile devices ... maybe even agnostic to the frequency (seems better than wait until critical mass has been reached at the 700 MHz).

But then again ... Google reported for full year 2007 a $16.4 billion in advertising revenues (up 56% compared to the previous year).(see refs Google Investor Relations). Imagine what this could be if extended to wireless / mobile market. Still lower than Verizon's 2007 full year revnue of $23.8B (up 5.5% from 2006) but not that much lower considering the difference in growth rate.

The "successfull" proud owners (Verizon, AT&T Mobility, etc....) of the 700 MHz spectrum might want to keep in mind that Google's business case for entering wireless must have been far beyond the their proposed $4.6 billion.


Appendix:

The former analog TV spectrum auction has been divided UHF spectrum into 5 blocks:
Block A: 2x6 MHz bandwidth (698–704 and 728–734 MHz); $3.96 billion
Block B: 2x6 MHz bandwidth (704–710 and 734–740 MHz); $9.14 billion dominated by AT&T Mobility.
Block C: 2x11 MHz bandwidth (746–757 and 776–787 MHz) Verizon $4.74 billion
Block D: 2x5 MHz bandwidth (758–763 and 788–793 MHz) No bids above the minimum.
Block E: 1x6 MHz bandwidth (722–728 MHz)Frontier Wireless LCC $1.26 billion