Will the 3G iPhone break the network?

Ilan Seidner, Director of Marketing, RAD Data Communications

This summer is set to be the greatest test of 3G networks since their launch. The steady drumbeat of announcements about reasonably priced 3G iPhone service rollouts with unlimited data plans portends a major transformation in how consumers will think about and use the mobile phone. 

The 2G iPhone posed no major threat to mobile networks as the most data intensive applications, such as streaming media, could only be used over WiFi and therefore the consumer’s own backhaul. The 3G iPhone revolutionizes the situation by allowing users to access bandwidth heavy applications on the move and, most importantly, use the operator’s backhaul network in the process. According to a study by M:Metrics at the beginning of 2008 iPhone users are many times more likely to surf the web, engage in social networking and download video — compared to other smart phone afficionados.The big question is – will the backhaul network break under the strain? 

O2’s CTO Vivek Dev confessed at Mobile World Congress in February that sales of the2G  iPhone were “driving unheard-of levels of mobile internet usage, and the introduction of flat rate data tariffs is expected to increase this further. Both of these place huge capacity demands on our networks.”  This was quite the revelation at the time as few thought that 2G mobile internet services could ever create real capacity concerns.  Evidently the iPhone made the mobile internet much more appealing than anything before – a fact that must have sounded like sweet music to mobile operators who bet the bank on the premise when it came to buying 3G licenses but until now have endure lack luster sales and limited data take-up. 

But it’s not all good news.  The 3G iPhone poses a far greater threat to networks than its predecessor both because its HSDPA data speeds create a far greater load on the network and as it is being offered far more cheaply than the previous model. While the first iPhone was too expensive to be truly mass market, the 3G version has a brutal combination of low cost and real mobile broadband capabilities that is likely to signal a massive influx of new heavy bandwidth users. The spotlight is on the mobile operator’s all too frequently neglected backhaul networks to see if they’ll cope. The old adage ‘be careful what you wish for, it may just come true’ has rarely seemed so apt. 

Backhaul, more than any other area, is being transformed from a network no-brainer, to a fundamental issue that will determine the profitability, and success, of mobile broadband networks and services. This is a testament to the revolution that cellular radio networks have undergone from transporting just voice and SMS to being able to stream video and deliver broadband-speed Web surfing. This has created the unprecedented situation in which the bottleneck in mobile networks has shifted upstream from the radio interface to the backhaul network, meaning that operators are faced with the challenge of significantly upgrading their fixed network capacity to deliver the kinds of next-generation services being promised to users on the move. 

3G and HSPA technology have created a revolution in cellular radio that transforms the data rates achievable between phones and basestations. However, no such transformation has taken place in the backhaul network. Once data arrives at the cell tower and starts its travel back to the mobile core and the Internet – along the backhaul – it faces a significant bottleneck: existing legacy network architecture with its fixed bandwidth and rigid hierarchy. This is the new mobile broadband bottleneck, and unless it is resolved, the data rates on 3G networks will have to be throttled and the 3G iPhone’s users (and its clones) will have a disappointing experience.

The challenge facing the industry is stark.  A single iPhone user on HSDPA, where a single connection can offer up to 14.4 Mbps, could use the same capacity as over 1,000 GSM subscribers! With the 3G iPhone looking certain to massively outsell its predecessor, mobile operators need to put their networks under the spotlight to avoid either overloading the network or having to massively throttle data rates such that they more resemble the EDGE experience the handset was supposed to replace. 

One may ask, should this really be an issue? E1s are plentiful and their costs have dropped significantly in Europe in recent years. And if more backhaul bandwidth is required, higher bandwidth microwave can always be deployed. But whilst this is technically feasible, the business case simply doesn’t stack up. Today, with few widely used HSPA services in Europe, backhaul already accounts for 75 percent of mobile transport costs and can account for around 20 to 40 percent of total operating expenses. Jumping from a single E1 per base station for GSM to up to 14 for co-located 2G, 3G and HSPA services pushes the mobile service into the red. What's worse is that LTE, the technology that will follow 3G and HSPA in Europe, is being designed to reach data rate targets of 100 Mbps – a huge step above 14.4 Mbps. Someone has to pay for this significant increase in transport costs. Because mobile operators can’t charge more for their mobile broadband services than what subscribers are already paying for existing high-speed DSL and cable connections, the exponential rise in backhaul costs will come from the mobile operators’ bottom line. 

Clearly, backhaul technology is well overdue for a transformation. Most operators have adopted a strategy of simply adding more E1s as they never needed to deal with a bandwidth explosion – as might happen on July 11^th (3G iPhone D-Day). But with the growth in traffic being decoupled from the revenue stream, this won’t work. More recently mobile operators have aggressively begun to enact optimisation and aggregation strategies, deploying cell-site gateways at co-located 2G and 3G towers that free up E1 bandwidth and hand-off multiple E1s to a larger pipe, in which the economics of E3 or even STM-1 make sense.  And while this, along with statistical multiplexing of 3G data, can result in network simplification and 50 percent savings, it only addresses an incremental yet controlled introduction of early 3G and HSPA services. Widespread, heavy HSPA usage, which requires many more cell sites with more bandwidth, will necessitate a different strategy if the service is to be ubiquitous and profitable.

For many European operators, the key to the solution is self-evident: Ethernet. Ethernet delivers about five times the bandwidth of E1 lines for the same amount of money, is extremely flexible and is easier and less expensive to maintain and manage than legacy networks. Luckily for European mobile operators, there is ample low-cost Ethernet available over wholesale DSL and Metro Ethernet networks.

However, anybody familiar with telecoms knows that the path to a complete solution is never easily implemented. When the CDMA world was putting together its specifications, the ability to backhaul directly over IP was built in and, therefore, cable lines and DSL can be used for that purpose.  In the GSM and UMTS world, however, it was decided to use TDM and ATM as interfaces for base stations.  This means that if European mobile operators want to capitalise on the benefits of using inexpensive DSL, they need some means of emulating ATM over Ethernet.  

With an eye to the future, RAD Data Communications pioneered and patented a number of years ago a technology called TDMoIP® (Time Division Multiplexing over IP), which enables TDM type traffic, such as 2G GSM, to run transparently over IP. This technology was extended to ATM and has since been formally adopted by all the major standards organizations, including the IETF and the ITU-T, and is commonly known as PWE3 (pseudowire emulation end-to-end). Pseudowire technology is regarded as the pre-eminent solution for transporting legacy protocols such as TDM and ATM over DSL, Ethernet and MPLS. It is this solution that is now being used or adopted by major European Tier 1 operators, such as T-Mobile, to enable their HSPA network to run over DSL.

Now that the basis for the solution exists the question remains how best to deploy it. Since pseudowire allows not just 3G traffic but also 2G traffic to run over DSL, can operators now remove their E1s altogether?

Unfortunately, as ever, it is not this simple.  As we’ve seen packet-switched networks offer cost savings and enable flexibility and scalability but they are also asynchronous and non-deterministic by nature and as such they also introduce inaccuracies such as packet delay variation and packet loss.  In short: packet-based transport networks such as Ethernet differ from traditional TDM and ATM in that there is no clock source for synchronisation.

The Need for Timing over Packet (ToP) Solutions

For packet transport to meet required “SDH/SONET or better” performance levels for mobile networks, there is a need for resilient clocking and synchronization schemes. Without proper mechanisms to negate the effect of impairments such as packet delay, delay variation and packet loss, all of which are inherent to packet-switched networks, operators run the risk of service disruptions, impaired cell hand-offs and excessive dropped calls.

As a result, accurate clock recovery and distribution methods have been developed to handle packet transport’s inefficiencies, capture the average transmission rate of the original bit stream, and distribute accurate timing information to all network elements. These methods include NTR, NTP, ITU-T G.8261, Synchronous Ethernet, and the recently approved IEEE 1588v2.

However, these synchronisation solutions are only just coming to market now. Mobile operators are understandably reluctant to rush into any area that threatens their voice revenues.  However, they cannot afford to ignore the backhaul issue and as such are looking at two different options for employing packet-based technologies. 

One, the so-called “Big Bang” approach, would transport all real-time voice and video traffic, along with data, over a packet-switched backbone network. The other, the so-called “Step-by-Step” or “Hybrid” approach, would continue to let time-sensitive voice and video to be carried over E1s and the deterministic legacy network, while less time-sensitive statistical HSPA data would be handed off onto DSL.

Most major operators in Europe are currently mulling over this decision and will most likely come to the same conclusion over the next twelve months by adopting this migratory approach. By migrating to a partial IP system instead of a complete one, the operator achieves major cost savings without jeopardising QoS-sensitive applications, especially the old revenue workhorse - voice.  Either way, migrating to an Ethernet transport network also takes into account future network migration to UMTS 3GPP Rev. 6, the introduction of LTE and the implementation of fixed-mobile convergence: the holy grail of an efficient all-IP network.

As 3G iPhone sales ramp up towards Christmas and the network becomes increasingly congested, all eyes will be on how well the user experience stacks up.  Apple may be able to create the most compelling products of the past decade but the 3G experience is out of their hands and threatens to upset Steve Jobs’ biggest product to date.  Everything now rests in the hands of the operators. At stake isn’t just the success of the mobile internet and a revenue model based around data, but the very existence of an industry built around delivering bandwidth devouring content over the air.