Mobile core wars

When you think of 4G, the tendency is to think of radios — fast radios that pump out massive capacity into the wireless industry's tiny niche of the electromagnetic spectrum. While radio access infrastructure is where most of the attention is focused — and where most of the money is — another part of the 4G network is starting to gather a lot of attention.

The 4G mobile core isn't just an ancillary component to the network anymore. Its role in the wireless network is not only becoming more prominent as networks migrate to all-IP architectures and the sheer volume of network data exponentially multiplies, the 4G core is now becoming a key point of differentiation for wireless vendors. The highly standardized nature of the long-term evolution (LTE) and WiMax architectures makes it hard to distinguish between the major vendors' radio access portfolios, but in the 4G core those differences are becoming quite pronounced. Philosophical camps are emerging, pitting the relative importance of the data plane versus the control plane. Vendors have adopted wildly different development strategies, some focused on evolving existing 3G architectures for 4G, while others start from scratch.

Perhaps there no is bigger evidence of the growing importance of the mobile core than the sudden spurt of acquisition activity in recent months. Cisco Systems has agreed to buy 3G and LTE gateway specialist Starent Networks for a cool $2.9 billion, even though it had developed its own line of WiMax and LTE core products based on its routing platforms. Tellabs entered the wireless core market with a similar — though far cheaper — acquisition, purchasing WiMax access service network (ASN) gateway vendor WiChorus for $165 billion in December. Even Hitachi felt the progress from its core joint venture with Nortel Networks was significant enough to pick up Nortel's LTE evolved packet core (EPC) assets at a bankruptcy auction, though it got them for a mere $10 million.

Every generation of wireless technology is disruptive, providing an opportunity for one vendor to gain market share or an opening for new market entrants. But while the radio network vendors for 4G are the same faces as the vendors for 3G, there is a much bigger disruptive opportunity in 4G core, said Stu Bennington, director of global portfolio marketing for Tellabs. Tellabs may be a mobile backhaul specialist, but it saw in WiChorus the opportunity to expand into a new area of the wireless network, while complimenting its established product line, Bennington said.

"Every time you make that transition from one technology to another, there is an opportunity," Bennington said. "We see some inflection points in the market where we can break in with some really good technology."

Is the 4G mobile core primarily a data architecture or a mobile architecture? That is the question the wireless vendor community is vigorously debating. Alcatel-Lucent and Ericsson have firmly come down on the data side. Both vendors have revamped their mobile core portfolios for 4G, essentially turning over the development of the powerful gateways in the EPC to their IP divisions. In Alcatel-Lucent's case the packet data network gateway (PDN or P-gateway) and the service gateway (S-gateway) are built on the 7750 Service Router it has sold to wireline operators across the globe. Ericsson is basing its gateways on the Redback SmartEdge Router. They share in common enormous routing power usually associated with IPTV and IP transport networks rather than the more modest needs of mobile data — that is until the advent of 4G, said Lindsey Newell, vice president of marketing for Alcatel-Lucent's IP group.

The vendors have to start from scratch with 4G, Newell said. The GPRS gateway support nodes (GGSNs) that are the primary packet pushers in mobile networks today have certainly scaled as 3G data has grown, Newell said, but they were never meant to handle the enormous packet loads and robust data services of the 4G network. Newell compared the 3G core to the broadband remote access server (BRAS) in wireline. They successfully evolved from platforms supporting only dial-up subscribers to broadband servers handling multi-megabit-per-subscriber connections. But as fiber is being pushed to the neighborhood and then into the home, as individual broadband connections open wide, and as new high-bandwidth services like IPTV emerge, the BRAS has given way to the much more powerful multiple service routers.

"In wireline a much higher availability platform was needed — not just one support best-effort data — to handle IPTV service," Newell said. "We believe LTE is going to create the same kind of discontinuity with the evolved packet core. Those GGSNs were designed for Web browsing and e-mail. With LTE we'll see the use of large-scale applications the current platforms simply aren't designed for."

While Alcatel-Lucent and Ericsson tout the raw routing power of their platforms, another set of vendors are emphasizing the finesse of their 4G core solutions. On the surface, Nokia Siemens Networks is smoothly transitioning its GGSN architecture to handle the increased demands of LTE, utilizing a proven mobile data platform embedded within the networks of hundreds of operators globally. But below the surface NSN is taking a radically new approach. To hear Alkarim Haji, head of converged Internet connectivity for NSN, describe it, the company and other purpose-built gateway vendors like Starent are the ones taking the completely new approach to the evolved packet core, while the competition remain static.

Haji said that Alcatel-Lucent and Ericsson have always built their mobile cores on routing platforms — the names of the routers have just changed to the evolved packet core. NSN has taken the somewhat controversial step of designing its LTE packet core on the advanced telecommunications computing architecture (ATCA). The usual knock on ATCA is it provides a lot of processing power but little in the way of packet-pushing capabilities. Consequently other vendors have chosen to build the control plane elements of their EPCs — the mobility management element (MME) and the policy charging and enforcement function (PCRF) — on ATCA but use raw data throughput of routers for their gateways. For NSN, though, the P-gateway and S-gateway are each another set of blades in the same ATCA shelf.

Yes, 4G requires a lot of capacity, Haji said, but newer generation ATCA architectures support a lot more throughput than its detractors give it credit, and with the aid of multicore processors, NSN's EPC platform can support hundreds of GB/s of IP traffic. And in doing so, NSN isn't sacrificing any capacity in signaling and the control plane. Router-based solutions where made for static IP networks — DSL lines don't move.

"The fixed edge is nothing like mobile edge," Haji said. If you're supporting 5 million home broadband connections, you need the raw throughput of a massively scalable router platform, but if you're supporting 5 million mobile sessions, you need raw processing power, Haji added. "The mobile edge isn't just about throughput. It's about sessions and the control plane. It's about intelligence."

The rash in packet core activity of late doesn't necessarily signal a growing importance of the mobile core, said Peter Jarich, wireless infrastructure analyst with Current Analysis. The core has always been an important element in 3G and even 2.5G; what's changed is the importance vendors have assigned to it as part of their overall strategies, Jarich said.

Vendors have long been accustomed to selling complete networks, winning contracts to supply not just the radio access and data and voice cores, but often backhaul, transport and service delivery architectures, Jarich said. In 2G builds and earlier 3G builds, the RAN and the packet core were virtually inseparable. Then Starent came along and convinced operators that they could mix and match their elements. It won over CDMA operators first, and in recent years has been moving into the GGSN space, capturing HSPA gateway contracts around the world. Now many larger operators — AT&T and Verizon Wireless being two big examples — are starting to divvy up their networks into domains, and instead of looking for the best integrated end-to-end solution, they're looking for the best individual pieces and requiring their vendors to link them together, Jarich said.

Continue on Page 2