vRAN Tech Hits Resistance at SK Telecom

SK Telecom has met internal resistance to the virtualization of its radio access network (RAN), with its operations department worried that an overhaul might add to complexity and even pose a security threat, Light Reading has learned.

Operations employees at SK Telecom (Nasdaq: SKM) are concerned the use of open hardware and software in the RAN could become burdensome, and that it would raise questions about who takes responsibility if something goes wrong.

A talking point at last week's Mobile World Congress (MWC) in Barcelona, virtualization of the RAN would allow an operator to run baseband processing functions as software programs on common, off-the-shelf (COTS) servers. In theory, this could lead to cost savings and other service-related benefits.

But these are not clear-cut. France's Orange (NYSE: FTE) last week told Light Reading that the business benefits of RAN virtualization were not obvious and that it currently remains "cheaper and simpler" to upgrade each and every mobile site. (See Orange: Still No Clear Business Case for vRAN.)

In traditional networks, baseband processors can be found at any mobile site. In a virtual RAN (or vRAN), an operator could store the COTS hardware that supports baseband processing in some kind of centralized data facility.

Park Jin-hyo, SKT's chief technology officer, told Light Reading his company has been in discussions with Ericsson AB (Nasdaq: ERIC), Nokia Corp. (NYSE: NOK) and some "new players" about RAN virtualization but has yet to be fully convinced.

"From the point of view of the technology, it is good, but in terms of operations there are some challenges," he said during an interview at MWC. "For example, if we use open hardware and software, who can guarantee the combination, as well as safety and configuration."

SKT's operations department is apparently worried that -- in an open environment with a multitude of vendors -- any vulnerabilities appearing in the vRAN would be left to the operator to fix. "They [operations staff] cannot endure that kind of burden," says Park Jin-hyo.

A potential solution is to come up with a "robust scheme" ensuring SKT's technology suppliers "share this kind of problem."

One of the original attractions of virtualization was that it would allow operators to work with a range of suppliers, instead of being tied to one vendor's products. But long-running concerns about interoperability and commercial arrangements in multivendor networks have held the industry back.

Germany's Deutsche Telekom AG (NYSE: DT) last week criticized its vendors for developing products that work only with their own infrastructure. Orange, meanwhile, has only just started to invest in what it calls "full virtualization" due to concern about who bears responsibility in the event of a problem. "Resilience is always a big concern and when it came to real deployments, we had questions about who is responsible for what," said Yves Bellego, Orange's director of spectrum strategy and planning, in a discussion with Light Reading at MWC. (See DT Demands Automation, Cloud Tech From Pan-Net Suppliers and Is Orange's New Deputy CEO the NFV Champion It Needs?)

Park Jin-hyo echoes those remarks. "An operator could buy hardware from HP [Hewlett Packard Enterprise] and software from Ericsson, but if there are some issues, then who is in control?" he says.

SKT reckons the use of new vRAN vendors could be cheaper than relying on traditional products from Ericsson or Nokia, provided it can address the concerns of its operations department.

"The infrastructure itself could be very cheap but there is a processing cost," says Sungho Jo, a senior director at SKT's network technology R&D center. "I think that in five years, there will be more virtualization in 5G compared with 4G, but it has to be based on guarantees around security and safety."

For more NFV-related coverage and insights, check out our dedicated NFV content channel here on Light Reading.

A further, 5G-related issue for SKT is whether the kind of open hardware that comes with virtualization would measure up in terms of performance.

"The big difference with 5G is low latency, which requires a lot of processing power," says Park Jin-hyo. "With open hardware based on FGPA [field-programmable gate arrays], I am not sure we can guarantee this kind of low latency and powerful processing."

"This means there are some limitations compared with the core network, where we can use common hardware from HP or Cisco," he adds. "In this [RAN] area, you need specialization for processing power."

The concerns of Orange and SKT will serve as a reality check given the hype surrounding vRAN technology at the recent MWC show. (See Major Telcos Pool Efforts to Slash 5G RAN Costs.)

In one of the week's biggest news announcements, several Tier 1 operators clubbed together to form the ORAN Alliance, which says one of its goals is to "maximize the use of common, off-the-shelf hardware" in the RAN. (See Major Telcos Pool Efforts to Slash 5G RAN Costs.)

A member of that group, Orange says this goal covers the use of white box technology in radio units. "The reason is to ensure the openness of the RAN, and to desegregate it so that we can have more flexibility in vendor selection," said the French operator in comments emailed to Light Reading.

ORAN essentially unites two existing groups focused on RAN virtualization -- xRAN, whose members included AT&T Inc. (NYSE: T), Deutsche Telekom and SKT, and the C-RAN initiative backed by China Mobile.

But with the emergence of new industry groups, there is the usual concern about the fragmentation of standardization efforts. Vodafone is reported by Fierce Wireless to have steered clear of ORAN so that it can focus its resources on the similarly named OpenRAN initiative, which forms a part of the Facebook-led Telecom Infra Project (TIP).

ORAN is keen to play down concern about the emergence of another group. "ORAN's operator leadership and specific focus will enable it to make the rapid progress necessary to accelerate the transformation of 3GPP-based mobile RAN technology," it said in a statement. "TIP has a broad focus on telecommunication infrastructure projects, extending beyond 3GPP technologies. It is driven by a broad mix of companies."

— Iain Morris, News Editor, Light Reading

kuntal0 3/9/2018 | 9:08:22 AM
Addressing the concerns expressed in this post It's important to clarify the concerns expressed in this post.

First, use of CoTS hardware to run RAN baseband application isn't any different than a data center running mission critical applications such as Stock Exchange and Banking. At a data center, the hardware issues are resolved with commonly used tools and well-established faulty hardware replacement mechanisms used by companies such as Nasdaq, Citi Bank, Google, Facebook and Amazon to name a few. On top of it, the robustness that comes with NFV with auto failover, live migration, scale up & down and in & out makes vRAN much more reliable and fault tolerant than today's custom hardware based legacy RANs can ever provide. In case the operators like SKT want to implement such high availability platform for RAN, it will be stupendously costly to implement on a custom hardware & custom platform centric RAN. Operators like SKT should be moving towards a highly available and more robust infrastructure than what is available from the legacy RAN vendors.

Second, the end-to-end integration and delivery of the vRAN solution is the job of a system integrator. It's highly unlikely that the operator will be left to fend for itself to integrate the NFV/SDN platform and applications like vRAN and vEPC. There are very capable SIs stepping up to play this role, e.g. Tech Mahindra, Amdocs and Italtel.

Third, the use of NFV doesn't rule out implementation of RAN protocol stacks requiring extremely low latency. There are very capable telecom-grade NFVIs out there that can support applications requiring micro seconds of latency in a very predictable way. vRAN solution, by default doesn't require virtualization of all layers including PHY. However, NFVI solutions exist in the commercial market today to provide the ultra-low latency that LTE & 5G PHY and MAC require for all use cases including uRLLC.

Fourth, use of FPGA does not introduce more latency than some custom-built processor. ALL of the 5G NR products in 2018 - 2019 timeframe will be with FPGA based PHY. Only in the later part of 2019 and 2020 onwards, we will see some ASCI and SoC based PHY products. Hence, it's rather confusing to state that FPGA is an issue.

SK Telecom is highly respected for its leadership role in bringing transformative solutions to the telecom space. I sincerely hope that the concerns about virtualization of the RAN infrastructure that SKT expressed in this post get addressed in a rational way, and the operators such as SK Telecom don't get left behind due to the FUD.

Sign In