Monterey is first Ethernet switch-chip specifically for fronthaul

Broadcom may have had its hopes of acquiring Qualcomm dashed by President Trump’s intervention (see separate item), but it continues to eye other ways to expand its presence in the telecoms market. One is the launch of its Monterey Ethernet switch, which it claims is the first to be designed specifically for mobile fronthaul – the extremely high performance, low latency links between centralized baseband units and remote radio units in a virtualized RAN.

This is a strategic area for a chip company to target, especially one as strong in Ethernet switch-chips as Broadcom – indeed, one of its possible consolation prizes in the wake of its Qualcomm bid being blocked could be to bid for another Ethernet player such as Mellanox. The performance demands of fronthaul connections – and the expensive fiber/CPRI combination those typically require – have been a barrier to adoption of vRAN to date. However, one factor which could help reduce the caution of MNOs about vRAN, and speed up deployment, would be availability of high performance fronthaul solutions based on Ethernet.

Ethernet fronthaul solutions have been emerging over recent years, promising flexibility, greater multivendor openness and therefore better pricing than CPRI. However, there have been question marks over their performance, compared to CPRI, in very demanding environments. Switches like Monterey will aim to fill that gap, by promising technology at the top end of the Ethernet range, optimized particularly for fronthaul processes and requirements.

The IEEE has developed standards for Ethernet fronthaul in the 1914 family, which includes 1914.3 for radio over Ethernet (RoE) encapsulation and mapping. Importantly, as well as a native mode, this also supports two modes that enable integration with CPRI, the dominant specification for linking shared basebands to remote radio units. CPRI has proved expensive and inflexible, though it has now been joined by an Ethernet offshoot, eCPRI, geared to flexible vRAN fronthaul. However, eCPRI was developed by the same closed group of large vendors as its predecessor, and has the same key drawback – that each vendor’s implementation is slightly different, making multivendor interoperability very difficult, and so damaging the economics of vRAN.

Before the recent IEEE developments, Ethernet was unable to support synchronization and jitter requirements for CPRI-to-Ethernet bridging and so was confined to relatively small vRAN designs. However, products like Monterey now support IEEE’s new 802.1CM (Time-Sensitive Networking for Fronthaul) standard which addresses this issue, giving operators the option of an end-to-end Ethernet architecture for cellular networks, for the first time.

Bob Wheeler, principal analyst at The Linley Group, told EETimes: “The industry is moving away from point-to-point CPRI radio links and towards a switched Ethernet infrastructure based on new protocols like eCPRI and IEEE 1914. Broadcom has developed a unique solution by extending its terabit Ethernet switch to address this new radio fronthaul application by adding support for Ethernet-based 5G radios as well as installed CPRI-based LTE radios.”

The company has also announced the latest switch-router system-on-chip products in its StrataDNX family, including Jericho2, which it says is the industry’s highest bandwidth merchant chip in this category, and the FE9600. This is all part of Broadcom’s ongoing effort to woo network infrastructure makers away from custom ASICs and towards merchant silicon, by promising almost as much performance and customizability, at lower cost.

Jericho2, it says, delivers five times higher bandwidth at 70% lower power per gigabit than its predecessor, with 10Tbps rates supporting industry standard 400GbE, 200GbE and 100GbE interfaces. It works with Elastic Pipe packet processing and large scale buffering plus integrated High Bandwidth Memory (HBM), to deliver “disruptive economics” to the sector.

“The Jericho franchise is the industry’s most innovative and scalable silicon used today in various switch-routers by leading carriers,” said Ram Velaga, general manager of switch products. “Jericho2 will accelerate the transition of carrier grade networks to merchant silicon based systems with best-in-class cost/performance.”

“OEM ASICs dominated the router market for years as merchant silicon fell short of the features and performance required by this demanding market,” said Wheeler. “Jericho helped open the door for merchant silicon in routers. With the availability of the 10Tbps Jericho2, OEMs will find it more difficult to justify the massive investment required to develop new router ASICs.”

The company is also shipping FE9600, a fabric switch device with 192 links and 9.6Tbps fabric capacity. This claims over 50% reduction in power per gigabit compared to its predecessor.

AT&T has been a key driver behind the shift from very specialized architectures towards white box switches based on merchant silicon. Broadcom, together with Barefoot Networks, have provided chips for its conceptual white box switches, and the operator endorsed the new product too. Chris Rice, SVP of Domain 2.0 architecture and design, said: “At AT&T we are fully committed to leading the industry towards a new networking paradigm, one that delivers faster networks, quicker innovation cycles and lower capex and opex, while most importantly, staying ahead of our customer needs. Merchant silicon, as well as hardware/software disaggregation, are instrumental to this future. Broadcom’s Jericho2 10Tbps router chip is in line with this vision and could enable AT&T to accelerate this transition.”