Optical components

400G Hits Critical Mass; Deployments Coming

While the number of 400G products is still small, critical mass has been achieved; there is enough commercially available for companies to start developing and deploying initial systems, Heavy Reading reports.

The needs of the core, metro and access networks have all provided impetus for technology development and adoption, but the trends toward cloud computing and network virtualization have helped make data centers incredibly influential when it comes to driving technology markets. Data center operators are at the forefront of the trend toward new variants of Ethernet (25G/50G/100G), faster speeds (200G, 400G) and new interconnect options.

Hyperscale data center interconnects (DCIs) and connections between switches within data centers are key applications for 400G, concludes a new study by Heavy Reading , which projects that server input/output (I/O) will be almost 5Tbit/s by 2020 at the latest.

In an overview of the market, analyst Simon Stanley observed there are several complete 400G solutions becoming available, including optical components, optical modules, coherent and PAM4/NRZ transceivers, OTN processors and FPGAs. The report notes that the first 400G modules and active cables are already being commercialized.

The cost of 100G connections fell significantly in 2015, largely associated with the shift to CFP and CPAK optical modules. Costs will continue to drop through next year, Stanley projects, now driven by a shift to QSFP 28 and CFP4 modules.

Want to know more about communications ICs? Check out our comms chips channel here on Light Reading.

Meanwhile, the costs of 100G in other applications, such as metro and long-haul, will be driven down by other technology advances, specifically the introduction of CFP2-ACO modules and the next generation of coherent receiver DSPs.

This will all naturally lead to increasing demand for 400G. Heavy Reading expects that demand to manifest first with data centers, partly for DCI between data centers and long-haul/metro applications, but also to connect top of rack (ToR) switches, which currently make use of 100G interconnects.

Currently, connections from ToR to individual servers in each rack tend to be dual 10G. While there are some that are still skeptical about 25G and 50G interconnects, Heavy Reading expects that today's dual 10G connections will start getting replaced by dual 25G connections next year, a trend that should continue into 2018. The performance gain will be accomplished with minimal increase in cost.

The report includes a survey of activity from 19 companies providing faster connectivity options and enabling technologies, including, Altera Corp. (Nasdaq: ALTR) (Intel) Broadcom Corp. (Nasdaq: BRCM), Macom and Xilinx Inc. (Nasdaq: XLNX).

— Brian Santo, Senior Editor, Components, T&M, Light Reading

Be the first to post a comment regarding this story.
Sign In