400GE is Revolutionary, Not Evolutionary

2018-08-23 | 4 min read

Next generation optical transceivers are expected to use less power per gigabit, be less expensive per gigabit, and operate at four times the speed of 100GE transceivers. It is rather puzzling how 400GE transceivers will meet all these expectations. In fact, the move from 100GE to 400GE in the data center is revolutionary, not evolutionary.

It took 15 years for data centers to evolve from 10GE to 100GE. Data centers began implementing 100GE in 2014, yet full build-outs only became cost-effective over the last couple of years thanks to the availability of more affordable optical transceiver modules. Emerging technologies enabled by fifth generation wireless (5G), such as artificial intelligence (AI), virtual reality (VR), Internet of Things (IoT), and autonomous vehicles, will generate explosive amounts of data in the network. Current 100GE speeds available in data centers will not be able to support the speeds and data processing requirements needed by these technologies. As a result, data center operators are looking to evolve their networks from 100GE to 400GE.

There are three key challenges that need to be addressed to make the transition from 100GE to 400GE as smooth as possible:

Challenge 1: Increase Channel Capacity

According to the Shannon Hartley theorem, there is a theoretical maximum amount of error-free data over a specified channel bandwidth in the presence of noise. Therefore, either the channel bandwidth or the number of signal levels must be increased to improve the data rate or channel capacity to reach 400GE speeds.

Challenge 2: Guarantee Quality & Interoperability

As new 400GE transceiver designs transition from simulation to first prototype hardware, engineers face the challenging task of developing a thorough, yet efficient, test plan. Once deployed in data centers, marginally performing transceivers can bring down the network link, lowering the overall efficiency of the data center as the switches and routers need to re-route the faulty link. The cost associated with failed transceivers once deployed in the data center is enormous. Since large hyperscale data centers can house more than 100,000 transceivers, even a small one-tenth of one percent failure rate would equate to 100 faulty links.

Challenge 3: Reduce Test Time, Reduce Cost

Keeping the cost of the optical transceivers low is a high priority for data center operators. To be competitive, transceiver manufacturers must find ways to drive down production costs. Like most new technologies, the price of next-generation optical transceivers tends to drop sharply after introduction to the market, and development costs amortize as volume ramps. Test time contributes significantly to overall transceiver cost. Therefore, more efficient testing of the broad range of transceiver data rates accelerates innovation and lowers cost.

The Next Test Challenge

Many data center operators are moving to virtualized networks using software-defined networking (SDN) and network functions virtualization (NFV). They need full network test of Layers 2-7, including SDN/NFV validation and traffic loading, to ensure that traffic flows through the virtualized network as expected. This is the next challenge data center operators will need to overcome.

400GE in data centers will soon become a reality. Find the solutions to address these 400GE transceiver test challenges here.