The Road to 400/800 Gig is Paved!

In this edition of Voices of the Industry, Gary Bernstein, RCDD, CDCD, Global Data Center Solutions Specialist for The Siemon Company, shares his insight on the outlook and feasibility for data center migration to 400 and 800 Gig.

Gary Bernstein, RCDD, CDCD, Global Data Center Solutions Specialist for The Siemon Company

A decade ago, enterprise data center managers were looking at infrastructure requirements to prepare for a migration from 1 to 10 Gig for server connections and 10 to 40/100 Gig for uplinks. With many just now starting to make that leap, 400 and 800 Gig may seem a bit farfetched anytime soon. However, trends in digital transformation, bandwidth-intensive applications, and the need to cut cost have paved the road to 400 and 800 Gig.

The Need is There

With 92% of enterprises adopting a multi-cloud strategy, cloud data centers have been a primary driver in the need for 400 and 800 Gig speeds. The COVID-19 pandemic with work-from-home strategies and consumer demand for streaming and online retail has further fueled cloud adoption, with cloud spending increasing 37% percent during the first quarter of 2020. While more data is moving to the cloud, the size of that data is also increasing, and emerging applications are requiring more bandwidth to support lower-latency transmission.

Large enterprise customers are generating increasing amounts of IoT/IIoT data and adopting technologies like artificial intelligence, machine learning and advanced data analytics that require extreme low latency and 50 and 100 Gig server connections in cloud data centers. Telehealth and high-definition MRI images in healthcare and high-frequency trading and online banking in finance are further driving bandwidth demands and the need for near real-time transmission. In the professional media industry, uncompressed high-resolution video, computer animation and visual effects are also driving file sizes to an all-time high. While one hour of uncompressed raw standard definition video content requires only 13 Gig of storage, that same video content at emerging 8K resolution amounts to more than 7 Terabits. Large cloud data centers are also adopting data center interconnect (DCI) technology that connects multiple dispersed data centers and super-spine architecture that connects multiple leaf-spine networks, both of which increase the size of data sets. 5G buildouts, next-generation virtualization and cloud-native applications are also driving the need for 400 Gig in edge data centers and central offices.

While few enterprise applications and workloads currently require anything beyond 100 Gig uplink speeds, enterprise data centers too are driving the need as many are realizing that using new 400 Gig switches to breakout to multiple server connections offers a significant reduction in the cost per port.

The Route is Navigable

Enabled by newer encoding technology, the IEEE has already released several standards for 400 Gig over both singlemode and multimode fiber. The IEEE Beyond 400Gb/s Ethernet Study group is now working to define physical layer specifications to support 800 Gig and potentially 1.6 Terabit speeds. The current 800 Gig objectives include support over eight lanes of multimode to 100m and eight lanes of singlemode to 500m.

While the market viability of specific 800 Gig applications and transceivers will become clearer over the next two to three years, Arista, Cisco, and Juniper already have transceivers available for the 400GBASE-DR4 parallel optic application that uses eight singlemode fibers (i.e., four lanes) to a distance of 500m and the 400GBASE-FR4 wave division multiplexing application that uses four wavelengths per fiber over two singlemode fibers (i.e., 100 Gig per wavelength). While almost all cloud data centers use singlemode, most enterprises still deploy multimode cabling. The 400GBASE-SR4.2 multimode application therefore has excellent market viability for enterprise data centers enterprise. This application uses a combination of short-wave division multiplexing and parallel optics with two wavelengths per fiber over eight fibers of OM4 multimode (i.e., 50 Gig per wavelength) to 100m.

So what does migration look like? Enterprise data centers are currently running 1 or 10 Gig server speeds and 10 or 40 Gig uplink speeds with the option to eventually move to 25 or 50 Gig for servers and 100 or 400 Gig for uplinks. Cloud data centers are currently at 10 or 25 Gig server speed and 40 or 100 uplink speeds, with many already looking to increase their server speeds to 50 or 100 Gig with 200 or 400 Gig uplinks.

Both 400 and 800 Gig offer broad market potential for both enterprise and cloud data centers due to cost-effective aggregation that enables connecting a single 400 Gig switch port to break out to down eight 50 Gig servers or to four 100 Gig servers, with cloud data centers eventually connecting a single 800 Gig switch port to break out down to eight 100 Gig or four 200 Gig servers. Both applications are also highly achievable due to leveraging existing MPO connectivity and singlemode and multimode fiber. It is therefore recommended that enterprise data centers utilize a Base8 MPO OM4 cabling solution to ease migration to 400 Gig, while cloud data centers will likely utilize a Base8 MPO singlemode cabling solution for easy migration to 400 and 800 Gig and beyond.

Gary Bernstein, RCDD, CDCD, is Global Data Center Solutions Specialist at Siemon with more than 20 years of industry experience and extensive knowledge in data center infrastructure. Established in 1903, Siemon is an industry leader specializing in the design the design and manufacture of high quality, high performance IT infrastructure solutions and services for Data Centers, LANs, and Intelligent Buildings. Connect with Gary on LinkedIn.