Drivers of the optical revolution - Challenges in optical backplanes, Part 3

Despite projections for almost two decades that network elements are on the verge of ‘going optical,’ metallic interconnects have continually advanced to suit bandwidth demands. To date, optical backplane technologies have struggled to reach mass-market deployment due to cost constraints, but major upticks in data traffic expected over the next five years may be enough to finally bring optics to parity.

Data from research firm IDC suggests that by 2020 increased smartphone usage, Machine-to-Machine () connectivity, and the Internet of Things (IoT) will converge to produce roughly 50 billion connected devices, requiring an overhaul of the current network infrastructure (Figure 1). Combined with progressions of Moore’s Law, these factors will generate the need for advanced backplane technologies with the transmission capacity to serve emerging network trends, says Chuck Byers, Technical Leader and Platform Architect, Enterprise Networking Group, Cisco Systems, Inc. (www.cisco.com).

21
Figure 1: Increased data traffic driven by the Internet of Things (IoT) and growing smartphone usage will drive the need for advanced networking technologies in the years to come (Sources: 1: IDC, Intel, United Nations; 2: IDC Digital Universe Study, December 2012; 3: McKinsey Global Institute*, 2013).

“Looking at a time horizon of 5-10 years, we are approximately five Moore’s Law cycles away,” Byers says. “I would expect roughly a 32-fold improvement in the capabilities of processors, storage engines, and interconnect systems – both on the backplane in the box and between boxes. This is considered quite attractive for computing; storage systems for big data; UHD video engines; next-generation architectures; and other large-scale, high-bandwidth applications.

“According to the Cisco 2013 Visual Networking Index (VNI) report, by 2017 global mobile data traffic will reach 11.2 exabytes per month, growing 12-fold from 2012 to 2017 [1]. This implies that the nodes processing that traffic need to grow in capacity at similar rates,” Byers points out. “Today a good HDTV stream can be to about 6 Mbps; in 5-10 years we are all going to expect 4K/8K UHDTV, which could take up to about 120 Mbps per compressed stream. This 20-fold increase will need to be accommodated in all the servers, storage engines switches, routers, and service nodes in the video network (all of which could use xTCA). It is possible that could be valuable to accomplish this. The IoT is also going to create an uptick in M2M traffic by a couple orders of magnitude over this timescale, also requiring a massive increase in box bandwidth. Eventually, advanced backplane technologies (including optics) will be required to keep pace.”

Benefits of an optical revolution

As discussed in Part 1 of this series, most current systems don’t include provisions to support optical backplanes (’s Zone 3 connector being one exception), rendering the move to optical transmission a revolutionary step that requires serious cost-benefit analysis. Once deployed in the network, however, the advantage optical backplane solutions can afford is future-proof transmission capacity over many iterations of board technology. Using fiber optic arrays and Wave-Division (WDM) techniques, optical backplanes not only present a solution for over-burdened next-generation networks, but also enable new High-Performance Embedded Computing (HPEC) system architectures for the future, Byers says.

“The main benefit to consider with optical backplanes is that once the basic set of interconnect fibers are in place on the backplane, future generations of boards can crank up the interconnect bandwidth almost without bound,” Byers says. “More wavelengths can be lit up per fiber, and higher data rates can be used per wavelength.

“Short-reach components are capable of hundreds of meters and long-reach components can go hundreds or even thousands of km,” he continues. “For a backplane, the longest we would reasonably need would be a few meters. So, in that sense, fiber can be an overkill, and the components typically used for intermediate- or long-reach fiber interfaces are over-engineered for fiber backplane applications. 

“There is another route organizations can consider with the longer reach offered by fiber interconnect: enable multi-box systems,” Byers suggests. “If backplane signals can extend for 10 meters, this could place a dozen or more frames full of equipment all in the same optical backplane domain. Millions of processor cores could be included in a single system. However, there are flight-time latency concerns that would have to be accommodated in a high-level architecture and, for example, special protocols may be needed to account for transport delays over tens of meters round trip.”

For more information on Cisco’s involvement with optical technology, visit www.cisco.com/en/US/netsol/ns340/ns394/ns113/networking_solutions_packages_list.html.

  1. Read ‘Cisco 2013 Visual Networking Index (VNI) report– at www.cisco.com/en/US/solutions/collateral/ns341/ns525/ns537/ns705/ns827/white_paper_c11-520862.html
  2. Read ‘“5 years out” – Challenges in optical backplanes, Part 1’ at xtca-systems.com/articles/5-backplanes-part-1/.
  3. Read ‘Fiber, free space, or silicon? – Challenges in optical backplanes, Part 2’ at xtca-systems.com/articles/fiber-backplanes-part-2/.