For someone who "grew up" in the Internet/dot.com era (1993-1997ish), where men were men and the fastest commercial Internet backbone at the beginning of the '90s ran at T-3 (45 Mbps) speeds over fiber, it's alternatively humbling and cool to be on a 20 Mbps home cable connection with 100 Gbps optical network links slowly and steadily being turned up around the world. Just how fast can the world go in the next decade?
Moving from 45 Mbps to 100 Gbps in two decades is something a lot of people under the age of 30 most certainly take for granted. The wireless-obsessed press certainly takes it for granted after the fiber bust/glut a decade ago, but without optical networking providing high-speed transport, the wonder of the mobile world comes to a screeching halt - no Apple iPhone, no tablets, no LTE.
If anything, core network speeds are increasingly more important as consumers and businesses alike ratchet up speeds. The cable industry has a clear path to delivering multi-Gigabit speeds using DOCSIS 3.1, with an upper limit of up to 10 Gbps if everything is tweaked correctly. Back in 2010, Verizon demonstrated the ability to deliver up to 10 Gbps over existing infrastructure in Boston - an Elks Lodge, as a matter of fact. Verizon has more recently discussed delivering up to 1 Gbps via FiOS and those crazy guys at Google are conducting a few consumer deployments of gigabit fiber service.
For bandwidth hungry businesses, this is all good news. Faster consumer speeds at roughly the same or lower pricing ultimately translates to faster business speeds at better price points, driven in part by competition between cable and incumbent telco providers.
In some cases, businesses are encouraged to move to faster speeds in order to ditch legacy connections and services built upon copper-based hardware. Both AT&T and Verizon want out of copper as fast as possible for a number of reasons: it's expensive to run two separate networks; maintaining and repairing legacy gear becomes more costly every year, especially as the guys who run them retire (think of it as telco's COBOL problem); and last mile copper plant can't deliver the speeds needed to compete with higher-speed cable company offerings.
In the near term, 400 Gbps per link/wavelength seems to be the high-speed "standard" falling into place. Orange has a 400 Gbps network link running between Paris and Lyon, while numerous companies have issued press releases and papers discussing 400 Gbps experiments and trials. Expect current 100 Gbps circuits to migrate to 400 Gbps as the latter becomes the new high-speed core network speed standard.
Where life gets interesting is if you start thinking about the number of 10 Gbps users you can theoretically support on a consumer or business network, with business customers expecting measureable quality of service (QoS) guarantees for both speed and reliability. Consumer service tends to be pretty much an on-or-off thing, with credits only offered for outages lasting for 8 hours or more with most people assuming they are getting the speed they are paying for.
The Holy Grail for speed freaks is terabit Ethernet (TbE), followed by 100 Terabit Ethernet. A study group at IEEE has been formed to look at TbE, but going at terabit speeds may require throwing out the existing templates of doing business. Researchers at the University of California, Santa Barbara Terabit Optical Ethernet Center (TEOC) want to put 1 Terabit Ethernet over optical fiber in 2015 - that's about two years away - with enabling technology to reach 100 TbE by 2020.
Part of the pitch for TbE is a heavy emphasis on photonics, keeping everything in light as long as possible for better energy efficiency. Moving to traditional electronic circuits means more energy used, which means more heat generated and more cooling leading to more power consumption for the whole system.
Whenever the world moves to TbE, it will drive faster speeds and the need for upgrades within WAN networks and data centers. It may also drive improvements beyond traditional glass optical fiber. Defense Research Projects Agency (DARPA) and researchers at the University of Southampton have, in separate projects, created a hollow-core photonic-bandgap optical fiber. The new fiber gets rid of the glass fibers at the core of a cable - maybe not really good growth news for Corning - and uses a cladding that reflects light. Through glass, light travels 31 percent s-l-o-w-e-r than through air or a vacuum. No glass, faster transmission speeds.
A demonstration by Southampton demonstrated the ability to transmit data at 73.7 terabits per second at lower latency than glass across short distances. DARPA's hollow-core tech is designed to be applied for highly-accurate gyroscopes, but telecommunications companies and high-frequency traders have to be looking at the technology today to increase network speed.