Are Stress Fractures Threatening the Internet?

By Pam Baker

(Back to article)

Given that the Internet has become the cornerstone for businesses large and small, there is little wonder that any perceived threat to it is due cause for alarm. Now that worldwide demand for this all important resource is rocketing skyward, businesses are beginning to wonder if the Internet can withstand the g-forces.


A recent study by Nemertes Research claims that the increased demand “is pushing against the physical and logical limitations” and thereby stressing the Internet. Specifically, the study finds that Internet demand is outpacing growth in network capacity at the access layer by a wide margin, and that IP addresses are “quickly depleting.”


Other industry reports seem to support this conclusion. The current number of global broadband subscribers now stands at 400 million. This represents a 600,000% increase since 1998. Considering there are some 6.5 billion people on the planet, the Internet is far from reaching peak use. But is this truly a worry business folks should lose sleep over? Not really.


“None of this means the Internet will abruptly stop working,” said Ted Ritter, research analyst with Nemertes. "Instead, the slowdown will be in the area of innovation. Ultimately, access bandwidth limitations will hamper deployment of next-generation applications."


Even if the Internet itself can remain unscathed in the face of a bruising increase of use, doesn’t a slowdown in innovation and access still have the potential to cripple businesses? Again, not really. As Stu Elby, VP of Network Architecture at Verizon, put it: “IP stands victorious." Elby explains that over the last decade, enterprises have adopted IP as the primary convergence mechanism to handle all mission critical machine-to-machine (M2M) and client-server applications.


More recently, IP is being used as the primary person-to-person communications mechanism (voice and video) within the enterprise. And early on IP adopters quickly found out the Internet could not provide the level of availability or real-time performance needed to meet mission critical requirements.


To work around these deficits, Internet operators have built carrier grade IP networks either as totally independent networks from the public Internet or as a logical partitioning of some the router infrastructure used by the public net. These carrier grade IP networks do provide all of the latest performance and capabilities enterprise customers want; but at price points commensurate with the added features—typically much higher than basic Internet access prices. Enterprise IP and IP/MPLS services (e.g., IP-VPN) offered on these networks have been tremendously successful, for example.  


“The alleged architectural limits of the Internet are actually more related to the access infrastructure, for example, xDSL and cable modems networks, that is expensive and time consuming to upgrade, rather than to the core Internet architecture that, on the contrary, is very scalable and relatively inexpensive to upgrade compared to the access infrastructure,” said Paolo Gambini, chief marketing officer at Tiscali International Networks (TINet), a wholesale provider of global IP/MPLS network services,.


In effect, this means that the Internet is not suffering any stress fractures, but bottlenecks are forming at the points where users connect. But even these are fixable. “The business case for further network investments is the real challenge for access and content providers alike,” said Gambini, which, of course, underscores Elby’s point on cost restrictions. So, much of the problem lies in “the last mile” at the end of which is any given business’ customers.


Ben Zhao, assistant professor at the University of California, Santa Barbara, said there is a significant shift in bandwidth usage from the core to the last mile. Internet service providers (ISP) that have traditionally relied on statistical multiplexing to offer much more bandwidth than physically available, are now finding themselves overwhelmed by traffic demands from bandwidth hungry applications, including peer-to-peer (P2P), streaming video and remote PC access.

Most of these are application level traffic that does little to optimize for bandwidth or location, and ISPs are paying the price. Their response has been “to throttle and filter traffic, particularly P2P traffic, with limited success.”


Access bandwidth is already hampering current applications and video streaming. It's only likely to get worse as video streaming to the home takes off and becomes a major alternative to home media systems. Also, a rapidly growing dependence on cloud-based applications such as GoogleDocs and Amazon's S3 are only going to exacerbate the situation, as more and more computation moves from the home PC to the cloud, “making the access bandwidth even more important as a potential bottleneck for all things Internet,” concludes Zhao.


Here is where the illusion that the Internet is fracturing comes in: as people see more private networks forming they believe it is an indication that the public network is overloaded. 


“Large financial services firms, for example, where milliseconds of latency make a huge difference; these exchanges don’t want their customers’ billion dollar trades to get held up on the Internet by download requests for a new Coldplay song on iTunes or viewings of the latest viral video on YouTube, explains Ray Patalano, senior manager of Enterprise Solutions at Ciena.


“The need to maintain optimal, predictable, consistent high-quality network performance is essential in financial services, as well as other markets. That’s what private networks deliver and why businesses are building them instead of running those applications over the public Internet.”


Eventually, though, private networks are likely to fold back into the larger public network architecture.


“The growth of the ‘private’ IP networks to support enterprise customers is not an indication that the Internet architecture is broken or reached its limit,” said Verizon’s Elby. “Rather, it is recognition that the business models for enterprise and public Internet are quite dissimilar. With the emergence of consumer applications that require enterprise-like performance—HD video over the Internet, for example—we may find the private IP networks and public Internet converging in the future.”