• United States

Scaling networks for the web-scale effect

Mar 11, 20155 mins
Cloud ComputingNetworking

The network is adapting to the new web-scale realities of web traffic.

The world has gone web-scale crazy, but what does web-scale mean when we apply the term to networking? Isn’t the network how we got the web in the first place? Unfortunately, many of today’s networks leave a lot to be desired. For decades now, we’ve basically designed, built, and operated networks using a provision and monitor paradigm. WAN connections can take days, weeks, even months to correctly provision. If your connectivity needs were complex, timeframes could be even longer, if the desired services were available at all. Contrast this antiquated networking model with the real-time, on-demand environment that we have created for ourselves with applications, content, and businesses that can “go viral.”

Gracefully supporting “going viral” is what web-scale networks are all about. A web-scale network is one that is agile enough to get big, fast. Get big in terms of the numbers of connection points, as well as in terms of the size and topology of those connections. And do it in real-time, not in days, weeks, or months.

Some web-scale principles are already being applied in the networks run by providers of search, social networking, and web services such as Google, Facebook, Microsoft, and Amazon, but web-scale has now grown to become one of the principle driving forces for the whole telecommunications marketplace. While a number of these internet content providers are building networks, their internal connectivity needs are often focused on connecting their data centers or content centers together so that they can rapidly create, replicate, and distribute content. The critical connections between the content users and these content centers are more often than not provided by traditional telecom service providers. Carrier-neutral providers have emerged to provide co-location and interconnect functionality for data and content centers, and there is a growing ecosystem of operators offering data center and Ethernet exchange services.

The Web-scale Effect

In the past, providers could easily predict how much capacity they needed in the network, and when they needed to add it. After all, the traffic was really only comprised of emails and website browsing, not typical drivers of viral growth. Network demands were relatively predictable.

Then, the Googles, Facebooks, and Amazons of the world jumped on the scene, and new streaming video businesses such as YouTube and Netflix emerged and effectively reinvented the way services are delivered.

These web-scale giants, among others, have been driving the rapid adoption of cloud services, on-demand networking, and virtualization, dominating the networking landscape. And it’s relatively unknown that roughly 50% of all transatlantic traffic is now web-scale driven, in part because these players purchase large amounts of network capacity from traditional carriers. Given their continued growth, they may eventually come to dominate the traffic on submarine networks.

Because these web-scale players are able to deliver services so smoothly, businesses and end users expect connect, compute, and storage services to be available 24/7. But what is the secret to their success? Simply put, it is how they have evolved the architecture of the network to be more real-time and on-demand. They have focused their networks on two functions: one which interconnects their data centers, connecting content to content, and another which connects users to that content. Both portions need to embrace open networking concepts to maximize flexibility and interoperability. Application Programming Interfaces (APIs) are the desired model for communicating with networking equipment, and web-scale drives the network to becoming a programmable platform that can be orchestrated and controlled in real-time rather than just a bunch of provisioned static pipes.

Web-scale networks will ultimately become the standard that users come to expect. Most of us are already accustomed to the benefits of the on-demand virtualized app-driven world, and our acceptance of it has driven us to expect this same type of experience in the networking world as well.

This expectation – fair or not – is driving significant change in network architectures that leverage the dynamics inherent in web-scale networks: virtualization, openness, integration, automation, and massive scale. These dynamics are now being broadly applied in today’s IT environment in the form of private and hybrid cloud deployments, and now network operators of all types are following suit.

When we refer to the web-scale effect, we are talking about adopting the lessons learned from the leading web-scale networks and how they are architected, deployed, and operated. This is driving a shift in how ALL networks are designed and built. It’s why all types of providers across the globe – regardless of size – are looking at adopting the fundamental dynamics of web-scale networks – ensuring that they are scalable, fast, and agile.

Unpredictability is the New Normal

For network operators, the predictability of scalability requirements is long gone. Networks now need to scale in real-time, on-demand, without glitch, and accommodate all of the dynamics that make services like 4K streaming video a reality, small cells easy to deploy, and connections from users to content simple and reliable.

Clearly, web-scale is not a market; it’s actually the market driver. As the cloud continues to grow, the roles of service providers and the web-scale community are increasingly interdependent and dynamic. What started with data center operators has become mainstream thinking in major enterprise environments, and it’s now driving changes in service provider decision making, too.

Thankfully, there are technology advancements that help solve these specific problems for network providers – like more agile, even lower-cost metro Nx 100G, and network architectures that utilize open principles to allow unprecedented scale in both physical size and programmability. These new innovations allow all types of providers to build bigger, more agile networks that can gracefully respond to the unforeseen demands and the viral nature of today’s web-scale world.


With more than 20 years of telecom experience, Mr. Alexander is currently serving as Ciena’s Senior Vice President and Chief Technology Officer. Mr. Alexander has held a number of positions since joining the Company in 1994, including General Manager of Ciena's Transport & Switching and Data Networking business units, Vice President of Transport Products and Director of Lightwave Systems.

From 1982 until joining Ciena, Mr. Alexander was employed at MIT Lincoln Laboratory, where he last held the position of Assistant Leader of the Optical Communications Technology Group. Mr. Alexander is an IEEE Fellow and was the recipient of the IEEE Communications Society Industrial Innovation Award in 2012. He is currently an Associate Editor for the IEEE / OSA Journal of Optical Communications and Networking. He has served as a member of the Federal Communications Commission Technological Advisory Council, as an Associate Editor for the Journal of Lightwave Technology, as a member of the IEEE / LEOS Board of Governors, and was a General Chair of the conference on Optical Fiber Communication (OFC) in 1997.

Mr. Alexander received both his B.S. and M.S. degrees in electrical engineering from the Georgia Institute of Technology. He has been granted 18 patents and has authored a text on Optical Communication Receiver Design as well as numerous conference and journal articles.

The opinions expressed in this blog are those of Steve Alexander and do not necessarily represent those of IDG Communications, Inc., its parent, subsidiary or affiliated companies.