Loading...
You are here:  Home  >  Data Education  >  Data Architecture News, Articles, & Education  >  Data Architecture Blogs  >  Current Article

The Next Step in Data Routing: Sharing Information Between the Control and Data Planes

By   /  June 6, 2018  /  No Comments

Click to learn more about author Gur Shatz.

In a few short decades, the Internet evolved from an obscure way for the military and academics to communicate to the dominant network for the world’s business and entertainment industries. A lot has changed, especially when it comes to how routers – the beating heart of the Internet – operate.

One big change has not yet been made, however. Data planes and control planes continue to remain separate, even more so with the adoption of Software Defined Networking (SDN). While that might have been effective in the ‘80s, when application requirements were not all that different from one another, today it’s another matter.

Capacity needs vary significantly; driven in part by video and other demanding applications and services, real-time applications, such as voice and video conferencing, and other loss-sensitive applications.

There needs to be a way to treat these applications uniquely.

What is the solution?

Bring data planes and control planes closer together.

The Internet: Scalable but Not Terribly Smart

It’s a common challenge: The limitations of a system today are due to decisions made in an earlier era. In the case of data routing, the limitations were in access speeds, processing power, and memory. These all had a fraction of the power of what they have today.

In the beginning, router data and control functions were combined. As the Internet grew, designers met increasing demands for packet processing with custom hardware. This enabled billions of IP packets to keep flowing. Routing decisions were moved to a separate control plane, which only sent the data plane the information it absolutely needed, such as updates to the routing tables. This opened the door to the deployment of massive amounts of computing power to the data plane.

This was a necessary step. However, despite the fact that computing power has radically increased, the same basic paradigm remains: The control plane does the heavy lifting, and the data plane dutifully does what it is told to do.  There is little information sharing between the two planes.

Expanding the dialog, so to speak, will create greater efficiencies. Today, the control plane has no special insight into network conditions. Did the packet reach its destination? Is the router working correctly? Is the router that the routing table suggests for a packet’s next destination overwhelmed – while one just a tad farther away is open and ready for business? Since there is no meaningful information exchanged between the data and control planes, real-world conditions don’t play a role in what routing decisions are made.

Think of it as two versions of GPS: one in which the GPS system only considers the distance to a location and another that accounts for distance and real-time road conditions.

Make the Internet Smarter

The irony is that SDN and OpenFlow separate the control- and data-planes even further, moving the two domains to different machines. The lack of interaction may not be a big deal in a highly controlled environment, such as datacenters, but crossing the globe is very different than crossing a country.

Enterprise networks need to bring the control- and data-planes closer together — not further apart. By converging the two, we can route packets more intelligently, selecting the optimum path for each application based on what’s happening right now on the network. Voice traffic gets sent across paths with less packet loss, and file transfer traffic goes across paths with more capacity. Enterprises that are serious about using the Internet for their global networks need to solve the problem themselves.  Expecting a new update to BGP protocol or a change in ISP behavior is unrealistic.

It’s what SD-WAN allows. Connecting locations into an SD-WAN overlay allows organizations to overcome the limitations of the underlay. SD-WAN uses real-time latency and loss statistics to select the best available path for a given application. So just because an ISP advertises a path, it doesn’t mean traffic will be sent along that path. The SD-WAN may route traffic across an alternate network.

It’s how Cloud-based SD-WAN providers architect a global backbone; content delivery networks (CDNs) do something similar. This is no surprise, really. Sharing information is the next step for routing, after all.

About the author

Gur Shatz is co-founder and CTO of Cato Networks. Prior to Cato Networks, he was the co-founder and CEO of Incapsula Inc., a Cloud-based web applications security and acceleration company. Before Incaspula, Gur was director of product development, VP of engineering and products at Imperva, a web application security and data security company. Gur holds a BSc in computer science from Tel Aviv College.

You might also like...

Data Science in 90 Seconds: K-Means Clustering

Read More →