Linux and the Future Network

By March 23, 2015Blog

When Linus Torvalds released Linux in 1991, no one could have imagined the influence it would have on modern computing. It quickly took off in the web, HPC, and embedded spaces. Looking back, it’s easy to see the major advances like LAMP and Beowulf clusters all the way to modern cloud computing and forget the effort that individuals and companies put in to advance Linux. Those changes were primarily on the server side, and we’re starting to see the same thing happening in the network itself. The next generation network will require similar changes of individuals and companies.

Linux represents the nexus of open source, operating systems, and networking. Each of these contributes uniquely to the success of Linux, and each of them enables new opportunities for the future. For a much better explanation than I could give, read Ian Murdock’s (yes, that Ian Murdock) “Open source and the commoditization of software“. Nearly ten years on, and it’s still a great read about the rapidly changing technology business.

As Murdock explains, the OS has been commoditized. Linux is everywhere. Now it’s in phones, home routers, search engines, rifles, refrigerators, etc… Both consumers and the field of computing as a whole have moved past caring about a single system running an OS as the computing atom. Now we talk about clusters, fabrics and clouds. We just want devices to work and services to be delivered. Judging by this year’s CES and SXSW Interactive, the IoT buzz is reaching fever pitch and will ultimately be assimilated into yet another demand on the network. For the remainder of this post we’ll consider where Linux is headed, who will build and operate the future network, and what that will look like.

Linus Torvalds recently said “people who start writing kernel code get hired really quickly.” The Linux Foundation report confirms this. Knowing Linux is great, but it’s no longer a sufficient skill for IT and engineering staff. Remember what I said about open source, operating systems, and networking? Developers with skills in one or more of them are already creating the next generation of network technologies, while their engineering counterparts are starting to deploy them. These people are also in the greatest demand and also have the most job security. Linux is the underpinning of two important Linux Foundation projects: OpenDaylight and OPNFV. These projects will be integral to the next generation of networks.

Turning to hardware, we see that the effects of commoditization on both hardware and OS means that Linux readily runs on many platforms, foremost among them are x86 and ARM. As x86 and ARM continue to mature, they can each handle more specific use cases previously handled by ASICs. Merchant silicon is also enabling many new options in networking as companies can focus on innovating in software instead of having to reinvent the wheel and assuming the cost of manufacturing their own switching ASICs. We will discuss this more in a future blog post. Cumulus Networks and Pluribus Networks are both offering interesting and open Linux platforms at the network level driven by these innovations. Whitebox switching and server switches are new technologies to manage costs and complexity as well as deliver new capabilities into the network.

It’s important to remember that everything we’ve talked about so far is still networking. There was just as much early FUD about whitebox switches taking over the network as there was about network engineers being a dying role. It’s simply not the case. Like any other area of technology, those who want to learn and grow will adapt. Here at Itential, we’ve met many network engineers who want to be part of the solution to building and operating future networks. Some of them are learning programming, but this is not a necessity. Others are learning more about servers and Linux, no doubt aided by Cisco’s growth in the server market. We’re hiring people who want to grow in these areas!

To conclude, the old way of networking with boxes and cables worked great for a long time. However, the demands being placed on the network now make business as usual a game of chicken between cost and complexity. If you keep going that route, it doesn’t end well. SDN and NFV are going to be the way forward to manage costs and complexity. They will be driven by applications running on systems (physical and virtual) powered by Linux. Those with skills in these areas will be in-demand, and companies that embrace these changes will be best positioned to focus on the unique challenges of their business rather than being mired with the cost and complexity of a network that is difficult to manage, doesn’t provide needed visibility, and cannot offer new services quickly.