Organizations need to learn to think about networks as holistic entities. Networks are more than core routers or top-of-rack (ToR) switches. They’re composed of numerous connectivity options, all of which must play nice with one another. What role does automation play in making network heterogeneity viable? And does getting all the pieces from a single vendor really make management easier if that vendor has 15 different operating systems spread across their lineup of network devices?
Most network administrators are used to thinking about their networks in terms of tiers. Access is different from branch, which is different from campus, and so forth. Datacenter is something different again, and then there’s virtual networking complicating everything.
With networks being so big and sprawling that they frequently occupy multiple teams, it’s easy to focus on only one area at a time. Looking at the network holistically—both as it exists, and as it’s likely to evolve—is a much more complicated process, and increasingly important.
Networks grow, evolve and change. Some of this is organic; growth of the organization necessitates the acquisition of new equipment. Other times growth is more unmanaged; something that’s especially common with mergers and acquisitions (M&As).
Regardless of reason, change in the network introduces the possibility for a heterogeneous management environment. The greater the number of management planes, the more difficult network-wide orchestration is—this has real-world implications for IT teams beyond networking.
The tipping point when automation becomes a necessity
Automation is an absolute necessity to do IT operations at scale. There comes a point in any organization’s growth where adding more personnel to the IT department simply doesn’t enable additional growth. Past that point, there’s no path forward except automation.
None of this is controversial. Every organization with more than a few hundred employees has faced this truth, and many are struggling with the result. There’s so much automation in use that there’s now a requirement to automate the automation. In other words: orchestration.
But how viable is orchestration in a heterogeneous environment? Each vendor has their own management stack. APIs are different. Systems and devices may or may not have support in third-party configuration management tools such as Puppet, Chef, Saltstack or Ansible. The necessity of orchestration coupled with the difficulties posed by heterogeneous environments has traditionally been the loudest argument made for absolute loyalty to a single vendor, especially in the networking world.
That rationale, however popular, falls apart upon closer examination. There are few vendors who can offer every piece of the networking puzzle, and they all suffer from an explosion of management options, multiple operating systems (each with different features and functionality) and all the very same problems that come with heterogenous networks in the first place.
In addition, those unexpected events—such as M&As—do occasionally occur. They’ll bring their own diversity to the table, as will the expansion of cloud computing, edge computing, mobile and all the other places around the world that your organization’s bits must flow.
An organization’s survival is inextricable from the health of its IT, and that IT doesn’t work unless the network does. How will all that IT infrastructure be automated, orchestrated and managed? Today? Tomorrow? Ten years from now? The foundations laid today will be built upon year after year.
If there’s little or no real-world operational efficiency to be gained through futile attempts at networking monoculture, doesn’t pursuing a strategy building an infrastructure designed from the ground up to support heterogeneity make the most sense? Vendors should be chosen based on their commitments to open networks, open protocols and open standards. Not for any ideological reasons, but for pragmatic ones—networks won’t get less complex with time, and 10 years from now you’ll be stitching together something new on top of whatever automation and orchestration it is that you choose to build today.
The holistic solution
Cumulus Linux allows you to manage the switch like a server with network automation. With Linux-based switches, NetDevOps is easier than ever before, as IT teams can leverage the automation tools with which they’re already familiar.
Cumulus Linux is built for agility, and offers a standard base from which to begin building the adaptive, open network of the future—one designed for heterogeneity, founded on a consistent Linux operating model, and enabling scale through automation that’s easier than has ever been possible before.