Has SDN improved or just Kardashianized your network?
Software-defined networking (SDN) has taken center stage in the networkingindustry for the better part of a decade now. In the spirit of theelection season, its a great time to ask the question: Is your networkbetter off now than it was before?
The answer should be a resounding yes. For network engineers with aspecific problem to be solved, there have never been as many tools in thetoolbox to solve those problems. Indeed, with all the openAPIs/automation frameworks and languages available to program them, themix of custom/merchant silicon and x86 solutions available today and with the exponential reduction in cost and power consumption of networkequipment on a per-bit basis, we may be in the Golden Age of Networking.
But for many networks, a stark reality has emerged around SDN. Debates onthe merits of esoteric networking technologies have been overtaken by anunending stream of fantastic pronouncements of panacean wonder. With hypetrumping substance, networking technologies have fallen irrationally intoand out of fashion, with some becoming famous for being famous. ThisKardashianization of networking has left many companies crippled with fearof missing out and paralyzed with confusion about what to do next.
The following are some of the pitfalls that have befallen carriers andenterprises that have failed to master SDN and have instead become masteredby SDN.
Chasing the press release
It is important to remember that the reality underlying most pressreleases falls somewhere in a truthiness continuum between padded resumeand Soviet economic pronouncement. Trying to reverse-engineer acompetitors technical solution, or even divine its true intentions, from apress release is fraught with peril. Also, remember your mothers adviceabout not jumping off bridges just because everyone else might be doing it.
A variation of this pitfall is the But <Insert Cool_Web_Services_Company>is using SDN refrain that usually follows any critical questioning of SDN.Indeed, it is true that some Cool_Web_Services_Companies use SDN, but itis important to understand where they are using SDN and why.
When understanding the spirit of SDN, beyond merely its literal definition, the potential of this technology becomes limitless.
Cool_Web_Services_Companies usually have multiple networks, some of whichuse SDN, while others/most resemble traditional (Pre-SDN) carriernetworks. Fortunately, the engineers at Cool_Web_Services_Companies dotend to write long, detailed articles on the technical solutions they useand it is wise to read the whole article to understand the motivations andapplicability of the particular SDN technologies they choose. What workswithin a data center might not be very useful in a network that connectsdata centers together, or one that peers with the rest of the internet.
Need to do SDN
Perhaps the most frustrating pitfall engineers encounter is when they aretold by their management they must do SDN. The motivation to put pointson the board to demonstrate compliance with an executive SDN edict doesnot tend to lead to positive outcomes.
Where SDN has been a boon is when engineers have started with thequestion: What is the problem to be solved? This may seem obvious andself-evident, but too often SDN strategies involve top-down selection of afashionable solution first and then rigidly applying it to every problem(or worse, to non-problems).
Rigidly defining what is and is not SDN
The strictest definition of SDN, as described by the Open NetworkingFoundation (ONF), is the physical separation of the network control planefrom the forwarding plane, and where a control plane controls severaldevices. In terms of intuitiveness, this definition unfortunately betraysthe practical goals and motivations of SDN to reduce costs and accelerateinnovation and agility in delivering network services. Companies cansometimes get caught up in the semantic debates of what does and doesntcount as SDN and miss the point of this powerful movement.
A better understanding of the spirit and power of SDN could becharacterized simply as the marriage of software and networking to solveproblems. This opens the doors to a full panoply of options: networkoverlays, automation frameworks that drive down operational costs and riskfrom manual processes, open interfaces that reduce the friction of mixingand matching equipment in various roles to drive down costs, etc.
Oneunderrated and underutilized SDN application in particular is the use ofprogrammable APIs into the networking devices. These democratizing APIsenable permission-less innovation from the crowd to rapidly deliver newfeatures and functions months/years faster than are typically developed bygear-makers. When understanding the spirit of SDN, beyond merely itsliteral definition, the potential of this technology becomes limitless.
Not searching for the Ricky Martin in the Menudo of SDN
To those of a certain age, the mention of Menudo typically evokes vaguerecollections of a Latin boy band from the early 1980s that launched thecareer of Ricky Martin. What most dont know is that the band included more than three dozen members, spanned three decades and spawned severalspin-offs. When it comes to boy bands, certain patterns tend to emerge.
While a boy bands rise in popularity can be meteoric, its demise is almostalways inevitable, usually cruel and often amusing. But on rareoccasions, a star emerges Ricky Martin, Justin Timberlake, Bobby Brown,Michael Jackson leaving in his wake a jumble of mostly forgottenbandmates.
SDN has caused great disruption and noise, but has it actually made the network better?
Which brings us to SDN, an approach that most believe began in the late 2000s. In truth, the underpinnings of many SDN technologies were introduced in some form or another years/decades earlier. The large castof SDN technologies includes many that are destined for future Where arethey now? musings. But we can expect some stars to have an enduringimpact in the industry. Network overlay technologies (for example, VxLAN)appear to be one example.
The lesson here is to search for the winners, not just assume everytechnology with the SDN moniker is equal there can be quite a differencebetween Michael and Tito. Furthermore, even finding the star can becounterproductive if used in the wrong application. Ricky Martin in frontof a pop audience will have the crowd livin la vida loca; onstage infront of a heavy metal audience and la vida will not be so loca.
The greatest asset, worst liability and most overlooked component of anynetwork is the human element that must design, deploy and operate it.Too often, the raw technical facets of a solution are solely considered tothe exclusion of how humans will use and apply the technology. When thehype bubble inflates, critical thinking can grind to a halt, and the results can be calamitous. Opportunity costs can skyrocket as scarceresources pour into dubious misapplications of trendy technologiesdominating the headlines, crowding out less fashionable, but far morerelevant and productive solutions.
SDN has caused great disruption and noise, but has it actually made thenetwork better? The answer to this question is ultimately a reflection ofthe motivations and goals of those who have sought to deploy SDNsolutions.
SDN is an umbrella of technologies, some powerful anddisruptive, some of dubious value. Careers have been made, as well asbroken, by SDN. For those network service providers that have started outwith a problem to solve, objectively looked at the bevy of options provided within the umbrella and applied the right tool for the job, SDNhas been a godsend. For those that have been caught up in the hype, startedwith a solution and worked backwards in search of a problem, that have soughtto use SDN without fully grasping why, keeping up with SDN has been as productive a use of time as keeping up with the Kardashians.
The opinions expressed in this work are solely those of theauthor and do not reflect the positions of any of the organizations withwhich he is affiliated.