Does SDN Mean IT Will Be Able To Get Rid of Network People?

Over the last several months, Greg and I have talked a lot about software defined networking. SDN is a new and interesting way to look at moving traffic around a network. In fairness, some would argue that SDN is not “new” as such, but I think almost everyone would agree that it’s interesting. Yes, even those of you who are sick of hearing about SDN. SDN seems to strike fear into the hearts of network engineers just getting started in their careers. “The world is changing,” they – and perhaps you – reason. “Is there any point in going after networking certifications? Or being in networking at all?”

By way of example, missives like the following grace the podcast inbox from time to time. (Identity of the sender kept anonymous at their request.)

I was listening to Show 112 and the other SDN shows, and you are making me worry.  I got my (standard) CCNA last year, and am working towards a few more certifications this year.  Now with all this talk about SDN, is IT going to need as many ‘network people’ in the future?

I mostly do desktop support right now (have been for about 8 years now), and hope to move into doing more network type stuff at the enterprise where I am employed.  What I wonder is maybe within 5 years where each site for my company has at least one network person, will one network person be able to support all of our sites?  Right now, the company has about 15, maybe 20 network people spread around the state.

Therefore, what I wonder, do you think there will be continued growth for CCNP, and CCIE type people over the long run?  Or might SDN technology more or less put these type of technology people out of work…  And that would leave CCNA type people to rack switches/routers, plug-in cables then call the one CCIE for the whole enterprise and say ‘ok, can you ping it’?

Should maybe I give up on networking infrastructure and look into moving to servers and virtualization (shutter the thought)?  I don’t want to get to be a CCNP in a few years, only to have that kind of work ‘replaced by a machine’…

My short answer to the concerns raised here is that networking folks haven’t got a thing to worry about. Sharp network folks at all levels of ability and experience will always be employable, even in the coming SDN netpocalypse. More than that, I don’t think SDN will enable IT organizations to shed network staff members. As my friend Tony Mattke is fond of saying, “Complexity goes against robustness.” And software defined networks will inevitably be complex. SDN does nothing to make networks simpler – not down underneath, not down where it counts. They are every bit as complex as legacy traditional network are today; I could argue that they are more so. With complexity comes things not working right. And when things don’t work right, it will take someone who knows where their towel is to make it better.

And that’s us. The packet militia. The hardcore, stone-cold, heroes of the keyboard armed with sniffers, cluebats, and whiteboard markers. Yeah. The network engineers.

That was all very rah-rah…I know. So now for the longer answer…

Question 1 – With all this talk about SDN, is IT going to need as many ‘network people’ in the future?

As I mentioned above, my opinion is yes. The only specific burden I see SDN alleviating is that of rudimentary, day-to-day provisioning. And even that’s a stretch. Why? Because we have the ability with SNMP, expect, NETCONF/YANG, and APIs to automate network provisioning in a way that integrates with server provisioning, and we have for years. And yet, on the whole, there’s still not much of it going on. Or if there is, the provisioning is a part of workflow process that still involves human beings to review and approve – humans that need to know what they are looking at.

Also understand that server and virtualization people are not network people. Yes, there are smart folks in that realm like Scott Lowe (you do read his excellent blog, don’t you?) who are wrapping their brains around new network technology, but networking is a complex, detail-oriented, technical discipline. Just like virtualization is. And just like storage is. Etc. It is becoming increasingly difficult for any of us to be very good at more than one discipline, because it’s challenging enough to keep up with our own. As the problems unique to each discipline are increasingly complex, so are the solutions.

In my experience, server and virtualization guys are often stumped almost immediately when their system can’t communicate, and usually the issue is no more complicated than a bad VLAN tag. Do we think a complex connectivity overlay generated by an SDN bathed in unicorn tears is going to suddenly imbue those guys with the knowledge requisite to successfully troubleshoot a problem? In the words of John Pinette, “I say nay nay.”

Question 2Will SDN enable one network person to be able to support far more sites than today?

Not very likely. To my mind, SDN’s big promise isn’t that of fewer staff people. It is (potentially) faster time to bring network services online and dynamic reaction to changing conditions. In other words, SDN will add new and rich functionality, as opposed to replacing people who do mundane networking tasks. Not to beat a dead unicorn, but the richer that functionality becomes, the more human firepower that will be required to make it work right.

Question 3Do you think there will be continued growth for CCNPs and CCIEs over the long run?

Yes…although I think the sorts of skills that are taught in the CCNP and various CCIE tracks are going to expand from what they are today. Note that I said “expand” and not “change”. IP networking is still IP networking. IPv4 will be around for a long time. IPv6 is growing rapidly. Ethernet is the default transport, and the roadmap for Ethernet is long indeed. Switching and routing is still switching and routing. Those networking fundamentals will serve you well for the long haul, no matter if vendors decide to replace OSPF and BGP with some fancy new protocol to make forwarding decisions.

Therefore, dig in. Get certified on Juniper, Cisco, whatever. But then keep learning. One of the themes that comes up in SDN discussions is that of network programmability…using APIs to program network devices instead of the CLI. If you know how to do scripting and/or honest-to-goodness programming, I believe that this skill will probably serve you well in the networking industry. We’ll see how this develops over the next several years, but if you don’t know how to read & write code, start learning.

Question 4 – Should maybe I give up on networking infrastructure and look into moving to servers and virtualization?

Now, that’s an interesting question. If you’ve read this far, you know what I think the outlook is for network engineers in light of SDN. So obviously, no, I wouldn’t give up on the networking side of things. However, if you want to augment what you do, virtualization is, in my opinion, the absolute next best tech to get a handle on. Why? Network infrastructure is being virtualized. Part of SDN’s promise is to abstract physical network devices and treat the network as a virtual entity. Plus, there’s an interesting array of virtualized network appliances out there. Understanding hypervisors and how packets flow around them is a critically useful skill to have as a network engineer (something I’m still working on myself), and there’s a lot to know. As you start digging into this topic, you’ll find that there’s no one way every vendor handles their virtual switches, security between guest OS’s, etc. It’s a huge topic, but one that will make you exponentially more valuable to IT organizations as you learn it.

In summary, I wouldn’t move over to virtualization full-time, but I would dig into the networking side of virtualized infrastructures. Maybe add to your CCNP or CCIE a VCP.

Stay The Course

Change is here in the networking industry. Change isn’t on its way real soon now. It’s here. Pay attention to what’s going on, and keep your skills up to date as best as you can. Understand the problems that new technology is trying to solve – understanding the problem will go a very long way to helping you understand the technology in play. Networking isn’t about provisioning VLANs, adding routes,or turning up interfaces. It’s about getting traffic from one place to another quickly, reliably, and securely. If the task involved to facilitate that happens to be provisioning a new VLAN, so be it. But if that task turns into writing a script that an automated provisioning task calls when a new VM gets spun up, be okay with that, too.

We don’t know where SDN is going to take us yet, and SDN has gotten far more media attention than market adoption. The best thing you can do is stay on top of what’s happening, and in the meantime, stay the course.


  1. riw777 says

    Another point to add on to yours –learn old technologies, because every new technology is really probably a rerun of something that’s happened before. Understanding the business cases, the proposed uses, and the final resting place of older technologies will give you a leg up on understanding new proposals.

    There is nothing truly new under the sun; knowing enough of the old almost always provides the key to the new in some way.

    Don’t despair of your career, stay the course.

    • Riz Ali says

      Thanks Riw777 well said i agree with you and its an appreciable encouragement what you gave in your comment .i respect your words.

  2. says

    Another point I would add – don’t worry about SDN making any current technologies obsolete. Nobody is going to stop using established technologies overnight. While many companies run on a 3-5 year cycle for replacing equipment, some run on 10 years, or longer. As a consultant I still see networks running CatOS, I still see PIX firewalls, or XL series switches. For a lot of companies, if it works and is still supported, they will carry on using it.

    On the other hand, don’t stay rooted in the past. Learn about current (i.e. shipping) datacentre technologies. Learn about SDN. Keep learning all the time and you won’t go far wrong.

  3. says

    Hey Ethan, another great post. From a high-level view it appears that controller based networks will require that its subject-switches be of similar hardware type and that the network topology is either fixed or heavily constrained. Otherwise the controllers code would become increasingly complex (and thus fragile) in order to handle the variability in the network.

    Most companies would love to have a homogenous network, but I don’t see too many out there. The complexity in our industry comes from evolving and integrating disparate and heterogenous networks. I don’t see that SDN is flexible enough (yet) to play in the space when configs aren’t standardized and EOL equipment is common.

    I think the greater impact to our industry is that SDN will be used as a tool to commoditise network hardware. But even then, having a solid foundation from certification and experience will help your concerned emailer to learn new vendor hardware and OS really quickly as the need arises.

  4. ktokash says

    Good post and comments. I agree, start down the path. Even if someone came out with a flawless implementation of SDN tomorrow it would take a year for half the tech companies out there to even hear about it, longer for non-tech companies, and a few years after that to start any migration.

    Speaking of migrations, if you firmly believe that SDN is a game changer you can find someone with a good implementation of it and learn it. If you know “classic” networking and the hot new replacement you’ll bank a lot of money as a consultant migrating networks, much the way people are doing with IPv6 right now (though v4 is a concurrent technology). If you don’t know the current paradigm you won’t have that opportunity.

  5. Jason says

    no, you will have to get familer with a IDE of somesort , and be familer with API of some sort. And you can chuck that SNMP OID cheat sheet in the bin as well.

  6. Ciscosurplus says

    So with all that what is the compelling reasons for enterprises to make the switch,

    Can’t shed staff
    Just as if not more complex
    Doesn’t really buy me anything in terms of revenue or cost reduction

    Errr let’s not bother then

    • returnofthemus says

      Just what I was thinking, though at the time can’t find any reference to it in any Cisco Press publication, so it’s probably just another one of those fancy Buzz Words anyway, LOL!

  7. CiscoVarGuy says

    are those suggestions just for network engineers ? I work as a consultant for a Cisco VAR in LA area, have CCNP+Voice,ASA skills. What would a nice natural career path for a consultant. Just recently I saw a CCIE Voice consultant blog that recommends move into Big Data,Hadoop,hive,pig and Scripting Application market.

  8. Alex_Aus says

    Thankyou very much, I am a CCNA from Australia soon to be working on my CCNP the concept of SDN hit me a bit when I first read about it and killed my confidence for the future of my career path. This article assuaged my fears in this regard. Thankyou

  9. thema27 says

    Someone once said that if you want to become the best in your field, find a master and camp at his doorsteps. I am vCamping on this website lol
    But on a serious note, this is good stuff. As a newbie, I am relieved to hear this and, most importantly, to be provided useful pointers…Thank you for the invaluable advice!

  10. maybeedave says

    Ethan, my apologies for being blunt but you are wrong. SDN purists do target CCIEs. One of the reasons why networks are so complex as each router is a brain or has mulitple brains (meaning routing protocols inside). When these brains don’t talk correctly you get routing loops and other anomalies that we engineers take great pains to avoid. By having one God Box (SDN controller) it can now see all of the topology and ensure there are no loops. In the pure SDN environment this means no more OSPF, EIGRP, RIP, but maybe you still have BGP for your Internet and Extranet. This is the model that BigSwitch is pitching with their “Topology Manager”. In fact they are partnering with “white box” venders who simply put an Intel Alta asic in the box and run an OpenFlow agent. The new model is a fraction of the cost with buying Cisco gear and the support contracts. The new God Box can eliminate the complexities of engineers filtering and traffic engineering. We move from a world with complex engineering to a world where we have a complect controller and a lesser paid Network Operator to input data into a GUI so the God Box can figure it out. This will clearly eliminate CCIE jobs! It also has the potential to be a Cisco killer as now all the routing protocols are not needed. You are right that it will take years though your comparison to IPv6 is all not a good one as there is no killer app for IPv6. The killer app for SDN is MONEY. SDN networks have the potential to be a fraction of the cost of todays networks and the cost savings they tout is Operational Expense which is our salaries!

      • oldcreek says

        This is uncalled for, calling others name does NOT make you look smarter, and your post is garbage, I bet you don’t know sh*t about networking.

      • Jean-Christophe Perrault says

        That really doesn’t help the discussion. I sadly see what maybeedave is saying and agree with it. While I doubt we will see all Network Engineering jobs disappear in the next 5 years, the reduction of salary will be a huge incentive for companies with large networks to move to SDN and replace dozens of Network Engineers with a SDN network and one SDN Engineer. We might end up seeing CCIEs fighing for the networking jobs left in average size companies and CCNPs moving into different branch of the IT field. It is not doomsdays by any mean, companies will make more money and less staff so in THEORY the salaries should rise but a lot of knowledge we have acquired in the last couple of years might be useless for many of us and we will have to keep on studying to get new jobs. Sadly, knowing how to configure trunk links will probably never serve me again in this situation.

  11. Matt says

    Questions like – should I learn networking or SDN stuff are for the absolute beginners, or so it seems to me.
    I´m working on my CCIE… failed the lab once, but not giving up. And I do think this is the right way to go, but this does NOT mean that we should ignore the SDN growth. On the contrary!!! I´ve been reading a lot about the SDN, Openflow, NVF, OpenStack, Virtualization… it´s still not ready, but it´s definitelly the future, so… dont be lazy, learn as much as you can. you wanna be a CCIE? So be it!!! You like the SDN? So learn it!!! No rush, one thing at a time, but dont be lazy…

  12. pat says

    I think it will help offset costs in a data center environment and the switching world but it’s not going to affect routing and ISPs as much. The backbone of the internet cannot fully utilize SDN. I’ve worked as an network engineer for an ISP and Network support for a large data center and I can only really see SDN benefiting one of them. WAN and layer 1 will be around a great many years I think.

  13. Thomas says

    I am a software engineer (who loves the idea of SDN) and I work for the largest ISP in the United States on the Network Engineering Team and I will confirm that SDN is an (extremely) small blip on our radar as an organization, yes we do talk about implementing SDN “sometime in the future” but as for tossing away our current investments with Cisco, Juniper and Arris its not happening anytime soon. I would advise for those of us still in the work force for the next 20-30 years by all means learn some SDN concepts so you can join the conversation and maybe learn to code a bit so you will not be totally lost. However I would say for all the Network Engineers DO keep learning and certifying in your chosen technologies as they will be a huge asset to you as more and more vendors roll out their own implementation of SDN.

Leave a Reply

Your email address will not be published. Required fields are marked *