Will Virtualization Kill Networking?

August 20, 2009 - 26 Comments

TRIVIA: The first video MTV ever played was The Buggles’ “Video Killed the Radio Star“, which was sadly all too prescient as evidenced by the subsequent popularity of Milli Vanilli. So, a good friend of mine recently pondered if virtualization would similarly marginalize networking, if networking is becoming an inhibitor to innovation–I am not sure who plays the role of Milli Vanilli in his analogy. :)This is kinda curious stance to take. From my perspective, our level of connectedness as a planet is only increasing and we are still staying true to “Metcalfe’s Law” (the value of a network is proportional to the square of the number of nodes). Far from the network being an inhibitor, I see the network continuing to be an enabler, whether it is social networking or cloud computing, the interesting things that are going on are very network centric–Twitter is much less compelling if you have to mail your tweets on postcards. Similarly, the iPhone is hit not because it is a great phone, it is a hit because it is an exceptionally well connected phone. Looking at spending priorities and questioning whether budget is better invested in systems that can directly benefit the business, its important to remember that systems only create sustained value for the business if employees and customers can reliably, securely, and predictably access them. OK, so networking is a non-negotiable part of any system, perhaps all we really need are dumb pipes–people won’t notice or care. Well, if you want to test the assertion that end users and consumers view their “pipes” as a commodity, check out the various spirited discussions on the web about the carrier choices for the iPhone to see how opinionated folks actually are about their transport. This is pretty much in line with what we found when introduced our Cisco MDS Fibre Channel switches–once customers were educated, they started to have an opinion. Interestingly enough, there are companies out there that advocate the “dumb pipe” pipe perspective that networking is over-rated are demonstrating it by….investing to enhance their own networking offerings…?!So, the reality is that the data center continues to be three-legged race of technologies–sometimes one leg (technology) is leading the way,sometimes that same leg is being dragged along, but in the end, they all need to stay closely coupled to make any progress. The Cisco Nexus 1000V as a very practical example of this. The rapid adoption of VMware ESX forced us to reconsider how we deliver networking services and led to, among other things, the development of the Nexus 1000V. And now, the introduction of the Cisco Nexus 1000V has paved the way for much broader implementation of virtualization in the data center. And, so the race continues…For more thoughts on why Data Center infrastructure is sexy again, check out Data Center Infrastructure Gets Hip

In an effort to keep conversations fresh, Cisco Blogs closes comments after 60 days. Please visit the Cisco Blogs hub page for the latest content.


  1. The hypervisor is creating unprecedented flexibility, yet it is confined due to network limitations IMHOwireless reviewsThx”

  2. Virtualization rules. I like the energy efficiency about this topic and ease of creating virtual servers with no additional cost. Virtualiation is future.

  3. Jim,Great post. I couldn’t agree more with what you’re saying here. Not only is the network essential, but from what I can tell it’s becoming more and more complex. In the old days we worried a lot about just keeping our circuits up. I can’t count the number of days I spent on the phone arguing about whether or not I was seeing LMI and if their DLCI mappings were correct.With today’s technology network reliability has gone up significantly which has allowed us to become more and more reliant on the network and to spend more of our time on advanced services like QoS, NetFlow, and IP SLA.Everything I see in the community shows me a continued expansion of the role of networks and of network engineers. We have folks out there pushing the limits of 10Gb already and it’s not going to stop there.We’ve already added support for the Nexus 1000v in Orion and we’re seeing that customers love the enhanced visibility into their virtual infrastructures. Not sure where this technology will take us next, but I’m excited…Flame on…Josh

  4. Jim:We are working on improving energy efficiency on a number of fronts (internally, our products, and enabling customers). A couple of resources to check out on that front:Environmental Sustainability: http://www.cisco.com/web/about/citizenship/environment/index.htmlEnergy Efficient Data Centers:http://www.cisco.com/web/about/citizenship/environment/data_centers.htmlEcolibrium Blog:http://blogs.cisco.com/greenHope this is helpful,Omar

  5. Omar, can you tell us what is Cisco doing, generally, to go green, reduce energy costs etc. ?

  6. The hypervisor is creating unprecedented flexibility, yet it is confined due to network limitations IMHOwireless reviews

  7. I’m not nostlagic, even if the name and the access method didn’t change when i was installing 15 years ago Kalpana or Grand Jonction switches. They were too fast for some servers like VAX 3×00, APOLLO or Sun Workstations and i used to put 10Bt hub between to generate collisions … Future name would looking like … Ethernet Lossless 🙂

  8. Olivier:It is pretty funny–I think in the networking world we get some warped sensibilities. I remember my first job where I had a two week running argument with my boss to justify a T1 (he thought we could never fill the pipe), and yet today I have ~20Mb of bandwidth piped into my house.In a way, we are spoiled–we are used to growing by orders of magnitude. I think folks in the data center, with maybe the exception of the compute folks, are accustomed to more modest increases between generations of their products…?Omar

  9. DS45: You have to be careful when being nostalgic about ethernet. The ethernet we have now isn’t really related other than be very broad principles in design. For example, Manchester encoding and materials, etc. It’s mostly a marketing thing now. I guess fast”” by any other name is still fast. :)”

  10. In the way Omar takes …I have read carefully all comments and I found funny that XNS …. sorry, Ethernet reach soon 100 Gb /s and that we find it natural. Who would have thought at Xerox PARC in 1971? Like Bruce Springsteen, who has adapted to his time and musical tastes of its fans, Ethernet and Cisco have responded and anticipate our needs. Do not forget what is expected of us is that IT is most effective at the best price for the Company to be as competitive as possible. Computers are everywhere and we can’t actually know everythings. It’s time to automate our business and all the tools to help are welcome (even if they are funny to use).

  11. Omar- agree. More generalized skill sets and tools that provide for broader responsibilities.G

  12. There is a funny but relevant saying that sums up the future of networking..The Internet is a telephone system that’s gotten uppity””(Clifford Stoll). It used to be that network technology was way ahead of the applications but that no longer holds true. Just like the iPhone, it’s the applications that are and will define the architecture networks of tomorrow. I am seeing a shift from networking being a business tool to business as entertainment and that type of content requires a new type of thinking on the part of manufacturers. This is why I see Cisco having the vision to lead the way. Hey, Cisco bought Scientific Atlanta and recently Flip… It is obvious that the end devices are driving the shape of the network core. Who da thunk it that Cellular devices would be the number one devices accessing the internet and now new TVs will have Ethernet ports… I am waiting for Cisco to come up with a Wireless version of a XGig core switch and a switch that can access resources (content & endpoints) anywhere without fiber. Time for Cisco to but a wireless provider and turn it into a global switch…that will be the shape of things to come!”

  13. Greg:I am not so sure about the concept of the generalist, at least as a replacement for individual specialists–there is just too much to know. We see this now with server admins that have picked up network operations duties as they endeavor to manage their VM environment–neither the server nor the network team find that a palatable trend.I think leading organizations will be drive increased fluency amongst their silos and will benefit from making the effort. I think the other thing we may see is actually an additional generalist”” layer. We sometimes see this now, where some companies pull virtualization into its own operational “”silo”” separate from the existing teams.Regards,Omar”

  14. At this poing there are likely a range of possible outcomes, driven by innovation, vendor strengths/weaknesses, the vision of the IT organization (virtualization-lite versus full vmotion, etc) and the willingness of the network teams to embrace automation. IMHO it is likely that the rich environment will produce a wide range of outcomes for vendors, enterprises, service providers and IT careers.If I was to speculate the biggest impact, it would be the rise of hte IT generalist who understands the implications of virtualization and networking (delivery, security, etc) and policy (vs manual labor).ThxG

  15. Custom hardware of any kind will have to add new value that aligns with the emerging capacity, flexibility, automation and addressing demands of the dynamic mesh with VMotion. This may be a significant opportunities for new technologies to disrupt large, mature markets that have been all too dormant. Growth, margins return driven by automation and the second wave of virtualization (beyond the VLAN).Some network industry legends are starting to address some of these infrastructure 2.0 issues already.How about The Beatles Helter Skelter?

  16. Doug:Actually, I think if I were to extend the analogy, if would have to be someone like Springsteen or the like who has managed to adapt and stay current.So I certainly agree with both you and Greg–any laurel-sitting would be premature. I think the saving grace for networking is the speed at which it can change and adapt–I think Mr Linkin made this point once. Regards,Omar

  17. nicely written! But did you have to bring Milli Vanilli into it? At least use The Eagles comeback tour or something… I concur a bit with Greg’s assertion that while the underlying infrastructure of the network is continuing to evolve virtualization, for lack of a better word, transcended IT. It solved such a critical business problem that it got noticed. I think there is an opportunity for networking to do the same.

  18. Bon jour Olivier!Nice to hear from you again. Thanks for the education on Trevor Horn–I did not realize he was so accomplished.So, I agree with you in terms of many of these technologies being disruptive but not destructive. The thing that makes it so interesting is that these technologies all have unintended consequences, which keeps it entertaining for all of us.Omar

  19. Has vmware killed the servers ?Trevor Horn, the Buggles singer is probably the best music producer and didn’t killed radio stars …. He is behind FGTH, Propaganda, Art of noise, simple minds, Pet shop boys … Nexus 1000v is a new ‘out of box’ service which permit us be more imaginative. We now only have to enter in cloud networking.FYI : i made an howto around the Nexus 5K for the vmug.fr : http://ds45.blogspot.com/2009/08/howto-quest-ce-que-le-nexus-5000-en.htmlHave a nice day !

  20. MJK:I agree–as bandwidth increases, location largely becomes irrelevant. As much as we are consolidating now, it is inevitable that we will cross some threshold and the trend will reverse itself. As a example, I am already starting to see to opinion pieces that large scale data center consolidation is not tenable from a power and cooling perspective and perhaps we should be looking at a more distributed approach.Omar

  21. Greg:I will grant you that–there is certainly opportunity to either eliminate, or at least mask, some of the current complexity on all fronts. Maybe its because I am a network geek, but I have faith in the ability of the network to keep pace.Regards,OmarPS For those of you who do not know Greg, you should check him out at http://gregness.wordpress.com/

  22. I think the situation will change even more in favor of networking when the typical wireless bandwidth is equivalent to the bandwidth inside a typical computer system bus.Also, I thought Dire Straits, Money For Nothing”” was the first MTV video. Anyways, oldie but goodie.http://www.youtube.com/watch?v=VOD805iAqjY

  23. Omar:I think you’re right that there have been recent significant gains. Yet there are also layers of legacy complexity and manual labor in the network… as systems are increasingly automated. The hypervisor is creating unprecedented flexibility, yet it is confined due to network limitations IMHOThxGreg

  24. Greg:I don’t disagree with you, but I guess I don’t see networking as stagnant. In the last 18-24 months, we have seen the ramp up for 10GbE, 8Gb FC, and FCoE, with DCB standards and 40GbE and 100GbE on deck. We have seen port profiles on the N1KV, service profiles on the Cisco UCS and FEX move the ball forward in terms of automation. Finally we are seeing
    etworking”” expanding beyond the traditional switch and moving into the hypervisor (N1KV now, MR-IOV and the like down the road) and into the UCS without losing features or functionality, as in the past.I know we have a number of customers who wish we would stand still for a bit so they can catch up. :)That being said, I’m also the first one to say I might be a bit myopic :), so I would love to hear where you think networking is missing the boat–I appreciate your perspective on these things.Actually, I’ll open that up to all our readers…would love to hear hear what folks think.Regards,Omar”

  25. Omar:If networks don’t evolve/automate they will follow the factories overseas. Virtualization is just the first potential to shake up the empire of manual tasks.www.infra20.comGreg