Is SDN the innovation the industry never knew it needed?


iphoneDoes SDN bring with it the type of innovation the original iPhone brought? Innovation takes much more than just really great technology.  It’s having the right technology that disrupts the status quo and enables uses that couldn’t be predicted.  For example, the iPhone revolutionized the way we look at the smart phone.  Prior to the iPhone you couldn’t convince my wife that a smartphone could completely replace her MP3 player, HP IPAQ and Mobile Phone.  Apple wasn’t sure (or right) about how the application model would look on the iPhone but they knew the existing model didn’t work.

Prior to the current generation of smart phones, I didn’t even consider the smartphone experience to be substandard.  My Palm Treo was the height of the technology and looking back and comparing the technologies, it’s amazing to see how wrong the industry had it.

I’m asking the question if SDN is the enterprise equivalent to the original iPhone.  Is it solving problems in ways we never thought and enabling a slew of applications we never thought possible?  SDN proponents have called the traditional approach to networking “a big bag of protocols.”  Meaning we’ve always just looked at the problem from the challenges and limitations of our protocol based approach.

The industry has traditionally taken a hardware first approach to solving protocol related issues.  Looking back in history, as the BGP routing table for the Internet got bigger the solution was to add more memory to edge routers.  It was when it became too impractical to keep adding memory just for route tables that route summarization helped make sense of the true problem.  Just improving the specs on the hardware was building a better Palm Treo.  The industry doesn’t just need better network devices.  The industry needs a different way to look at networking.

Is SDN that new and better way of looking at networking? Is SDN the iPhone-like paradigm shift the industry needs to develop a whole new set of apps that have not been possible without a new model?

There’s a lot of excitement around SDN.  I can’t seem to write or talk about this topic enough.  My most popular TechTalk segments are on SDN.  At VMworld, the hottest topic was VMware’s SDN play, NSX.  The media, customers and bloggers are intensely monitoring the conversation on Cisco vs. VMware SDN strategies.  But, excitement alone doesn’t bring industry wide change.  Organizations have to see tangible benefits to technology in order to implement it.  Network departments have to see a clear value proposition for implementing disruptive technologies such as SDN.  Thinking about what a virtual network SME looks like alone gives me perspective in the difficulty of convincing an enterprise to undertake an SDN transformation.  How many networking SME’s with virtualization expertise are out on the market?

What is the value proposition?

Both VMware and Cisco, who have the mindshare on the SDN topic right now, are talking similar benefits.  According to both vendors, SDN is all about cloud, big data, operational flexibility and agility.  Looks like they are covering all of the hot buzz words.

The reality is that these applications and operational challenges are great examples of the promise of SDN even in the current state of the technology.  The less obvious but I believe more important value of SDN is in the applications we have yet to realize are possible.  Without the innovation of today’s smartphone, none of the new mobile economy would be possible.  This is the true promise of SDN.  Today’s network limits our vision in the same way that yesterday’s smart phone limited the vision of mobile application developers.  SDN is the potential starting point for applications that are just not possible with today’s network.  I’m going to be honest and say, I’m not sure what applications will be built or what the model for building the applications will look like in a mature model.  It’s a similar lesson from the iPhone; we don’t know what will get created until we deliver the capability into the hands of the developers.

Published by Keith Townsend

Now I'm @CTOAdvisor

9 thoughts on “Is SDN the innovation the industry never knew it needed?

  1. Innovation is a strange beast, but the beast is mostly fed by hardware and price.

    It’s the hardware that makes ‘innovation’ possible.

    But use of the hardware is what makes it sell, what is use ? Use is applying it to a certain task. Also called application. 🙂 So hardware without an application is useless. So applications are what makes it valuable to customers. It’s the applications that fuel trends and trends can also help drive hardware development.

    With hardware I don’t mean the iPhone, I mean the parts. The touchscreen, battery technology and networking. Always in combination with price (the price needs to be low enough).

    The first iPhone-like phone on the market was the LG Prada, not the iPhone. The software and marketing of the iPhone was better. Why ? Because Apple has been wanting to make such a device for over 20 years. But the hardware wasn’t available. So they made sure their software was ready when the hardware was available for an acceptable price.

    Why do we know they wanted to make such things ? Because of their neighbors that produced the first iPad-like device in 1994*. And Apple created the Newton, their first attempt at such a device completely failed. So they waited a long time before attempting it again.

    Back to SDN. So SDN is also only possible because of hardware.

    But there are at least a few pillars I can think of of the top of my head.

    General purpose processors gained the ability to do virtualization (something which used to be the domain of only big systems from for example IBM). Because of virtualization we got virtual switches. Why ? Because systems got large and cheap enough and to run operating systems.

    OpenFlow is possible because general purpose CPUs have become so powerful and network communication so fast that it’s possible to separate data plan from control plane.

    Overlay networking exists because people want to deploy SDN without buying new hardware and still have all the benefits. It also can reduce complexity.

    Network function virtualization is possible because general purpose CPUs and other chipsets have evolved (specifically Intel has worked on this).

    An other trend is clear, partly fueled by hardware advances and partly by the advantages of Open Source many embedded systems now run Linux. Some of these things you have at home or in your car. Pretty much all DSL-routers, Digital Televisions, WiFi access points run Linux for example, but also in the enterprice we see this, many server BMC (read: IPMI) modules and SANs run Linux.

    And many enterprise switch vendors have now chosen the Linux as the OS for running their firmware. Juniper pretty always had their FreeBSD-derived Unix OS on their hardware. So there is now a general purpose OS on the switch. Some switch vendors even give you direct access to that OS. I keep wondering if that trend will somehow influence how we’ll build our networks in the future.

    * http://www.youtube.com/watch?v=JBEtPQDQNcI
    http://articles.washingtonpost.com/2012-03-09/business/35449604_1_tablet-computer-ipad-apple-and-samsung
    http://blog.thephoenix.com/BLOGS/phlog/archive/2012/03/13/from-the-pre-boston-com-globe-the-man-who-invented-the-ipad-in-1994.aspx

  2. I see SDN as analogous to the iphone in many ways. I wasnt really around for the mainframe era … but today it is hard to even imagine what type of application ecosystem might exist for the network? I imagine this would be similar if we could take a time machine back to the mainframe days and tell them just what type of app ecosystem would emerge if we opened the systems up – I think at that time it would have been hard to imagine. Just as very few would have imagined a year into the iphone we would have 100,000 apps.

    I am scared for SDN though – what if in phones there had been an extremely dominant provider – we could look at blackberry but the extensive bleedover between consumer and business tech isnt happening with network devices. But imagine there was some uber-dominant player, and because they were so dominant nobody bothered coming out with an iphone but instead pushed and pushed and pushed until the dominant player made an iphone and then bought the iphone from the dominant player. Now in this imaginary scenario, the reason why the iphone had to be pushed onto the dominant incumbent is because their dominance resulted in an unhealthy market that stifled innovation. Now the player finally gets the iphone years after it could have been possible, and the cycle begins anew. The root cause of the problem never being addressed, the market continues with a lopsided, non-competitive space that will just create the same problems all over again, continuing to stifle innovation and the benefits that healthy levels of competition can bring to markets. That is my fear for SDN. But, I am optimistic.

    1. I wasn’t around when mainframes were introduced either, never worked with them. But they are still being sold today. They can be very efficient. To give you an idea a couple of years ago a hosting provider replaced many servers with one mainframe. Much more power efficient I would guess and they would only have done it if it made financial sense to them. And many are rented I believe. You can even pay IBM per CPU-in-use.

      Blade servers always remind me of mainframes. I’ve mentioned silicon photonics below, this could lead to more use of blade server like systems. As components can be more easily separated. (because you can run a system I/O over silicon photonics, probably pretty much everything except for RAMCPU traffic).

      I wouldn’t be scared of SDN, most of it currently seems to be centered around general computing and software. Software development and knowledge can be distributed very well I think.

      What might stifle innovation ?:
      – (software) patents
      – Maybe a possible Intel dominance and eventually complacency:

      Intel has the biggest fabs with the smallest ‘processes’ I believe, that is their advantage.

      They are almost the biggest player in CPUs (ARM is bigger I believe), especially on the server of course.

      They do storage (SSD) and graphics but not all that great (their latest SSD they have on the market isn’t completely their own I think).

      They do switches (ASICs) now.

      Intel will be unleashing silicon photonics on the market as well. Other companies will have to license that from them if they didn’t also invest in 10 years of research on that topic.

      What scares me about what could be bad for IT-technologies in general ?: NSA/broken encryption/no privacy, DRM/Secure Boot/UEFI.

      1. Great points again Lennie, I agree. I should have been more clear – I do absolutely think there is a place for mainframes and other vertically integrated systems. I think the problem was that – at a time that was all that was really available – or accessible is a better word. The market dynamics left no to very few decent options for those that would prefer a different approach. The dominance of the largest players made it possible for them to convince customers whose businesses were not a good fit for the mainframe model that they should be using mainframes. Like you mention with the LG prada, it is rare when a new tech hits the market that is actually a new tech. Often whoever becomes successful with a new tech is not the one who created it – there are many technologies that exist today that may be accessible in the sense that someone could potentially buy the technology, but if its not available via one of the mass-market channels so few adopt that the technology never fully matures are gains adoption. I think as the industry transitioned from mainframes towards x86 the challenge was probably more sociological than technical – masses of people had to become educated while overcoming the best arguments from their trusted incumbent providers and they had to become comfortable with new models. They had to become willing to make significant changes to their day to day lives and their careers. I think much is the same way today with the networking industry being in a very much mainframe like state. The industry is filled with people who clock in and clock out every day, comfortable in their jobs, confident in their knowledge of the aging technology and older approaches to forming standards and norms they either have to be willing or be forced into change.

        I should have also been more clear, I am not afraid of SDN, I am afraid for SDN. In my analogy I meant to imply that the big incumbent is analogous to Cisco. If someone had leveraged SDN to its potential, they already could have been attacking the market with fully functional systems the way that thin wireless companies did, changing the industry almost overnight. SDN no doubt has disruptive potential, but how much it will disrupts the status quo really depends on the market and how the players execute. Nobody has as of yet used SDN to successfully disrupt Cisco’s incumbency, and that is largely because of the market risk – other established companies have been very sheepish about investing because they were afraid, what if customers dont adopt. The past decade is littered with startups that offered perhaps superior technology to Cisco that all failed or got bought, the only companies that have survived have been the ones that essentially do things in very similar fashion to Cisco. Not because there are not perhaps superior alternatives, but becasue of the sociology of the market – you could even point to earlier technologies that were essentially the same as many things gaining traction in SDN today – that werent really considered viable in the past – not because of the tech, but because of the sociology of the market and general inability to get past the ossification of the market.

        SDN is a case study in how bad the market became under a monopolistic paradigm. We could inject current technology into the market, and it would not solve this problem if we do not attain greater competition and solve the sociological challenge of changing attitudes and buying behaviors.

      2. I see the OpenDayLight project as a very positive development. If they can produce something that even almost works, then these kind of developments can’t be stopped. Doesn’t matter what incumbents like Cisco do.

      3. I sure hope so Lennie, I am optimistic and tend to think you are right about that .. but I’m also not going to stop pushing as I think it still could go either way. The more diverse the market becomes I think everyone will benefit.

      4. Here is one company http://cumulusnetworks.com/ that is doing what I mentioned above:

        “Cumulus supports a broad range of ODM produced routing platforms built upon Broadcom networking ASICs. They provide everything it takes above the bare metal router to turn an ODM platform into a production quality router. Included is support for both layer 2 switching and layer 3 routing protocols including OSPF (v2 and V3) and BGP. Because the Cumulus system includes and is hosted on a Linux distribution (Debian), many of the standard tools, management, and monitoring systems just work. For example, they support Puppet, Chef, collectd, SNMP, Nagios, bash, python, perl, and ruby.

        Rather than implement a proprietary device with proprietary management as the big networking players typically do, or make it looks like a CISCO router as many of the smaller payers often do, Cumulus makes the switch look like a Linux server with high-performance routing optimizations. Essentially it’s just a routing optimized Linux server.”

        http://perspectives.mvdirona.com/CommentView,guid,872dd193-eb1e-4ba6-b65b-65609d8ef7a7.aspx

        I found out about them because of their presentation on LinuxCon/CloudOpen:

        Click to access ptm_0.pdf

        http://viodi.com/2013/06/25/a-new-open-data-center-network-disaggregating-the-network-operating-system-from-switchrouter-gear/

        This isn’t against SDN per se as that the article seems to imply, maybe an alternative to OpenFlow though.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: