SDN: Open, Fragmented Chaos

I wanted to follow up on a prior blog post after attending a recent SDN conference where I also moderated an investment panel.  In summary, I walked away from this conference and reading other recent SDN news thinking that 2013 will be a year of increased entropy for the SDN market.  VCs will continue to fund new start-ups, incumbent large IT companies will announce SDN plans/roadmaps, users will demand open standards to avoid vendor lock-in and the press release and marketing onslaught will be intense.  From an investment perspective, I still think it is too early to make any definitive conclusions given all this disorder and timing of SDN revenues being significant still being a couple of years away, but I believe that existing merchant silicon suppliers like Broadcom can only benefit from the deployment SDN with little threats from start-ups, while Layer 4-7 based appliance companies like F5 are most at risk given the ultimate SDN architecture and significant VC start-up funding in this area. 

Users Want Open Standards: Not surprising, both enterprise and service provider end users are weary of being locked-in to single vendor or vendor coalitions. After all, one of the main goals of SDN is to unlock monolithic data center hardware and appliances to allow for more innovative and faster feature development.  What I think will be challenging here is any standards efforts typically involves multiple constituents with different agendas.  This tends to slow down the standards process and results in compromises in the ultimate standards that limit functionality and flexibility.  A very relevant example in the recent past was the standardization process for IP Multi-media Subsystem (IMS) in the telecommunications industry.   When I asked a senior technical executive from the telecom industry at the SDN conference on whether there were any lessons learned or best practices from the IMS standardization process that could be applied to SDN, the answer was not encouraging. Specifically, the executive mentioned how the standards process around IMS was tedious, took longer than expected, and resulted in compromises that ultimately left the standard somewhat inflexible for some future unforeseen requirements (e.g. certain aspects of machine to machine communications over 3G/4G wireless networks).   Although not yet formally announced, the new open-source Daylight controller consortium from traditional networking and IT vendors Cisco, Citrix, HP, IBM and NEC will be interesting to watch.  Is this is a true open-source initiative, or a coalition effort from those that benefit from the current processes in data center design, implementation and hardware sales that just want to keep the status quo as we transition to SDN over the next few years.

Fragmentation and Chaos: To me, the SDN market right now is both fragmented in terms of company functionality and chaotic in terms of vendor positioning and marketing.  I say fragmented as most start-ups (and public companies for that matter) I see are offering one or a few pieces of the overall SDN solution, but not one is all that encompassing. That make sense given how broad SDN is both in terms of architecture and functionality.  It is likely we will see continued funding of start-ups to fill in functions within the SDN functional grid as well as acquisitions as existing public companies and mature SDN start-ups seek to fill out their SDN offerings.  The F5 acquisition of LineRate was a recent example of this as an existing appliance based Application Delivery Controller company F5, acquired an SDN start-up focused on Application Delivery Control.  Thus, the fragmentation of the SDN market is likely to remain and supportive of continued VC funding, which will continue to fuel consolidation as mature start-ups, traditional networking, hardware and software companies seek to fill in the gaps of their SDN solution.  I continue to believe such a cycle and the ultimate timing of significant SDN revenues being 3 years away will make it highly unlikely any true SDN start-up goes public in the next two years.

I also characterize the SDN market as chaotic right now given the marketing onslaught of large technology companies in 2013.  In the past several days alone, we have seen initial indications of the open-source Daylight controller (e.g. expected to be supported by Cisco, Citrix, HP, IBM and NEC), SDN announcements from traditional vendors Ericsson and Huawei and ONUG releasing its top five recommendations to enable Open Networking.   While 2012 was the year start-ups garnered virtually all the attention in the SDN market, 2013 seems to be the year that technology incumbents are scrambling for mind share through coalitions, product launches/roadmaps and acquisitions.  On top of these developments, venture capitalists on the panel I moderated at the SDN conference indicated they expect further investments in 2013 for new SDN start-ups.   Seems like the SDN crescendo will only intensify throughout the year.

Investment Thoughts:  My investment thesis around SDN continues to evolve as I continue to digest new information.  I provide some takeaways from the investment panel I moderated at the SDN conference below, which are of-course subject to change in the future as new information becomes available.   

  1. Little Competition for Merchant Silicon Companies: Everyone agrees that there will be strong demand for merchant silicon for new hardware platforms as the SDN market develops.  On the other hand, it appears VCs do not want to fund merchant silicon start-ups given the high R&D and other costs associated with semiconductor companies vs. software companies.  Thus, my conclusion is that Broadcom and maybe Marvel (if they can get some traction with their merchant silicon products) could be companies to benefit from the growth of SDN, although it will take a few years for SDN to truly drive merchant silicon sales. Intel would be another beneficiary, but merchant silicon would likely be too small of a business for such a large semiconductor company.
  2. Layer 4-7 Companies More At Risk Than Cisco: While all incumbent data center equipment suppliers are potentially at risk from the future of SDN, I think special purpose appliance based Layer 4-7 companies like F5 are more vulnerable than Cisco.   I say this because the ultimate SDN architecture will still require physical switching fabrics in the data center. Perhaps these fabrics will be merchant silicon based and Cisco will suffer share loss or margin pressure, but perhaps not.  On the other hand, the stand alone Layer 4-7 appliance is not present in the future SDN based data center, but rather replaced by a pure software solution in the application layer.  While its possible companies like F5 can pivot and transition their business models to be the suppliers of such software, the VC community seems intent on funding talented start-ups to attack this technology discontinuity while at the same time they are not funding merchant silicon companies at all and seem to be rarely funding data center fabric companies. 

Disclosure: I currently own shares of Cisco, HP, Marvel and Ericsson mentioned in this report. I may currently or in the future solicit any company mentioned in this report for consulting services for NT Advisors LLC.

SDN Permeates Ethernet Expo Conference in NYC

I attended the Lightreading Ethernet Expo Conference held in NY this week. While technical topics at this annual conference like 100G, Ethernet based wireless backhaul, small cells, Carrier Ethernet 2.0 etc.… were all prominent, I was fascinated to see how much Software Defined Networking (SDN) permeated the presentations and discussion as compared to the conference a year ago.  With the first Open Networking Summit taken place only about a year ago in October 2011 at Stanford University, it is spectacular to see in the course of only one year how much discussion in both the networking industry and investment community SDN now comprises vs. the actual level of SDN network deployment and revenues.

My current observations regarding the pace of the SDN deployment and the impact to the investment community are as follows:

–       SDN Is Unstoppable:  SDN is unstoppable and will be disruptive across Cloud, Carrier and Enterprise Networks, likely in that order.  There are too many smart people and disruptive companies within the technology industry working to make this happen.  Early high visibility adopters like Google and the recent significant increase in VC funding into the SDN area are also adding to this momentum.

–       The Pace of SDN Deployments Will Not Match the Current “Hype”: The pace of truly open SDN deployment via agreed upon OpenFlow standards will likely be slower relative to the current level of industry and Wall Street “hype”.  This is not, however, unusual for a new disruptive technology like SDN.  A simple example is key features that are required in the Carrier community like MPLS, IPv6 and VLAN Stacking still need to be broadly developed and tested across the vendor community.  In addition, open standards around the Northbound Interface for APIs from the Controller Layer to the Orchestration and Feature Layer are not yet fully specified.

–       Risk To Proprietary Northbound Interface Implementations High: I think there is material risk to a truly open standards approach on the northbound controller interface.  I believe early adopters/innovators in the cloud operator and/or vendor community will drive pre-standard implementations of the northbound interface to facilitate a competitive and time to market advantage over the competition.  The standards work right now seems more focused on controller to data plan standards work (and rightfully so) given the magnitude of work that needs to be done in that area.  To me this raises the question, will SDN see a couple of leading operating systems around the control layer through early leadership in the northbound interface that will lead to only a couple of dominant winners like we see today in server virtualization (e.g. VMW) or Smartphones (Apple and Android operating systems that drive application development)?   When I posed this question to the carriers at the Ethernet Expo conference, they all were cognizant of the risk, but felt that the SDN northbound interface will not follow the server virtualization or smartphone model. The carriers point to the standards process in place to avoid this risk.  Even so, a cloud provider like Google or a leading SDN controller company’s desire to differentiate itself through a more rapid feature deployment via a time to market advantage of the northbound interface could be very tempting.

–       The Window For Early Exits for VCs Is Narrow:  There is a short window for early exits for VCs in their SDN portfolio companies. While there have some nice early exits in 2012 for SDN companies (i.e. Nicira), I think the window will be short for such exits as actual SDN revenues and deployments will be minimal in 2012 and 2013.   My advice to companies and VCs in the SDN market, is don’t be greedy if your exit strategy is to sell the company.  The iron will not always be hot as it is now and don’t let investment bankers fill your head about lofty valuations just because of the VMWare/Nicira deal.  Facebook announced the acquisition of Instagram for over $1B in April of this year.  Do you think Instagram would sell today for over $1B????

–       Valuation Impact to Publicly Traded Networking Companies Is Semi-Permanent:  Even though the migration to SDN will be gradual over the next 3 years, the negative impact to valuation metrics for publicly traded hardware based networking companies is likely to remain.   SDN will take time to penetrate carrier and Enterprise networks, but as I said earlier it is unstoppable.  Investors looking at companies like Cisco, Juniper, F5, Brocade etc.… will constantly have the overhang of what SDN will do to the business models of these companies over the next 3-5 years.  A case in point at the Ethernet Expo conference in NY this week was the presentation from AT&T.  AT&T clearly articulated that their vision of an SDN based network of the future will not only utilize standard, low cost and high volume Ethernet Switches, Storage and Servers, but also the ability to eliminate the deployment of application specific appliances (e.g. DPI, Firewall, Load Balancing, SBC, CDN etc.…).  AT&T envisions these specific network appliances being replaced by virtual appliances written by independent software vendors.  Thus, SDN not only creates an overhang on Switch/Router companies but also network specific appliance companies.   We are likely to see the successful networking companies transform themselves by quickly endorsing the SDN model and selling features as software/virtual appliances rather than hardware with their embedded features.   The challenge they will face is start-ups are being funded to break into the virtual appliance/software model given the technology discontinuity created by SDN.