8 Comments

Summary:

As startups sign up more customers, legacy network vendors are talking about starting groups to form their own standards for software-defined networking. But there are already standards, and it could be too late.

dell_logo

Dell is joining a growing list of vendors that want to develop their own standard for software-defined networking (SDN), and perhaps dilute the influence of customer-led standards in the data center.

Last week, the nonprofit Object Management Group (OMG) said Dell intends to create an OMG committee on SDN. The Dell news follows rumored moves by Cisco, Citrix, HP and other vendors to sponsor or contribute to a new consortium code-named Project Daylight. Reports of the project surfaced last month, and others have followed.

Vendors’ efforts to set standards boil down to a matter of securing as big a place in the data center as possible. In recent years, webscale companies such as Amazon, Facebook and Google have effectively prompted the efforts by moving away from legacy IT vendors and toward custom gear to better fit their needs at the their huge scale.

In 2011, Facebook announced the Open Compute Project, an effort for customers such as Facebook to name their own needs and wants for servers and other data center components. Dell has joined the Open Compute Project. But now not one server going in to Facebook’s newest data center, in the Swedish city of Luleå, comes from a traditional server maker like Dell.

Open Compute isn’t the only new customer-led standards organization. Facebook and several service providers, including Verizon and Comcast, created the Open Networking Foundation in 2011 to build a standard around the OpenFlow networking protocol. Dell’s SDN committee looks like an attempt to ensure a place higher up the stack.

But it could be too late for network-appliance makers to get out in front of enterprises with standards of their own, as SDN startups capture more and more customers and their products become more easily adaptable.

“The networking industry needs clearly defined leadership in the SDN technology space, and Dell is taking an important step to coalesce a standard under OMG through an open, international transparent standards process,” OMG said in the announcement that OMG it released Wednesday. That sort of language probably grates on the ears of Embrane, Pertino and other SDN startups, as they have already made inroads and could have trouble gaining a foothold in a standard with Dell and other hardware-focused players, let alone taking the lead.

As companies throw around new definitions for SDN, it becomes harder to understand what it is, and what it’s not, just like what has happened with the term cloud computing and the variations on it. SDN already has OpenFlow, thanks to the Open Networking Foundation. Do vendors really need to try to challenge those standards? Shareholders might want to see that sort of maneuvering, although it could be too little, too late.

  1. Cisco owns like 80% of the network market share. HP owns around 10% which is larger than the next five combined.

    Basically if both Cisco and HP get behind a standard then that will be the standard. From what I hear both companies are behind “Project Daylight”. So this is a done deal in my mind.

    Share
  2. Here is a deeper view of the SDN startup landscape, learned from interactions with networking & IT customers and key figures within the networking landscape.

    SDN startup ecosystem reminds of the networking bubble of the early 2000s, when there were dozens of networking and optical infrastructure companies, with no real business prospects in sight, except for tons of cash-burning trials and minor wins. Couple of early startups got sold to fuel big hype, and ones that came that came later had to sell their IP or fold.

    SDN Startups can be grouped into the following categories;

    (1) Virtual Network Infrastructure, L2/L3 Overlay Networking, Service Chaining Framework; This startup segment seems to have been mostly played out already, with Nicira getting acquired, followed by Contrail [at 1/10th the price tag of Nicira]. SDN startups in this category [like Big Switch, Midokura, Plumgrid] are most vulnerable to commoditization now and here is why;

    # All major networking vendors now either have SDN technology in the market or in works to offer in this category. It includes Cisco, VMWare , IBM, Microsoft, HP, Dell, Oracle, Juniper, ALU.

    # Enterprise market is jointly dominated by Cisco, VMWare, IBM, Microsoft, HP etc. Major enterprise customers who use VMWare as the Orchestrator have easier options to stick with VMWare, Cisco or IBM Overlay Networks, vSwitches and Service Chaining framework. For these SDN startup to win significant and sustainable overlay networking business against this incumbent ecosystem is just simply implausible.

    # Big public cloud providers [like AWS, Azure, HP] have their own networking stack. Smaller public cloud providers would look for free and open source combinations like OpenStack and Daylight.

    # Cisco and IBM will open source the Daylight SDN framework for overlay networking and network service chaining. It will likely work with OpenStack also. In essence, Daylight can open source the software that startups in this category are trying to sell for money!

    Just the way OpenStack open source was destructive to proprietary software companies [like Nimbula, Eucalyptus], Daylight open source can be destructive to SDN startups in this category.

    (2) L4-7 Applications;

    This appears to be the new hotbed of recent investor activity. Couple of companies have been funded by major investors in this segment. However, incumbents like F5, Citrix, Riverbed, Palo Alto, Silverpeak, Trend Micro have all come up with virtual machine versions of their offerings. Objectively, the performance of these virtual machines is pretty damn good for today’s use and new startups would need to do something very different and compelling to build a case against them.

    One startup in this category [Embrane] has released products last year but still has to see any significant business success against incumbent vendors. As couple of heavily funded new startups in this segment are still under stealth mode, lets give them benefit of doubt and wait till they make their offerings public. Will comment more on it when they choose to come out.

    The two other heavily funded startups in stealth mode in this category have teams coming out of Cisco and Juniper, respectively.

    (3) Fabric / Top of the rack switches;

    Openflow seems to be losing its hype so I won’t add that in this discussion.

    There are couple of startups in this category [Cummulus, Pica8], who aim to commoditize top-of-the-rack by providing complete stack of software bundled with a white-box top-of-the-rack switches, to compete against big guys. These companies seem to be selling and has momentum behind them. Their business prospects appear good.

    Then there is Pluribus, which is trying to host networking services within top of the rack switch itself. It makes TOR expensive. Many believe that such an architecture can create bottlenecks. May be there are some niche enterprise applications where they can sell. For public clouds, its unlikely to see traction.

    (4) Misc SDN:
    As SDN has become a catch all term, rest of the companies can be put in this catch all bucket.

    Pertino is offering a simplified VPN aka overlay network (the new cliche). This concept has been there since early last decade and there isn’t anything new. Check out LogMeIn Hamachi, which offered this functionality since early 2000’s, and you will understand the legacy nature of their solution. Hamachi offers much richer functionality than Pertino today! In addition to Hamachi, there are many other competitors in the market.

    Its not for love of large incumbent networking companies, but we can see the similarities that today’s SDN startup ecosystem has with later stage networking bubble of the early 2000’s.

    Share
    1. @JohnShaw

      Not sure where you got your Pertino G2, but it’s way off the mark. However, it’s not entirely your fault. Given the functionality of our current limited availability Network-as-a-Service offering, I understand your Hamachi comparison. But looks are deceiving.

      Under the hood of Pertino’s Cloud Network Engine lies a robust SDN architecture by just about every industry pundit’s measure; including Guru Parulkar, chair of the ONS, Executive Director of Stanford’s Open Network Research Center, and an early investor/advisor at Nicira. Elements of our cloud-based SDN platform include:

      • Full network virtualization via a WAN overlay architecture complete with a non-sequential, customer-unique IPv6 address space;
      • Segregated, redundant, and stateless control plane;
      • Cross-cloud, geographically distributed L3 switched data plane with a real-time control channel for forwarding primitives and feedback;
      • An advance orchestration layer that has more in common with Google and Netflix than VMware and Cisco;
      • Use of “merchant VMs” across multiple cloud providers;
      • An extensive northbound API and service enablement architecture for cloud-based NFV services.

      We have close to 1,000 customers to date who have used our limited release Cloud Network Engine to build secure cloud networks in 3 clicks that connect their remote sites, users and resources using just an Internet connection. Creating a secure global network in minutes with ZERO planning, ZERO hardware (beyond an Internet connection), ZERO configuration, and ZERO user impact seems pretty radical to me having built a good chunk of AT&T’s internal IT network back in the day. Most significantly, Pertino cloud networks are self-provisioned and self-optimizing. We do not need to pre-populate POPs with racks of equipment all over the world in advance of customer acquisition. As a customer is setting-up their Pertino account, our Cloud Network Engine platform will automatically turn-up a geographically aligned data plane access point if one is not already available. And, if the platform detects diminished performance, it can automatically spin up additional capacity or even move affected customers to other cloud data centers that are yielding better performance.

      We’ve just scratched the surface of what our platform will enable us to deliver to the market. Bringing enterprise-grade WANs to the masses would not be physically or economically feasible without the breakthroughs embodied in SDN.

      Isabella Guis offers a high-level summary here… http://www.sdncentral.com/sdn-blog/sdn-start-ups-you-will-hear-about-in-2013-part-3/2013/01/

      Share
  3. +1 to the comment above @JohnShaw – insightful and hypeless unlike every vendor claiming SDN halo.

    Share
  4. +1 @JohnShaw, great comment, thx

    Share
  5. +1 indeed @JohnShaw. V insightful and articulate

    Share
  6. Brent Salisbury Tuesday, March 19, 2013

    Certainly worth while. I would expect quite a few vendors to jump to OMG. Not too mention a chosen NOS should play a factor. Most vendors have been developing with Floodlight, based on Beacon. Daylight is undoubtedly not using Floodlight since that appears to be an attempt to push Big Switch out. Ironically Big Switch caught $8mil late round B funding from Intel a week later. Either way, both will emerge with Java controllers.

    The important factor imo is we have thrown out two years of network OS wars and realized out the gate the OS is a commodity. The value is in the application, not the controller. Southbound APIs still have another year or so of haggling while vertical integrators still try to differentiate based upon who has the better proprietary forwarding protocol mousetrap, but the winners will go after differentiation and embrace disruptive horizontalization.

    Big revelations and positive signs from those with nothing to lose, that they may actually be ready to take a little risk. Early entrants are going to start winning sales based on no product support. Those who wait on fortune50 to demand it, will lose seed equipment.

    Good read, keep up the coverage gigaom.
    Respect,
    -Brent

    Share
  7. Hey Jordan,

    I think there is more behind Dell’s action here than meets the eye. I work for Dell in networking and have worked on some SDN projects, but this is one I am not privvy to the details. (Also these comments are my own and Not Dell’s) Reading between the lines though on things I have heard internally and externally I would note you mention consortium but Dell is the only name mentioned … and Dell is mentioned as the leader of said consortium. It is true that Dell made a proposal to OMG, but there is a lot of speculation about who makes up the rumored consortium and what the pecking order is or isnt. I am eagerly awaiting to hear more details about who makes up the rumored consortium. While a commentor highlighted that HP and Cisco drive the vast majority of networking switches, SDN is a disruption that would never have even happened if left to the traditional network vendors. It has been lead by large cloud service providers & other SP’s, the academic and open source community, and Software vendors. I have no idea yet if/when/who/how might make up this rumored consortium that Dell would be part of but I will be eagerly anticipating more details … I do not see this yet as a vendor lead or a Dell lead consortium, to me that is prematurely speculative and I would wait for more details to emerge before drawing that conclusion.

    though one thing to note is … how exactly would a member of the community get involved with the Daylight consortium … thats right there is no way as it is not public and not open unlike with OMG where any proposal is subject to participation from any vendor as with other standards bodies.

    Share

Comments have been disabled for this post