Lippis Report 206: What the Market Said about SDN at the Open Networking User Group
Open Networking User Group (ONUG) meeting in Boston that we organized with partner Fidelity Investments. ONUG was a first for many reasons, but the biggest contributor to its success was its Board and the fact that the conference was closed to press, vendors and any recording. The ONUG Board, made up of chief network architects of the largest financial and retail concerns, developed ONUG sessions plus invited 10 pre-IPO firms to demonstrate their SDN solutions. The fact that the audience of 180 IT executives was able to openly discuss their open networking challenges, successes and hopes created a special mystique. Jayson Noland, Managing Director, Senior Analyst of Robert W. Baird, described the mystique best:
“There’s talk about standards and consortium, and what was apparent at ONUG is that there are very innovative early adopter IT leaders who will drive the market and determine the speed and success of true network transformation.”
Awais Nemat co-founder & CEO PLUMgrid on Open Networking at ONUG
Keep that in mind over the next several weeks, as there will be many vendor Software-Defined Networking (SDN) announcements that are significant and large. If you want to understand how the market will process these announcements, then you have to come to ONUG this fall in New York on Oct 29th and 30th; more on ONUG Fall 2013 in a few weeks. In this Lippis Report Research Note, I reflect back on ONUG and provide the top 10 take-aways.
Karl May President & CEO of Vello Systems on Open Networking at ONUG
At ONUG, IT business leaders shared their SDN experiences and key requirements to enable the growth of open networking. The output of ONUG was the identification of five recommendations to expedite the deployment of open networking, which include interoperability, vendor neutrality, programmable networks with northbound APIs, offering increased network monitoring and visibility, and a new business model. These five recommendations embody the promise of Open Networking, and ONUG will work hard to assure that this promise is delivered.
Guido Appenzeller, CEO Big Switch Networks on Open Networking at ONUG
1) Open Networks Must Be Interoperable Networks—Across SDN controllers and multi-vendor physical switches and hypervisors supporting industry standards, such as OpenFlow. Software-Defined Networks must address the entire network, including physical and virtual switches, and though overlays deliver value today, they are not the end game.
Craig Elliott co-founder and CEO of Pertino on Open Networking at ONUG
2) Open Networking Means Vendor Neutral Platforms—Support for multiple hypervisors, controllers, physical and virtual switches, network services and network silicon in a seamless multi-vendor environment with no vendor lock-in.
3) Open Networking Means Programmable Networks via Northbound APIs –Abstract network Command Line Interfaces (CLIs) to interface with applications and orchestration stacks to radically reduce the operational cost. Auto provisioning of physical and virtual network gear, compute and storage scheduling, and workload placement via an orchestration stack that does not require coding. The northbound APIs should also serve as an innovation injection to speed-to-market new applications, network services and design options.
Dante Malagrinò President & CEO Embrane on Open Networking at ONUG
4) Increased Network Visibility and Monitoring—Open networking also means visible networking so that troubleshooting, design, traffic flow optimization and others are enabled. Open networks should emit real-time network statistics to various traffic analytic and Big Data engines to determine network operational state.
5) Open Networking Business Model Needed—The industry needs to develop a business model that includes, but is not limited to, financial, support and service models. ONUG believes that for open networking to accelerate, the industry needs a viable, altruistic, truly open networking business model to drive innovation, fuel research and development, and deliver best of breed solutions without allowing individual vendor proprietary interests to derail SDN deployments. Who will be the Red Hat of Open Networking?
Aligning Network Management with Converging Operational Priorities
As a little more than a month has passed since ONUG, I provide my top 10 take-aways:
1) Top-Down Permission: Both service providers and Enterprise IT business leaders have received permission to change the way they approach 25 years of networking. Both came in the form of economics. Service providers see their revenue and demand cost curve bending at a rate of Moore’s law, but their network cost is not and will intersect revenue. That is, network cost will outpace revenue, leaving them with an unprofitable, and of course, unsustainable, business model. IT business leaders, operating under the echo effect of 2008, were directed to lower network cost, as network Total Cost of Ownership or TCO is getting larger than server and storage TCO. Try as they did to lower TCO, the only thing they could control is capital or acquisition cost, which only represents approximately 25% of TCO. They were given permission to change the way they design networks if it meant lower OpEx. Therefore, both service providers and Enterprise IT have the same mantra, and that is to change network architecture to radically lower TCO.
Best Practices in Big Data Monitoring
2) Bottom-Up Demand: One large bank showed a graphic plotting the rate of adoption of business unit self-service Virtual Machine adds, pauses and deletions. The curve was exponential, and it’s just the beginning as virtualized network overlay adoption is in its infancy considering only 48% of servers are virtualized and a much smaller number of virtual machines are networked. This is a tsunami that’s forcing network architecture change as business units demand self-service IT delivery that’s rapid, meaning minutes; network provisioning is stuck in weeks, thanks to firewalls, appliances, IP addressing, etc. Chief Network Architects (CNAs) are feeling the pressure to automate compute, storage and network provisioning, again providing permission to change network architecture to obtain it.
Arista EOS™ Precision Data Analysis with DANZ
3) White Box Networking Is Coming: CNAs want network gear that can be provisioned by orchestration tools, such as OpenStack, vCenter, etc. An open source reference design for network switching is desired.
4) Linux Is the New Network OS: Linux is the common denominator across IT silos of server, storage, networking and applications. Linux, as the network OS, will contribute to breaking down organizational barriers and enable networking to be part of orchestration.
5) OpEx Has to Fundamentally Change: A number of CNAs expressed that if network equipment cost was zero dollars, they still could not afford it. While, perhaps, tongue-in-cheek, the key point is that their operational cost is so high that they cannot add more tasks or load to that system. In fact, they need OpEx relief and a fundamentally new operational model for networking.
Save Your School Money with Lower Networking Costs from Avaya
6) Fundamental Re-thinking of Networking Required: Networking has not changed significantly in 25 years. Yes, speeds have increased, protocols have been added, WLANs have come into the market, etc. But layer 2 and 3 packet forwarding with three-tiers of network equipment has not changed in two and a half decades along with its management. For most network engineers, this is all they know. In fact, all large corporations would hire with CCIE certification requirements, which has resulted in a staff whom all think the same. Network engineers need permission to change the way they think about networking and in essence, be re-programmed.
Intelligent Networking for the Optimized Agile Datacenter
7) Engagement Model Changing: Some of the above is starting to manifest itself in a change in the engagement model between buyers and sellers of network equipment. The engagement is starting to be more technical in nature vs. relationship based. In addition, IT leaders are intending to use their key leverage of engineering talent and power of the purse to get open networking the way they want. In essence, power is shifting back toward the buyers of network gear.
8) SDN Network Visualization Is the Killer Application: While there is much discussion about SDN use cases in the data center, service provider, cloud providers, etc., at ONUG, the killer application was network visualization as it lowers acquisition cost by a whopping 95%, yes, 95%. Put another way: non-SDN based Network Visualization solutions are 20 times more expensive. In addition to lower capital spend, SDN-based network visualization solutions increase flexibility to tap the network from virtually anywhere. This flexibility is being afforded to business unit managers by offered self-service initiated monitoring. Lastly, SDN-based network visualization solutions cover more of the network so that a larger amount of the network that can be visualized. The solution set included Big Switch Network’s Big Tap plus OpenFlow switches from Arista, IBM and others with GigaMon, cPacket, Anue etc., providing the critical service node functions.
9) SDN Pilots Up, But Limited Deployments: At ONUG, 56%, 28% and 16% of this early adopter audience were in review, piloting and limited deployment of SDN projects, respectively. When asked when reviewing will turn to pilots, the bulk of the audience responded one to three years. That’s a long time and a surprising result. First, most deployments are network virtualization, and they will continue to dominate for some time. ONUG members are waiting for next generation silicon for physical SDN deployments, as OpenFlow implementations are currently limited. The next generation of merchant and custom ASIC silicon will natively support OpenFlow, increasing the number of simultaneous flows per switch, reducing the time to flow initiation, increasing the Switch Flow Learning Rate and increasing Switch Flow Table Capacity. If this rate of adoption does not change—that is, if pilots and deployments do not accelerate—then most start-ups may find that they are in a race against financial runway and revenue ramp-up.
10) ONF + OpenStack + Open Compute: Open Networking is bigger than OpenFlow. To deliver on the promise of Open Networking above ONUG believes that Open Networking will include standards and open source assets from the ONF, OpenStack, Open Compute, W3C or World Wide Web Consortium and others.
As Jayson Noland says, “there are very innovative early adopter IT leaders who will drive the market and determine the speed and success of true network transformation.” I couldn’t agree with him more. These top ten take-aways are a good measure of the state of open networking from those with the engineering and budget resources to deploy them.