2013 IDC Directions Part III- Where Are We Headed with Software-Defined Networking (SDN)?

Introduction:

In the third article on the IDC Directions 2013 Conference (March 5th in Santa Clara, CA), we take a hard look at Software Defined Networking as presented by Rohit Mehra, IDC VP for Network Infrastructure.

Note: Please see 2013 IDC Directions Part I for an explanation of the “3rd Platform” and its critical importance to the IT industry and Part II on New Data Center Dynamics and Requirements


Background:

IDC firmly believes that the “3rd Platform” is the way forward and that the network is the vital link between cloud computing and mobility.  “The Cloud is evolving into a comprehensive, integrated application delivery model incorporating all four elements of the 3rd platform,” said Mr. Mehra.

  • Cloud Apps require network agility, flexibility and must support higher east-west traffic flows (between servers in a cloud resident data center).
  • Mobile access is crucial with the proliferation of mobile devices (e.g. smart phones and tablets) and continued exponential growth of mobile data traffic.
  • Variable end points and different traffic patterns must be supported.
  • Social networking is being integrated with other enterprise applications. This is resulting in increased volumes of cloud data exchanges with client devices and more server-to-server traffic flows.
  • Big Data/Analytics results in scale-out computing which needs scale-out networking. Greater application-to-network visibility will be required.

As a result of these strong 3rd platform trends, Mr. Mehra said, “Application access/delivery is dependent on the  cloud resident data center and enterprise network.  Both will need to become more dynamic and flexible with SDN.”

IDC asked IT managers: What was the main reason you needed to Re-Architect The Network to support Private Cloud? The top three reasons were:

  • We needed to ensure security between virtual servers
  • We needed more bandwidth to support the virtualized applications
  • The network became a bottleneck to new service provisioning

Rohit said that SDN could address those issues and was gaining traction in the data center.  “”SDN provides better alignment with the underlying applications, along with improved flexibility and command of the network,” he said.  Through SDN models, companies will likely find it easier to implement virtual cloud hosting environments, according to Rohit.

A recent IDC study SDN Shakes Up the Status Quo in Datacenter Networking projected that the SDN market will increase from $360 million in 2013 to $3.7 billion in 2016.

SDN Attributes include:

  • Architectural model that leads to network virtualization
  • Dynamic exchange between applications and the network
  • Delivering programmable interfaces to the network (e.g., OpenFlow, APIs)
  • Management abstraction of the topology
  • Separation of control and forwarding functions (implemented in different equipment)

Rohit stated that SDN was NOT another name for “Cloud-based Networking” and that they were each in functionally different domains:

  • Cloud-based Networking involves emerging network provisioning, configuration and management offerings that leverage cloud Computing and Storage capabilities.
  • It’s a “Network As A Service” model that can apply to routers, WLAN, Unified Communications, app delivery, etc.

Rohit expects network equipment and network management vendors to add these capabilities to their platforms in 2013.

Three Emerging SDN Deployment Models are envisioned by IDC:

1. Pure OpenFlow (more on the role of Open Flow later in this article)

  • Driven largely by being open and standards-based (by Open Networking Foundation or ONF)
  • Inhibited by fluidity of OpenFlow release schedule; limited support in existing switches

2. Overlays

  • Exemplified by Nicira/VMware’s Network Virtualization Platform (NVP), IBM’s DOVE, others
  • Some vendors that started out offering “pure OpenFlow” have adopted overlays (Big Switch Networks)

3. Hybrid (Overlay, OpenFlow, Other Protocols/APIs)

  • Put forward by established networking players such as Cisco and Juniper
  • Offer SDN controller, with support for distributed control plane for network programmability and virtualization, etc.

Image courtesy of IDC.

Image courtesy of IDC.

SDN vendors are offering SDN solutions from four different perspectives. Many of them solely target one of the four, while others offer a combination of the following:

  • SDN enabled switches, routers, and network equipment in the data/forwarding plane
  • Software tools and technologies that serve to provide virtualization and control (including vSwitches, controllers, gateways, overlay technologies)
  • Network services and applications that involve Layers 4-7, security, network analytics, etc
  • Professional service offerings around the SDN eco-system

SDN’s Place In The Datacenter-IDC sees two emerging approaches:

1. Some vendors will push SDN within the framework of converged infrastructure (servers, storage, network, management)

  • Appeals to enterprises looking for simplicity, ready integration, and “one throat to choke”
  • Vendors include HP, Dell, IBM, Cisco, Oracle and others

2. Some IT vendors will offer a software-defined data center, where physical hardware is virtualized, centrally managed, and treated as an abstracted resource that can by dynamically provisioned/configured.

  • Vendors include VMware, Microsoft, perhaps IBM

Image courtesy of IDC.

Image courtesy of IDC.

SDN Will Provide CapEx and OpEx Savings:

OpEx

  • Better control and alignment of virtual and physical resources
  • Automated configuration, and management of physical network
  • Service agility and velocity

CapEx

  • Move to software/virtual appliances running on x86 hardware can reduce expenditures on proprietary hardware appliances
  • Support for network virtualization improves utilization of server and switch hardware
  • Potentially cheaper hardware as SDN value chain matures (long-term, not today)

Role of OpenFlow as SDN Matures:

  • Initial OpenFlow interest and adoption from research community, cloud service providers (e.g., Google, Facebook) and select enterprise verticals- e.g., education
  • Led to successful launch of Open Networking Foundation (ONF)
  • Centralized control and programmability is the primary use case- but that may be its limitation
  • At a crossroads now- OpenFlow taking time to mature and develop, while alternate solutions are emerging
  • As the market for SDN matures, OpenFlow is likely to be one of the many tools and technologies (but not the ONLY protocol to be used between Control plane virtual switches/servers and Data forwarding equipment in the network)

SDN Challenges and Opportunities- For SDN Vendors and Customers:

  • Vendors will need to consider adding professional services to their SDN portfolio
  • The value chain will benefit from these services early within the market adoption cycle
  • Need for SDN certification and training programs to engage partner and customer constituencies and to reduce political friction associated with change
  • Education on use cases is critical to getting vendor message across, and for creating broader enthusiasm for change among customers
  • Customers must ensure that they have the right mix of skills to evaluate, select, deploy, and manage SDN
  • The battle to break down internal silos will intensify alignment of applications and networks means an alignment of teams that run them

Image courtesy of IDC.

Image courtesy of IDC.


Conclusions:

1.SDN is rapidly gaining traction as a potentially disruptive technology transition, not seen for a long time in networking
2.SDN is riding the wave of a “Perfect Storm”, with many individual market and technology factors coming together:

  • Growth of Cloud Services/Applications
  • Focus on converged infrastructures (compute/storage/network)
  • Emergence of Software-Defined Data Center (SDDC)
  • Lessons learned (and benefits) from server virtualization

3.SDN brings us closer to application and network alignment with next-generation IT
4.Incumbent vendors will need to find the right fit between showing leadership in SDN innovation and balancing existing portfolio investments


Addendum: Software Defined Networks and Large-Scale Network Virtualization Combine to Drive Change in Telecom Networks

In a March 7th press release IDC wrote that SDN along with large-scale network virtualization are two emerging telecom industry technologies that will combine to drive a more software-centric and programmable telecom infrastructure and services ecosystem. These complementary and transformative technologies will have a sustained impact on today’s communication service providers and the way they do business.

“IDC believes that the rapid global growth of data and video traffic across all networks, the increasing use of public and private cloud services, and the desire from consumers and enterprises for faster, more agile service and application delivery are driving the telecom markets toward an inevitable era of network virtualization,” said Nav Chander, Research Manager, Telecom Services and Network Infrastructure, IDC.

“SDN and large-scale network virtualization will become a game shifter, providing important building blocks for delivering future enterprise and hybrid, private, and public cloud services.”  he added.  Additional findings from IDC’s research includes the following:

  • Time to service agility is a key driver for SDN concepts
  • Lowering OPEX spend is a bigger driver than lowering CAPEX for CSPs
  • Network Function Virtualization and SDN will emerge as key components of both operator service strategies and telecom networking vendor’s product strategies

The IDC study, Will New SDN and Network Virtualization Technology Impact Telecom Networks? (IDC #239399), examines the rapidly emerging software-defined network (SDN) market, the developments in large-scale network virtualization, and a new Network Functions Virtualization ecosystem, which are likely to have an impact on telecom equipment vendors’ and CSP customers’ plans for next-generation wireline and wireless network infrastructure.


References:

http://community.comsoc.org/blogs/alanweissberger/fbr-sdn-result-40-drop-switchrouter-ports-deployed-service-providerslarge-en-0

http://community.comsoc.org/blogs/alanweissberger/googles-largest-internal-network-interconnects-its-data-centers-using-software


IEEE ComSocSCV had the two leaders of the SDN movement talk at one of our technical meetings last year. Their presentations are posted in the 2012 meeting archive section of the chapter website:

Date: Wednesday, July 11, 2012; 6:00pm-8:30pm
Title: Software Defined Networking (SDN) Explained — New Epoch or Passing Fad?
Speaker 1: Guru Parulkar, Executive Director of Open Networking Research Center
Subject: SDN: New Approach to Networking
Speaker 2: Dan Pitt, Executive Director at the Open Networking Foundation
Subject: The Open Networking Foundation
http://www.ewh.ieee.org/r6/scv/comsoc/ComSoc_2012_Presentations.php

6 comments for “2013 IDC Directions Part III- Where Are We Headed with Software-Defined Networking (SDN)?

  1. March 30, 2013 at 10:49 am

    SCALABILITY CONCERNS IN SDN: What fundamentally differentiates SDN from traditional data networks is the separation of control from the forwarding plane. This decoupling leads to interesting properties. Most important, data and control plane components can evolve independently, as long as we define a standard application programming interface (API) between the two. Data forwarding elements are responsible for switching and are usually built from highly specialized application-specific integrated circuits (ASICs), which evolve significantly slower than software based control plane components. Also, the possibility of creating a centralized view of the network creates tremendous potential opportunities for simplifying the control applications, and therefore accelerating change and innovation in the control plane.

    This decoupling, however, has its own pitfalls too. First, defining a standard API between the two planes is not trivial. Technically, this API should be able to handle the needs of various architectures, and should be able to
    facilitate the independent evolution of both planes. Also, all or a majority of switch vendors should adopt the same standard API for it to be useful; otherwise, networks will be tied to specific vendors, which might lead to proprietary layers, preventing rapid change and innovation in networks. Second, decoupling data and control planes brings about scalability concerns. Moving traditionally local control functionalities to a remote controller can potentially result in new bottlenecks. It can also lead to signaling overheads that can be significant depending on the type of network and associated applications. READ the entire article at:
    http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=6461198

  2. March 28, 2013 at 12:45 pm

    2013 TiECon track on Software Defined Infrastructure:
    Computing has seen a significant paradigm shift away from the old norms of hardware based silos of compute, network, storage, into a much more dynamic and responsive Software Defined Infrastructure (SDI). The SDI segment is witnessing tremendous innovation with brand new ways of computing, networking and storage allowing for incredible cost savings for enterprises. Everything you want to know about this market will be discussed by industry experts at TiEcon 2013 SDI track on Friday, May 17th. More info at: http://tiecon.org/sdi

  3. Anonymous
    March 25, 2013 at 5:40 pm

    Very infomrative article reflecting IDC’s view of SDN. Here’s what Gartner Group just wrote: Software Defined Networking Creates a New Approach to Delivering Business Agility

    Software defined networking (SDN) is creating a lot of excitement in data centers, but current technology is still relatively immature. In the new research note “Ending The Confusion Around Software Defined Networking (SDN): A Taxonomy”, Joe Skorupa, vice president and distinguished analyst at Gartner, explains that SDN is not only limited to data center and service provider networks. Skorupa answered some questions around the current state of SDN and how it will evolve:
    Q – What is Software Defined Networking?
    A – SDN is a new approach to designing, building and operating networks that supports business agility. SDN brings a similar degree of agility to networks that abstraction, virtualization and orchestration have brought to server infrastructure.
    Q – What Models Exist for SDN Deployment?
    A – Three deployment approaches are possible – switched-based, overlay and hybrid. For greenfield deployments, particularly when the cost of physical infrastructure and multi-vendor options are important, a switch-based model will be common. The biggest limitation to this approach is that is currently does not leverage existing L2/L3 network equipment. Press Release: http://www.gartner.com/newsroom/id/2386215
    Joe Skorupa Bio: http://www.gartner.com/AnalystBiography?authorId=24834

  4. Anonymous
    March 25, 2013 at 11:02 am

    Thanks for an excellent summary of this IDC Directions session. Could SDN be a bit over-hyped? During the fiber build out boom from 1998-2001, many vendors of Multi Service Provisioning Platforms (MSPP) referred to their products as “God Boxes” because of all the functionality packaged in their equipment. In fact fiber to the building really never happened so there was no need for a MSPP.

    Now the “God Box” comes to SDN- according to this article:
    http://www.sdncentral.com/sdn-blog/god-box-idea-hits-sdn-network-tap/2013/03/ Caveat Emptor: “The networking industry is littered with the carcasses of companies that tried to put too much into one box.”

  5. March 23, 2013 at 3:28 pm

    Juniper evidently recognizes the power & potential of SDN (even if I don’t). Network World says the company will unveil a programmable core switch for software-defined networking. http://www.networkworld.com/news/2013/032213-juniper-sdn-268023.html
    I will refrain from expressing my own opinion till I learn more at the next SDN Conference in mid April. However, I’m always a skeptic on new, highly touted networking technologies as I’ve seen almost all of them fail or be significantly delayed.

  6. Brad Stern
    March 23, 2013 at 9:51 am

    According to Barrons on-line Mar 21, 2013 (subscription required), FBR wrote:
    “Could routers and switches go the way of the dodo? Yes, but slowly and not entirely. Our research does suggest that service providers and large enterprises are moving toward a significant reduction in the number of routing and switching portsdeployed as a result of:
    1) the continued shift from legacy packet and time-division multiplexing (TDM) networks into Ethernet networks;
    2) a networkinfrastructure transition to common compute-based platforms; and
    3) consolidation of network functions to compute platforms based in metro or regional network data centers.

    Together, we expect these trends to drive more than a 40% reduction in the number of switching routing ports in service provider networks.

    New networking technologies enable new nontraditional competitors. Enabled by new silicon-based innovation, network virtualization, and advances in open-source software, new computing/switching platforms may significantly reduce the need for purpose-built hardware, potentially diminishing demand and margin for many of the network components that Juniper offers to service providers andlarge enterprises. We expect this trend to drive networking vendors into software more quickly than many believe in an attempt to preserve revenue growth, creating a situation whereby the decline in hardware revenue may more than offset gains in software revenue.”

Leave a Reply