05:04 PM
Greg Ferro
Greg Ferro
Connect Directly
Repost This

SDN Doesn't Mean Cheaper Networking

The most common prediction about SDN is that networking will get cheaper. It won't. Here's why.

The most common prediction about SDN is that networking will get cheaper. In fact, IT will spend the same as it always has on networking, but the money will be allocated in different ways. Meanwhile, market pressures will force traditional vendors such as Cisco to adopt new revenue models.

Let's start with network hardware. I predict network hardware will get cheaper because it must be replaced and upgraded more often in the years ahead.

Today, a network device has a lifespan of at least five years; in many cases, it may be up to 10 years before it is considered for replacement. However, technology developments will speed up hardware refreshes; the next generation of networking is likely to be replaced every three years as the drive into 10-GbE ports will soon be followed by 40-GbE and 100-GbE interfaces.

We'll also see new silicon with lower power consumption, support for overlay networking and increased forwarding performance compel faster product churn.

It's reasonable to expect unit prices to fall as customers rapidly cycle out network assets in the same way that server assets are removed to gain new features or functions. However, while unit prices drop, overall network spending will remain the same because of the need to replace hardware more often.

[Network overlays virtualize the network for SDN. Get details in "Network Overlays: An Introduction."]

Another way hardware costs could be lowered is by changing vendor practices around expensive optical and copper interface modules. A significant percentage (often more than 50% ) of the standing cost of network devices is largely due to these modules, which deliver little customer value. The future of networking is services, not connectivity.

It's not necessary for "vendor supported" SFP modules to cost in excess of $300 per unit--even less so to enforce this in software. If the standard is badly written that there are reliability or production problems, then vendors should kick the IEEE into action to produce better standards that are resistant to low-quality manufacturing and prevent unreliable manufacturers from making poor copies.

It is beyond the scope of this article to explain the foolish choices made by the IEEE during the standards process that have resulted in extortionate pricing for what could be low-cost items.

SDN could get cheaper if there was a concerted effort from vendors and customers to drive new standards with reasonable technology choices for cheaper interface modules. Shipping SFP modules adds zero value to customer features or functions and customers resent paying for them.

Off-Brands Won't Flourish

Some make the argument that unbranded hardware will erode the market for premium-priced equipment. I don't think so. There's a lot of inexpensive networking equipment available today from vendors like Netgear and D-Link, including low-cost chassis switches. Yet the majority of customer continue to buy branded hardware for quality and assurance reasons.

Why would SDN change this purchasing practice ? The underlying reasons for purchasing branded or known hardware--customers want a trusted brand and partner in their data centers--aren't changed by SDN.

Niche use cases from mega-scale data centers simply aren't relevant to enterprises or service providers. Google and Facebook purchase low-cost network equipment direct from the manufacturer, but they also perform extensive testing to prove the hardware is viable and fit for purpose.

There is identifiable risk that low-cost products may have higher failure rates, poor quality software or other serious operational issues. While the mega-scale companies have software platforms designed to handle failure easily, enterprises are not experienced in handling these kinds of risks.

Cheaper, Simpler Software

The price of the network OS on the device is also likely to fall in the years ahead. Today, customers purchase a license for an integrated NOS that runs on the specific device. These NOSes often have thousands of features, functions and services that ultimately deliver network services.

However, a key value of current SDN platforms such as VMware NSX is that network services will be provided in a hypervisor or in a virtual machine.

Two factors should lead to lower prices. First, a software license today comes with the device that is intended to be used for at least five years. This timeline is factored into the customer value proposition (that is, the purchase price). As customers rotate hardware faster, the software license has less value over the life of the device, which will lead to price pressures on suppliers.

Second, market competition for cheaper NOSes is being introduced by startups. For instance, Cumulus Networks and Big Switch Networks have introduced NOSes that operate independent of the network hardware. ODM suppliers are already shipping small numbers of network switches that support these open NOS environments.

These startups license their software on a recurrent revenue model, and any purchase has a serious impact on the short-term revenue of incumbent vendors. This should drive lower prices on existing equipment.

As stated earlier, rapid device rotation is likely to occur, so software costs must go down to match perceived and/or delivered product value. Product rotation is needed to support greater bandwidth, faster forwarding performance and increased density.

Next Page: SDN Platforms Add Up

1 of 2
Comment  | 
Print  | 
More Insights
Newest First  |  Oldest First  |  Threaded View
User Rank: Apprentice
1/16/2014 | 2:10:47 PM
re: SDN Doesn't Mean Cheaper Networking
Why does the name "Greg Ferro" always remind me of "IEEE bashing?" I generally like your posts, but sometimes I think it's just the same old story GǪ If IEEE does such a terrible job, why don't you join them, attend meetings and show them how to do better?
User Rank: Apprentice
11/13/2013 | 10:14:58 PM
re: SDN Doesn't Mean Cheaper Networking
ACM - You are correct. This is all about the applications and not so much the network. SDN is almost exclusively focused on the data center and will likely yield an initial increase in expenditures with the plan to reduce expenditures over time through efficiencies. But in truth, the costs of non-uniform and inconsistent application delivery combined with the business opportunity costs associated with not being able to deploy new applications because of bandwidth limitations, security policy conflicts and IT resource costs are the real drivers. Critical business applications must work and new ones must be deployed to meet business objectives.

ADN (Application Defined Networks) - wide area virtual overlay networks - addresses how application-secific traffic reaches the data center where SDN may reside, ADNs change the enterprise network model completely by enabling application-to-application (A2A) networks versus the traditional closed site-to-site (S2S) networks. ADN implementations are live today and have provided immediate returns on investment (or savings) that lower IT budget costs at all remote sites trying to access the data center.
User Rank: Apprentice
9/29/2013 | 6:18:07 PM
re: SDN Doesn't Mean Cheaper Networking
I think we need to be more nuanced on the idea of "commodity". Today, "network connectivity" is a differentiated service, tomorrow "network connectivity" will be a commodity. The networking industry will move to new methods to scale network connectivity for larger systems and this will simplify network hardware. This is similar in concept to x86 servers having high levels of differentiation 5 yrs ago but today are much more similar.

In this article I postulate that this is a convergence of increased sales volumes driving less product types, and the increasingly rapid rotation of assets. In servers, products got cheaper but even more products were purchased.

It seems logical to conclude that the same will happen in networking. Simpler products made in volume that are replaced more quickly.

Also, in a similar way that VMware used software to enhance services derived from servers, then SDN software & platforms will be used to create new and enhanced services from network hardware.

So if you agree with those statements, then yes, at a very high level the same thing will happen. At a more detailed level, networking is dramatically different because it is a tightly coupled and connected system that spans a larger space (WAN, Internet, Security and Networking) whereas servers are standalone elements with loose coupling.

I see storage changing in similar way to virtualization because also a loosely coupled system that is located in one place, the data centre.

Rob Parten
Rob Parten,
User Rank: Apprentice
9/24/2013 | 9:32:40 PM
re: SDN Doesn't Mean Cheaper Networking
I agree this will make the hardware layer more of a "commodity service" and drive lower prices and I was singing that tune while at VMWORLD 2013.

However, do you believe this is going to shake up the giant and allow the other competitors to shine because they're, typically, already at a lower cost point? Or, from your experience and perspective, do you see people remaining loyal to what they've always known?

I can remember the days of Proliant vs Dell vs IBM because everything ran on bare metal. Now, what I hear is people wanting tighter configuration and management solutions with their virtualized infrastructures (Cisco UCS and HP FlexConnect) and I also hear "who cares, compute is compute is computer at the hypervisor".

Will SDN do to physical networking equipment what ESXi did to servers?
User Rank: Apprentice
9/24/2013 | 8:02:12 PM
re: SDN Doesn't Mean Cheaper Networking
Yes, SDN is not additional spend. At the same time I would propose that SDN will provide better services and capabilities and thus justify maintaining the network spend at current levels.
User Rank: Apprentice
9/24/2013 | 8:01:05 PM
re: SDN Doesn't Mean Cheaper Networking
That's the premise of my argument. Network equipment must get cheaper so it can be rotated more quickly and to allow purchase of SDN software that provides new services.

Customers don't love the network to spend _more_ they will ONLY be happy to have better networking at the same price.
Drew Conry-Murray
Drew Conry-Murray,
User Rank: Apprentice
9/24/2013 | 6:07:04 PM
re: SDN Doesn't Mean Cheaper Networking
It seems to me that one of the major value propositions for SDN are the applications that will supposedly take advantage of network programmability. Assuming the SDN plumbing gets worked out, applications may be where the money goes.
User Rank: Apprentice
9/24/2013 | 5:22:42 PM
re: SDN Doesn't Mean Cheaper Networking
Interesting. I am with you on the hardware arguments.

On the software side, in practice, the operating system is typically given away as part of the embedded system or discounted to zero. I don't know how much further down the software can go. I think when the hardware prices go down, you will see a shift in pricing mix to try to recapture some of that as software. I would expect to see a bump in software pricing before we start the decline there.

But that is just details. You overarching thesis that SDN is substitutive (ie, not creating net new spend) is exactly right.

Where there will be opportunity is in professional services. If the future of IT is more integrated, tightly orchestrated, and automated (through DevOps or some middleware layer), there will be professional services required to stand it all up. This won't be net new spend either, but it could be a lucrative space.

Given that controllers appear to be moving closer to free than premium entities, this might be where a good chunk of that money flows. The rest (as you are suggesting) would go to controller applications that actually solve some use case.

-Mike Bushong
More Blogs from Commentary
Edge Devices Are The Brains Of The Network
In any type of network, the edge is where all the action takes place. Think of the edge as the brains of the network, while the core is just the dumb muscle.
SDN: Waiting For The Trickle-Down Effect
Like server virtualization and 10 Gigabit Ethernet, SDN will eventually become a technology that small and midsized enterprises can use. But it's going to require some new packaging.
IT Certification Exam Success In 4 Steps
There are no shortcuts to obtaining passing scores, but focusing on key fundamentals of proper study and preparation will help you master the art of certification.
VMware's VSAN Benchmarks: Under The Hood
VMware touted flashy numbers in recently published performance benchmarks, but a closer examination of its VSAN testing shows why customers shouldn't expect the same results with their real-world applications.
Building an Information Security Policy Part 4: Addresses and Identifiers
Proper traffic identification through techniques such as IP addressing and VLANs are the foundation of a secure network.
Hot Topics
Converged Infrastructure: 3 Considerations
Bill Kleyman, National Director of Strategy & Innovation, MTM Technologies,  4/16/2014
Heartbleed's Network Effect
Kelly Jackson Higgins, Senior Editor, Dark Reading,  4/16/2014
White Papers
Register for Network Computing Newsletters
Current Issue
Twitter Feed