Posts Tagged ‘SDN’

Well, the 2013 version of Interop is now in the books and while conventional wisdom dictates that “what happens in Vegas, stays in Vegas,” I thought there were a number of themes at the show that rose above the general noise of the event and are worth sharing. In no particular order, these themes were:

  • Software Defined Networking (SDN) needs a better definition. It’s amazing to me the number of vendors that now claim to be in the SDN market. It seems that if a vendor is programmable, virtual or software-based, it can claim to be an SDN vendor. When you look at these attributes, who doesn’t fall into one of these categories? And when a definition is so broad that it means everything, it actually means nothing. There’s certainly a common theme that all the vendors talk about – legacy networks are complicated and rigid and this is out of alignment with the rest of IT. How this gets solved is still up for debate, and SDNs will not become a pervasive technology until the debate is at least partially answered.

All eyes were on Silicon Valley and the Open Networking Summit this week. One of the big topics of conversation was Intel’s push into an already highly competitive software defined networking (SDN) space.

In theory, this move by Intel makes sense. SDNs transform the data center and create an opportunity for low-cost switch manufacturers to become a more important part of the data center. However, “theory” and “reality” are two different things, and I don’t believe a pure, white box switch really works in this market.

Over the past decade, chip companies such as Broadcom, Marvell, Mellanox and now Intel have tried to create software to complement the switch chips, in a bid to to sell the switch chips. Intel has WindRiver, Broadcom acquired LVL7, Marvell has Radian and even Mellanox uses open source software. Software plus chips equals success, right? Again, in theory it makes sense, but in practicality, this hasn’t worked.

Last week, Alcatel-Lucent (ALU) held its annual Industry Analyst conference in Annapolis, Maryland. Unified Communications has historically been the primary focus for ALU’s go-to-market strategy, but the company has spent the last few years beefing up its OmniSwitch data networking portfolio as well. In fact, if you recall, ALU was the focal point of this Network World Article where the company beat out Cisco for a network project in its own home state.

Like every other network vendor, ALU has been trying to jump on the market opportunity created by the rise and complexity of server virtualization. I recently did some research that pointed out that a small amount of server virtualization saves both capex and opex. However, highly virtualized environments, meaning those that are more than 50% virtualized, have actually seen operational costs rise by as much as 20%. High amounts of server virtualization create unpredictable traffic flows that can wreak havoc on the network.

Big Switch stole many of the networking headlines this week when it announced its Switch Light software release. Switch Light, based on the open source technology Indigo, can be used on commodity white box switches to create an OpenFlow-based switch than can be used as part of a software defined network implementation. One of the elements of the press release that I felt flew under the radar, was that Extreme Networks would be the only “mainstream” network vendor that was committing to this reference architecture for a rack switch.

Extreme will later this year introduce its Slalom switch, which will be an optimized SDN switch running the thin virtual network software. Slalom will complement Extreme’s Open Fabric portfolio that has a number of SDN-capable switches in both stackable and chassis-based forms. Slalom will be a 48-port Gig-E Ethernet switch that will be based on a “white box” solution.

Despite the hype around software defined networks (SDNs), the industry has yet to find a legitimate “low-hanging fruit” for the network technology. It appears, though, that one might be emerging, as several vendors have announced TAP aggregation as an SDN application. Earlier this year, when Big Switch launched the company it announced its own product called Big TAP. Later, Cisco released TAP aggregation for its Cisco ONE controller.

Well, last week Arista Networks announced its DANZ data analyzer, which is an application based on EOS (Extensible Operating System).

To say the TAP aggregation market is hot is an understatement. This once niche market has become one of the fastest growing markets in the network industry. In the last year, market leader Gigamon started prepping for IPO, VSS was acquired by Danaher Corporation, NetScout bought ONPATH, and Ixia purchased Anue.

With the hype around software-defined networks (SDNs) having grown as high as it has, almost every vendor is looking for an angle to capitalize on the opportunity. I’ve noticed recently that many of the vendors, particularly the big-box vendors, are focused on the concept of “network programmability.” While I agree that programmability is a component of SDN, it shouldn’t be the sole focus of the technology. As an example, Cisco has been pushing the Python-based ONE (Open Networking Environment), Juniper has JunosV App Engine and HP has its own programming environments. Like I said, I think these are important parts of the overall SDN solution, but it’s not aligned with where buyers are today.

I recently ran a survey with TechTarget and one of the questions asked was “How do you think SDNs can help your company?” The number one response, at 50% of the respondent base, was “simplify network architecture.” The No. 2 response (46.9%) was “reduce network hardware costs.” The third most popular response (45.4%) was to enable network management. Way down on the list, at 7.7%, was to “provide a more programmable network.”

Programmability could be used to improve network management, but it really doesn’t have any impact on the first two options: simplification of architecture and reduction of hardware costs. Those problems are solved through the use of low-cost, standards-based switching hardware that is simple to deploy and manage. I’m not saying that the big box vendors are trying to slow innovation or aren’t taking SDNs seriously. In fact, it’s quite the opposite. The mainstream vendors do want to offer a credible SDN story, but they do need some level of vertical integration to keep the “end-to-end” value proposition intact.

For most mainstream companies, the limitations of a vertically integrated solution will probably be fine, at least in the near term. In fact, I’m not sure most mainstream companies would even know where to procure low-cost, non-brand-name network hardware. However, for those organizations with hyper-scale data centers where the network is the business, being able to cut the cost of switching through the use of low-cost, simplified network infrastructure can be significant. In a sense, it’s the same market where SeaMicro thrived by offering simpler, lower-cost rack servers.

While much of the industry will focus on the programmability of the network, the companies that want to leverage the cost benefit of SDNs now should think “open” first, and then look to leverage programmability once the architecture has been simplified and the overall network has become more manageable.

As most people who follow the networking industry know, Juniper outlined its vision for software-defined networks (SDN) at last week’s Global Partner Conference in Las Vegas. I’ve had a bit of time to digest the information, answered a number of questions about it, and I thought it was time to share my thoughts on what they announced and the impact to the industry.

First, summarizing what Juniper announced, they began the presentation with a number of myths regarding software-defined networks and then articulated Juniper’s principals of SDNs. I’m not going to repeat these, but if you missed the webcast, Jim Duffy’s article does a good job of highlighting them.



Insight and Influence Through Social Media
ZK Research: Home
Google+
Twitter
LinkedIn
Facebook
RSS Feed
ZK Research is proudly powered by WordPress | Entries (RSS) | Comments (RSS) | Custom Theme by The Website Taylor