The Future of White Box Networking in the Enterprise

Don't feel compelled to choose between white box and proprietary networking hardware just yet. Art Cole explains why.

By Arthur Cole | Posted Feb 20, 2015
Page of   |  Back to Page 1
Print ArticleEmail Article
  • Share on Facebook
  • Share on Twitter
  • Share on LinkedIn

To white box or not to white box? Is that really the question, or should the enterprise hold out for a more nuanced view of network infrastructure in the software defined future?

It seems that the farther along we get on the road to SDx, the more pertinent question is, what role will white box play in an increasingly distributed network environment?

To be sure, white box hardware will see a dramatic rise in web-facing hyperscale operations in the years to come, but the advantages the technology brings to the table start to erode as scale drops. This means the traditional enterprise facility, which still has a vital role to play in the emerging data ecosystem, won’t benefit as much from the conversion to commodity solutions.

I outlined some of the reasons why in a post on ENP’s sister site, itbusinessedge.com, a few weeks back; namel,y the overhead and management challenges that accompany white box deployments and the fact that the real cost savings fall primarily to organizations that can ODM their machines directly from Asia. Key applications like web transactions and Big Data processing will most certainly benefit from white box, but the jury is still out when it comes to traditional enterprise functions, which have shown a remarkable resilience within today’s integrated infrastructure.

This is the primary motivation for Cisco and other network giants to maintain the course when it comes to their leading enterprise portfolios. As Cisco CEO John Chambers told analysts recently, sales of the Nexus 3000 and 9000 switches are on the upswing even as more open SDN solutions hit the channel. With an integrated stack build on specialty ASICs, Cisco says it can deliver higher value than pure software solutions even as the headlines revolve around the wonders of automated hyperscale infrastructure. Google and Facebook may have devised innovative solutions for their particular needs, but it is by no means certain that the same models can be implemented successfully across the broader IT industry.

Companies like HP, on the other hand, see some middle ground between white box and customized hardware. HP recently teamed up with Cumulus Networks and Accton Technology to deliver “brite box” solutions to web-scale customers. This involves HP providing sales, distribution and support to what is essentially an Accton open switch running the Cumulus Linux distribution. The arrangement is intended to provide a branded but still standards-based open network solution through normal distribution channels, offering enterprises the flexibility of commodity hardware with the backing of an established datacenter partner. Initial products will be a 10G/40G spine switch and 10G leaf device, followed by 25/50/100G solutions for larger-scale operations and 1G configurations for specialty deployments.

Still, it is questionable whether even a company like HP can deliver all the tools necessary to implement a successful commodity infrastructure – white, brite or otherwise. A broad development ecosystem has already emerged around the white box industry, with companies like Gigamon offering extensive visibility and other functions across dynamic fabric architectures. This will become even more important as the enterprise encounters the unique security, application distribution and data isolation issues that characterize the software defined network.

This all contributes to the overhead I mentioned that skews the cost advantages of commodity hardware in limited-scale environments. Not only do you have to factor in the cost of the management platforms, but the skillsets to operate them as well.

So far from looking at a stark choice between commodity and integrated infrastructure, the enterprise would do well to keep both options on the table, perhaps with an eye toward integration should applications emerge that require scale-out functionality and uniformity up and down the stack.

If applications become as smart as developers say, than they should have no problem finding the right set of resources wherever they are deployed.

Comment and Contribute
(Maximum characters: 1200). You have
characters left.
Get the Latest Scoop with Enterprise Networking Planet Newsletter