Think Outside the White Box

So I was chatting to John Chambers the other day, as you do [Ed. note: Or don't? It actually took Steve 15 years to get an interview with the CEO of Cisco… nice try, Mr Saunders], when he told me something jaw-dropping. (See CEO Chat With John Chambers, Cisco.)

In the future, Chambers said, he expects Cisco's principal competition to come from "white box" solutions -- that is, open source software running on generic server and switch hardware, rather than the proprietary solutions from incumbent providers such as Alcatel-Lucent (NYSE: ALU), Ericsson AB (Nasdaq: ERIC), Huawei Technologies Co. Ltd. and Juniper Networks Inc. (NYSE: JNPR) that currently dominate the market.

The concept of white box networking isn't new of course. However, until now, it's only been considered viable for enterprise networks, data centers, and, just maybe, the very edge of service provider networks (and even then in special instances).

But what John (obviously, we're on first name terms now) was talking about was something far wider in scope: extensive use of white box platforms across vast swathes of wide area networks, reaching right up to the carrier core -- a near network "white wash," if you will.

Chambers's track record of predicting (and emerging victorious from) massive transitions in communications technology is second to none, which means there's every reason to take his prognosis seriously. And if he's right, then it has incredible ramifications for our entire industry.

Chambers is predicting that, as the digital economy takes shape, as many as half of today's service providers could become "irrelevant" during the next ten years -- that's CEO-speak for "go bust." (See Cisco CEO: Get Ready for New Digital World .)

And the effect on today's incumbent equipment manufacturers could be equally dramatic. Each of these companies has genius-level, visionary executives within their ranks, people such as Basil Alwan at Alcatel-Lucent and Ulf Ewaldsson at Ericsson. Such individuals are more than capable of creating strategies that embrace the digital, open standards revolution. But just having some high IQ mavens on board is not enough, and it remains to be seen whether the organizations beneath these folk have the cultural qualities required to make the shift to a white box universe. (See The Lowdown on Service Provider SDN.)

Unless they are careful, and smart, and act aggressively both internally and externally, the mega players that have dominated networking for the past 30 years could end up in the same place as "mainframe IBM" at the end of the 1980s -- purveyors of powerful and proprietary products which have lost their relevance.

Of course, should that happen, the solution would be the same for them as it was for IBM -- transition into a world class services organization, something that, reading between the lines, seems very much part of Chambers's plan to prevent Cisco falling into the "Where are they now?" file of networking history.

So who stands to benefit from the white box revolution? (Other than service provider and enterprise purchasers of white box networks themselves, obviously, who in theory should see their equipment costs tumble as they start to source low cost hardware and software, rather than much more pricey and proprietary solutions).

Well, Chambers's prediction is good news for the likes of Broadcom Corp. (Nasdaq: BRCM), Cavium Inc. (Nasdaq: CAVM), Intel Corp. (Nasdaq: INTC) and Mellanox Technologies Ltd. (Nasdaq: MLNX), all of which make silicon chip-based switches that sit within white box solutions. It should also benefit ARM Ltd. and Intel (again), which sell processors that run the code inside white box products.

Users of these solutions should be able to download open source, standard network operating software from the Internet, but they also have the option to buy fancier software systems from specialist outfits such as Cumulus Networks or Pica8 Inc. , which bolt on extra features, such as integration with analytics. Good news for that type of company too, then.

UBiqube Plc , which makes management software that allows service providers to configure large, multivendor, virtual networks, will also be happy with Chambers's viewpoint. And in the equipment world, Brocade Communications Systems Inc. (Nasdaq: BRCD), which is "all in" on open standards such as SDN, also stands to benefit.

Who else? It was impossible to listen to the keynotes from Google's principal architect, Bikash Koley, at last year's Big Telecom Event (BTE) or this year's New IP event without being convinced that Google will be a leading white box protagonist.

And Asia has a slew of companies that demonstrate the right techno-cultural attributes to have a chance of success succeed in the white box world, including NEC Corp. (Tokyo: 6701), Samsung Corp. and perhaps even Sony Corp. (NYSE: SNE).

But the clearest beneficiaries from the white box revolution will probably be the mega system integrators -- the likes of Accenture , Capgemini , HP Inc. (NYSE: HPQ), IBM, NEC, Tata Consultancy Services Ltd. , Tech Mahindra Ltd. and Wipro Ltd. (NYSE: WIT). (See NFV's Looming Battle: Systems Integration.)

Why? Because service providers, enterprises and others (think municipalities, utilities and vertical industries, such as healthcare and automotive) that want to take advantage of white box solutions will, in so doing, lose the tremendous level of support and business advice they currently get from their equipment partners today. Expect the system integrators to fill the "help" void, and clean up.

(Aside: At Mobile World Congress in Barcelona earlier this month, Accenture gave me a demonstration of an eye-popping tool called the Accenture Intelligent Migration Solution, a perfect platform to help service provider executives model and evaluate the business case for a migration to white box networking. It also works for virtualization, and IoT applications. Expect to hear more about this remarkable product in the near future).

There are a few reasons why I personally find the concept of white box networks appealing. For one thing, once the network becomes commoditized, it forces the emphasis onto the things which are really important to customers -- the services and applications that run over the networks, and customer service they get from their service provider. To be successful in this new world, service providers are going to have to significantly up their game -- and that's good for the entire customer community.

Another thing I like about the white box trend is that it feels like it just arrived from a science-fiction movie. It's all very 2001: A Space Odyssey, with more than a hint of Space 1999 for flavor. This impression, incidentally, was reinforced by one of the Samsung booths at Mobile World Congress; essentially a gigantic glowing white box (no doubt with its own gravity well).

I mean, who doesn't want a sci-fi-style white box network from the future?

The 64-thousand-gig question is: How long before that future becomes a reality? Listening to the media and analyst buzz around our industry you could be forgiven for thinking that open, virtualized networks are ubiquitous today. The truth is completely different, with most instances of virtualization living in labs, and less than 1% of 1% of 1% of live service provider networks making use of virtualization.

Technology standards present a significant hurdle that must be resolved before heterogeneous white box networks can thrive. Currently, there's an almost farcical number of de facto, industry and official standards organizations, associations and lobbying groups working on standards for virtualization -- an interoperability train wreck waiting to happen.

Standards miasma notwithstanding, the obvious benefits of white box solutions, combined with my conversation with John Chambers, have me thinking that things really will be "all white in the end."

For more on white box trends visit Light Reading's sister site, The New IP.

— Stephen Saunders, Founder and CEO, Light Reading

COMMENTS Add Comment
Page 1 / 2   >   >>
mhhf1ve 4/13/2015 | 6:49:21 PM
Re: Has it happened before? I'm going to guess it'll take at least a decade.. that's when Moore's Law is supposed to start to fail.. then it'll be more up to software tricks to make things perform better.  
Steve Saunders 3/19/2015 | 9:39:46 AM
Re: Has it happened before? how long, do you think? 
danielcawrey 3/18/2015 | 8:09:20 PM
Re: Has it happened before? While I certainly enjoy envisioning this type of future, I still think it is a long way off. Yes, I think we'd all like to be able to deal with more standardized white box type of equipment. But there's going to be a lot of changover for that to happen. It's a realistic proposition, but will take time to become a main component of IT networks. 
Steve Saunders 3/17/2015 | 11:37:34 PM
Re: history repeats itself i hadnt seen that
jhawk81 3/17/2015 | 7:24:08 PM
Re: history repeats itself Interesting that you would use the Space Odyssey theme given Ciena's recent advert:



brooks7 3/17/2015 | 11:19:32 AM
Re: history repeats itself You mean like OSMINE....

JUST KIDDING...but it was....


Steve Saunders 3/17/2015 | 8:51:40 AM
Re: history repeats itself thanks Victor; i agree with you on the need for a standard API
VictorRBlake 3/16/2015 | 9:11:12 PM
history repeats itself Wasn't the first IP network gear on standard compute platforms ? Think PDP-11 IP nodes (aka routers) or more commonly -- later routed ?

Specialized hardware used used to deliver a level of scale (performance both in # of ports, volume of traffic per port) reliability, and ease of maintenance.

When the L4-7 (now called network functions) emerged, there where those who played whitebox (Cisco and Arrowpoint which Cisco later acquired) -- both sets of products which folded. And there were those who went special purpose hardware Foundry (Brocade) and later A10 who took major market share and equally important took the largest scale projects and wins. Now as many of those technologies can be done in VM's we're seeing that transition for many vNF. But let's be clear -- it's one thing to load balance in front of app x or app y, but cleary there are vNFs that are going to do something like 4/6 mapping for 200Gbps in a vm (no time soon at least)!! So there's life in the hardware size of the biz -- but it will certainly be fewer and fewer customers who handle that scale.

Looking at existing vm technology, some vswitches scale up to under 1000 vids, they have quite a way to go to even match 15 year old network hardware that can do 4096x4096 VIDs no to mention hardware based TPID combos that can effectively let you have and addressable vid space of 4096x4096x65536 (4 byte tpid). Yeah, that's over 1 trillion tpid + svid + cvid combos or about 1,099,494,850 times (9 orders of magnitde) larger addressable forwarding space (at layer 2 only) than a typica vswitch today. Don't get me wrong -- I think that ground will be covered quickly. But the point is that the next leap forward is likely to be something like broadcast optical switching and routing (think Compas EOS or the like) which clearly will be able to surpass what can be doen in vms in terms of scale and throughput.

At the same time, I do think we are long overdue for standards based APIs to program these devices. Its something that's been an "ask" for almost 20 years by operators.


dwx 3/16/2015 | 4:20:42 PM
Re: Has it happened before? It kind of depends on the environment with open source.   Comcast and Apple have built their CDNs using primarily open source components, and then they have contributed the intelligent pieces they have developed back to the OS community.   Most large cloud infrastructures are built using open source components.  On the storage front almost everything people use today is open source.  

Networking is a whole different story.  There are some okay open source network implementations but for the most part they have lagged far behind traditional vendors. The most promising one, Vyatta, is now part of Brocade and not so open.  While you will see white box devices in the future, they will likely be running a network operating system from a traditional vendor.  

It's also not like the white box switches are completely open either, right now when you say "white box switch" you are talking about Broadcom (and maybe some Intel).  Broadcom doesn't open source their SDKs, there is nothing open about it.  

You also are already seeing vendors like Juniper eclipse what even the next-generation of Broadcom chipsets (Trident2+, Tomahawk) will do with their own custom silicon.  
jabailo 3/16/2015 | 1:30:55 PM
Re: Has it happened before? I was thinking also about NVidia and ATI -- the GPU manufacturers.

So yes, you could say that PCs are White Boxes, except that they also moved significant proprietary graphics technology (and audio) off to separate cards that required specific drivers to operate (as any frustrated Linux user will be well aware of).

What happens when a breakout company builds a network device that is superior in some function to the White Boxes?   Same thing.

Page 1 / 2   >   >>
Sign In