MKM: Cisco Biggest Loser in AT&T SDN Plans
Cisco is the biggest loser, and Ciena is among the biggest winners, as AT&T pursues the new approach to hardware and software acquisition it announced in the fall, according to a new report from MKM Partners. (See AT&T Revamps Supplier Program for SDN/NFV.)
Released just before the Dec. 31 deadline for vendor responses to AT&T Inc. (NYSE: T)'s Supplier Domain Program 2.0 Request for Information (RFI), the MKM report sees a decided shift away from purpose-built smarter hardware and toward commercial off-the-shelf boxes as part of the carrier's move to using software defined networking (SDN) and network functions virtualization (NFV) to save significant capex. (See AT&T Puts SDN/NFV in Driver's Seat.)
That's particularly bad news for Cisco Systems Inc. (Nasdaq: CSCO), which has the greatest exposure in terms of existing sales of Ethernet switches, MKM Partners notes. AT&T is moving closer to the datacenter approach taken by Google (Nasdaq: GOOG) in order to compete with those giants and their cloud services.
"We do not think AT&T will be particularly receptive to Cisco’s recently announced Application Center Infrastructure (ACI) architecture, including Nexus 9000 switches, because it still seems too complex and proprietary compared to more white box-oriented architectures," MKM writes in the report.
In general, AT&T is asking vendors whether and how they are developing SDN controllers, and how they are adapting their equipment to be controlled by SDN, MKM says. Based on those responses, the network operator will be developing more RFIs and requests for proposal in the months to come.
The research note predicts what will happen over the next five-plus years as AT&T transforms both its datacenter and wide area networks, concluding that "within our coverage universe, the long-term implications are most negative for Cisco and most positive for Ciena Corp. (NYSE: CIEN), Finisar Corp. (Nasdaq: FNSR), and F5 Networks Inc. (Nasdaq: FFIV). Outside of our direct coverage, the implications appear to be most positive for VMware Inc. (NYSE: VMW)."
Critical for companies such as Cisco and Juniper Networks Inc. (NYSE: JNPR) is whether or not AT&T moves completely away from ASIC-based hardware with some intelligence and toward the "white box" approach favored by software players such as VMWare, notes MKM. Those choices are likely to be made first for its datacenter network.
"Changes are likely to occur first in the carrier's datacenter domain, perhaps as early as 2015," MKM notes. "We believe VMware's NSX platform has a strong chance of being selected as the Data Center virtualization platform, and foresee NSX-related Controllers and standardized high volume Ethernet switches being deployed starting in 2015."
Further down the road, possibly in 2016, SDN controllers for the WAN would be deployed with the possibility of packet-optical convergence taking hold in the AT&T network, MKM states.
Of course, AT&T's decisions have the potential to resonate throughout the industry, impacting the choices made by other network operators pondering the same move to virtualization and packet-optical convergence.
Interestingly, Juniper isn't as directly impacted in a negative way as Cisco, given that it has the opportunity to pick up AT&T edge and core router business in 2014 and 2015, according to MKM. But both Juniper and Cisco may need to acquire optical assets going forward, the report notes.
— Carol Wilson, Editor-at-Large, Light Reading
Good point - the telcos have for years benefitted from getting a lot of "free" work from their suppliers and would-be suppliers through the RFI and RFP process. Many years ago, a very bright man - whose name I have unfortunately forgotten - who came to what was then either Lucent or AT&T Network Systems from Digital Equipment Corp. told me he was shocked by how much free information telecom equpiment providers gave away in their responses to RFPs and RFIs. He maintained that in the computer/IT world, clients paid for that kind of info.
I guess there's nothing stopping AT&T from choosing to gather as much information as possible before following the Google/Amazon route you describe, other than that's not what they've done in the past, and they may not (yet) have the personnel to adopt such an approach, which isn't to say they couldn't buy the talent.
I do think we have to continue to mention companies such as VMware when talking about the major spending programs of large carriers going forward.
The reality is Cisco makes Ethernet switches which are almost as cheap as what you would call a "COTS" device... It's all just Broadcom or some other NPU vendor's reference design in the end, most recently based on the Broadcom Trident II.. Google/Amazon just work directly with those chip vendors instead of going through a 3rd party, no reason ATT couldn't do the same. You just have to put in the work and stay ahead of the curve, to date the big vendors have come to market with higher speed/densities faster than the "white-label" boxes like Pica8 because they have a lot more resources to throw at it. Cisco also some good features in the 9K series apart from ACI appealing to SDN folks with regards to automation since it all just runs Linux.
It doesn't make sense to mention VMWare in the same article when talking about CapEx savings. VMWare is really really expensive compared to other open source host virtualization software like OpenStack or Xen. So I don't see NSX being a solution if they are really wanting to save any money.
The biggest cost savings for Google/Amazon is in building their own servers, which are far more costly and prevalent than the network equipment. They work with Chinese manufacturers like Foxconn directly to build hardware for them, also no reason ATT couldn't do the same. There are no Ethernet switches when you build the network into custom frames like that, apart from upstream aggregation. It works very much like a big blade chassis, except it's the size of a cabinet.
There is a lot more that needs to change in thought paradigm of Telco/MSO versus the Google/Amazon/Facebooks of the world. Those companies start looking to build from the ground up, they aren't issuing RFIs to 3rd party vendors...
If the pattern repeats itself, we'll be wondering where SDN went in a few years. But for now, I'm sticking to the belief that this time is different.
Carol
I think there are paths to some form of SDN or another, but with no single vision of SDN it's hard to tell whether these paths lead anywhere useful. They could; I believe in SDN. But I also believe that you have to create an architecture for it, and I can't find one out there yet.
Cisco has clearly signaled an intention to continue building more smarts into their hardware than COTS requires, in the belief that's a better approach -- or for the more cynical, as a way of protecting their market share.
But companies such as Google and AWS are proving, at least in their data centers, that COTS gear can work just fine, thank you very much. And they are the competition for the future, are they not?
So while the clearly defined path to SDN may not yet be there - and I personally don't know that it isn't -- the desire and the need to get there is stronger than it ever has been.
While this is only an RFI response, it likely also represents ATT's first shot across the bow to all vendors - get you capex prices low and your opex value lower and demonstrable.
VMWare gains from this event, but only inasmuch as large enterprises DC architecture teams charged with SDN/NFV eval take notice. Operators have a more vested interest in vendor redundancy and therefore standards and fully supported open source.