If you're skeptical about the intentions of the OpenDaylight SDN consortium, you're not alone.
Several questions remain regarding the true objectives of this consortium, pulled together by Cisco and IBM and involving several of their partner and competitor vendors in an effort, on surface, to induce an application ecosystem for open source SDNs. Several established and startup vendors, including Cisco, IBM, HP, Juniper, Arista Networks and Big Switch Networks, are collaborating under the auspices of the Linux Foundation to develop an open source framework for SDN application development.
But there are several elephants in the room, of various sizes:
- The conspicuous absence of the user-driven Open Networking Foundation (ONF). Not even a proactive endorsement or statement of support, let alone participation.
- Co-founder Cisco's effort to have its Cisco ONE controller as the de facto controller for OpenDaylight
- Guarded optimism - read: skepticism - from some members, like Big Switch, that this effort stays true to its stated goals. Indeed, Big Switch's participation is as much a watchdog over the process as it is contributor.
- Cisco's downplay, and even dismissal, of the role of OpenFlow and the decoupling of control and data planes in SDNs. OpenFlow is the popular open source SDN API and protocol supported by many users - like Google and other founders of the ONF - and vendors, like Big Switch, HP and IBM.
- The likelihood that Cisco's Insieme Networks programmable networking spin-in will not include Daylight in its product development.
- And just the unlikelihood that fierce competitors can collaborate and/or agree on any strategic product development. Indeed, efforts like this in the past have ended up as nothing more than failed PR exercises and thinly veiled smokescreens for other less-than-altruistic agendas.
"We're skeptical of this effort for several reasons," says Gartner analyst Joe Skorupa. "Unlike the ONF, this effort is controlled by large vendors and has cut out the voice of the consumer. In private conversations with a number of vendors, they have expressed the same concerns. They may be members, but clearly aren't supportive."
So why should anyone view this as anything more than Cisco and IBM attempting to freeze the SDN market and potential momentum as they get their own unique products ready in the background?
Good question, according to Jason Matlof, vice president of marketing for Big Switch.
"We are very cautious and even skeptical that this body will not be politicised by some vendors' behavior," Matlof says. "We are Platinum members because we believe that the spirit of Daylight is the right thing for the industry and users, to the extent that it remains committed to standards, including OpenFlow; it delivers on the explicit expectation that the community be 'Merit-Based'; and it is open to all contributions."
Matlof and Big Switch are concerned about Cisco's recent remarks about OpenFlow, which Big Switch believes undermines the capability of OpenFlow in SDNs. The company is also concerned that Cisco will use OpenDaylight to push its own Cisco ONE controller as the de facto controller for OpenDaylight.
That concern is legitimate, according to Skorupa.
"While the effort is supposed to be governed by a 'technical meritocracy' it looks to be 'pay to play' from a governance perspective," Skorupa says, referring to the tiered membership fees for participation in OpenDaylight. "Furthermore, the selection of a foundation controller that hasn't been proven from a vendor with a mixed record on new software and that has been openly critical of OpenFlow seems suspect."
Cisco is "not ready to coronate" its Cisco ONE controller as the de facto controller for OpenDaylight - but Cisco did submit it for the project and it's in the hands of 50 customers who are writing applications for it, says Dave Ward, Cisco vice president of engineering, CTO and chief architect. So that, and that Cisco co-founded OpenDaylight with IBM, might make Cisco ONE's chances pretty good.
"There's no momentum around one controller of all the open source controllers out there" like Floodlight, Nox, Trema and others, Ward says. "There's no reference architecture."
But it's not about the controller, Ward asserts - it's about the applications.
Without OpenDaylight "there's a chance in three to five years we're all chasing a controller," Ward says. "But all revenue comes with the applications on top. All the action is in the applications on top of a common controller."
That may be why Big Switch is so cautious about OpenDaylight: It may disrupt its business plan. Big Switch charges about $1,700 per month for its Big Switch Controller and $4,200 per month for its Big Virtual Switch.
"Most of the vendors are saying, 'Let's get together to create an OpenFlow controller,'" says Peter Christy, Networks research director at 451 Research. "Vendors don't plan to differentiate on it. They agree that they don't want to compete at some lower layer. They're doing an OpenFlow controller like they collaborate on OpenStack."
Christy says this effort could either undercut or accelerate Big Switch's business plan, but Big Switch says that's not what worries it.
"What we are contributing is overwhelmingly already free through Project Floodlight, and our business model is already largely open source," a Big Switch spokesperson states in an email to Network World. We derive our revenue from subscription and services around the Open SDN Suite. The [business] model is analogous to Red Hat's enterprise edition.
"We are concerned that Cisco will politicise the selection process ... and that they will begin deprecating the code to move it more toward the proprietary model that serves them and no one else, most especially customers."
That may be the ONF's concern as well. The ONF's participation, endorsement and support was missing from the OpenDaylight press release and related material, but the OpenFlow-focused organisation, founded by Google, Facebook, Yahoo, Verizon and Deutsche Telekom, issued a statement upon request that it is watching Daylight and other initiatives to see how they "measure up" to openness.
"ONF welcomes all who support our mission of furthering the commercialisation of open, standards-based SDN, and we are pleased to hear that the OpenDaylight Project will focus on the OpenFlow protocol standard - the first and only vendor-neutral standard communications interface between the control and forwarding layers of an SDN architecture - that ONF provides and continues to evolve," states ONF Executive Director Dan Pitt.
"We encourage our member companies to participate in industry initiatives - including open-source ones - designed to foster innovation, implementation, and deployment. These member companies are taking leading roles in the OpenDaylight Project. As a voice of the user community, ONF supports those initiatives that are true to our guiding principles by being based on multi-vendor standards and open to broad, merit-based, multi-vendor input. We are eager to see how this and other initiatives measure up to these principles and meet the needs of users."
The OpenDaylight work will include an attempt to standardise a northbound API from the network infrastructure to the SDN orchestration layer to aid in the programmability of the network. OpenFlow is one of the southbound network control protocols both OpenDaylight and ONF will support, but ONF passed on working on a northbound API and has no open source resources with which to attempt it, Ward says.
"There's no standards body working on northbound APIs," he says. "So we're creating the de facto standard. Let's use open source to get there."
Some analysts believe OpenDaylight will be more targeted to enterprises than the ONF, which is founded by large Web-scale companies and service providers.
"The vendor community realises that the Googles and Facebooks of the world will continue their pursuit of SDN-style network virtualisation and network programmability," says Brad Casemore of IDC. "That said, the enterprise remains up for grabs. The ONF speaks articulately and cogently for the interests of web-scale data centers, but it hasn't meaningfully espoused the cause of the typical enterprise, or the enterprise data center.
"In a certain sense, you could view the formation of OpenDaylight as a vendor-led attempt to reframe the discussion of SDN," Casemore adds. "It's fair to wonder, then, about whether OpenDaylight will be, as its proponents contend, a complement to ONF. Rather than a complement, it could actually develop into something more akin to an alternative to ONF."
And while this OpenDaylight work proceeds, Cisco and the other vendors in the consortium will continue to work on their own unique SDN architectures apart from the OpenDaylight project. Cisco's is Cisco ONE and the onePK API set for the company's three key operating systems: IOS, IOS XR and NX-OS.
Cisco clearly believes customers should go the onePK route, if a recent meeting with networking bloggers is any indication.
"OpenFlow is a fantasy you cannot counter because it's promising everything," says Prashant Gandhi, senior director of Cisco's Service Access Virtualization Technology Group, during the meeting with the Networking Field Day bloggers. "And there's nothing in production."
OnePK itself is only in the proof-of-concept phase, Gandhi acknowledged. Cisco is adopting a hybrid implementation of OpenFlow where its decoupled control and data planes work in unison with the control plane of IP. OpenFlow is also a new protocol with unanswered questions around security, survivability and scalability, Gandhi notes, to which a blogger countered that onePK has yet be proven bulletproof against buggy code, attack vulnerabilities, scale, etc.
So with so much suspicion of OpenFlow within Cisco, is it likely that Cisco will aggressively market the OpenDaylight controller vs. onePK? Is it likely that a strategic SDN investment like that made in Insieme Networks will include OpenDaylight in its product development?
Cisco would not comment on whether Insieme plans to embrace OpenDaylight in its product development. Analyst Christy believes Insieme will provide an interface to it but not base any product capabilities on it.
And then there are the vendors participants themselves, which represent some of the fiercest rivals in enterprise and data centre networking. It's unlikely that they would collaborate strategically on something as strategic as SDNs.
It didn't happen with the Network Interoperability Alliance, led by IBM, Bay Networks and 3Com; Microsoft's Web-based Enterprise Management -- WBEM -- effort was viewed as a thinly veiled cabal against Sun's Java. It didn't happen with the Open Software Foundation's distributed computing and management environments, and it didn't happen with the OSI protocols.
Why would OpenDaylight be any different?
"One should not conclude that perfect amity will prevail among the membership," says IDC's Casemore. "Common membership does not denote uniform agreement. Each vendor has its own reasons for joining OpenDaylight, and, quite often, those reasons differ. Even Cisco and IBM have different agendas. So, yes, cooperation, except on the most fundamental code contributions, will be difficult to achieve."
Which leads one to ask, just how strategic is OpenDaylight?
"OpenDaylight will not guide vendor strategies," says Casemore. "More likely, it will serve as a foil and tactical bulwark for individual vendor strategies."