OpenDaylight Project Rumors & SDNCentral Analysis: Five Questions from Network Operators
As the networking community gears up for the 3rd annual Open Networking Summit April 15th – 17th. Details and rumors continue to emerge about the OpenDaylight Project, which we expect to be announced any day now.
What’s happening with OpenDaylight Project is top of mind for our cloud service provider clients – many of them are working with us to finalize their initial multi-tenant network virtualization offerings to release in Q4 ’13 and Q1 ’14. We are helping these network operators define their technical and business requirements, identify and evaluate viable technologies and business models, and prototype and customize various network virtualization approaches to work in their unique environments. One attribute that’s common for all of our clients is that they are all aiming to make technology and vendor selection in the Q2 – Q3 2013 timeframe.
Our clients express a fear of selecting a solution that could potentially a) lock them into a single vendor (either proprietary or open-source) ; b) end up a dead end technology (say a non-mainstream standard or product); or c) requires too many resources (people, process changes, or CPUs) for deployment at scale. These clients are also smart enough realize when vendors proclaims ‘Open’ for no other reason than to shorten sales cycles or eliminate code escrows and are rightfully skeptical of open-source projects and initiatives (See OpenSource: Biggest Risk to SDN) .
Here’s the five most common questions we receive about OpenDaylight – and our answers and analysis based on the details and rumors we’ve heard:
1. What organization will host OpenDaylight?
We’ve been told my multiple sources that OpenDaylight will be hosted at the Linux Foundation. Based on our review of the Linux Foundation website – we’d expect to see OpenDaylight to be a collaborative project (just like MeeGo from Intel and Nokia) Assuming this information is accurate – our advice to customers is the Linux Foundation is a credible organization and approach to manage open source – everyone uses and trusts Linux. If this is true, then OpenDaylight will be run out of a viable organization that is independent and not under control of any one vendor.
2. Who are the key members?
We’ve heard similar reports to SearchSDN’s report that OpenDaylight will cost the highest level members between $250,000 – $500,000 and 10+ engineers annually. For argument’s sake – if OpenDaylight follows the sponsorship levels at Linux Foundation then, we can assume there will be Platinum, Gold, and Silver sponsorship levels plus Affiliates for key non-vendor participants. Based on this taxonomy and various rumors SDNCentral projects the initial OpenDaylight roster to include:
Some combo of Platinum and Gold Level Members (mostly carry overs from the rumors we heard last last month): Arista, Big Switch, Brocade, Cisco, Citrix, HP, IBM, NEC, and VMware. We’ve heard conflicting rumors about Dell, Intel, Juniper, and Microsoft joining as founding members. Initially, the shocker for us was hearing market rumblings about VMware’s involvement – though given that VMware owns / controls Open vSwitch , via Nicira, leads much of the development of Quantum for OpenStack; and is focused on closing cloud service providers as customers — if true — it makes sense for them to join the party.
Silver Level Members and Affiliates: This feels like the catch all for the masses once the project is launched. (i.e. logically – would you pitch your CEO that is was important to join as a founder, but only contribute at ‘silver’) We expect enrollment for silver partners to come some time after the initial announcement.
3. What code do we expect from OpenDaylight?
Based on our work with cloud service providers and leading edge enterprises – we define the pieces that constitute a functional network virtualization solution as the following:
– SDN controller
– Virtual multi-tenant network slicer
– Network Overlay
– Plug-ins for virtualization environments such as OpenStack and VMware
If we take the initial reports in our original OpenDaylight post plus add in potential VMware contributions – we make an educated guess that the following bits may be included in OpenDaylight:
Likely Code Origins
|Virtual Multi-tenant Network Slicer||HP & NEC||Programmable Flow|
|Network Overlays||Cisco / IBM
|vCider / DOVE / Nicira|
|Virtualization Plug-ins||IBM and VMware||Quantum (OpenStack)DOVE (VXLAN)|
Again – this is our own assessment based on experience, rumor, and conjecture – we don’t recommend you make purchase decisions based on this chart – as the real solution will emerge shortly.
The quickest way to offer these disparate technologies as one bundle would be for them to operate a separate projects under OpenDaylight – much like other projects within the Linux Foundation.
4. What is the Open-Source License?
We’ve heard multiple rumors: a) Apache 2; and b) Eclipse. Our experience suggests that customers are well protected by either license — as they both meet FOSS requirements. Not being attorney’s, we advise our client to speak with their legal departments, though in our experiences we haven’t seen issues with either license.
5. What will code be available?
We have no clue as to when to expect OpenDaylight to produce code (we have no insider knowledge and our sources go mum on this topic), but this is where experience give us free opportunity to make a prediction of: initial code in 1H 2013.
Normally, we are skeptical of any project that claims code in less than 12 months. In OpenDaylight’s case, it is starting from projects with existing code. Based on our hands on experience, one could build a functional demo out of our hypothetical pre-Daylight technologies listed above. Production level code is separate discussion.
To build a demo – you need to download, install and integrate:
– OpenStack as your virtualization environment
– Beacon as an SDN Controller
– FlowVisor for network slicing and multi-tenancy
– Open vSwitch for the vSwitch
– Quantum for OpenStack networking
Granted this is work – though a few, experienced network engineers could build and document a demo in three to six weeks. In theory, once you had this demo, one could then to extend to start testing specific vendors products by adding a vendor specific Quantum plugins – such from Arista, IBM, Cisco, or VMware, etc. Remember, this is a demo that would be no where ready for production.
Now translating this into production grade capabilities that an organization can run their business off of is a completely different thing. Though I expect we’ll see a number of Cloud Service Providers looking to take some of these components to build custom testbeds and then look to see which OpenDaylight vendor can scale this setup the quickest.
While there are clear vendor benefits from OpenDaylight, — there is still one question from our clients that we are unable to answer: – if we can replicate a significant portion of Daylight’s purported value with existing open-source and some professional services, then what is the customer value of Daylight?