Defining SDN & NFV

COMMENTS

Newest First | Oldest First | Threaded View

Page 1 / 2 > >>

Eddie_ 1/24/2014 | 4:00:15 AM

Re: my favourite SDN definition wildcard22,

today's leading article from Carol is a perfect fit for our discussion, I think. In my opinion it shows a multifaceted picture of the term 'open' and it underlines it's importance for a definition of SDN. Therefore your definition is a good approach, anthough Carol's article claims much more openness, e.g. torards North, East/West.

Eddie

Reply | Post Message | MESSAGES LIST | START A BOARD

t.bogataj 1/24/2014 | 2:25:57 AM

Re: my favourite SDN definition wildcard22,

I agree with your remarks. Regarding open interfaces, SBI is still the easiest challenge, regardless of whether it is OF or any other standard (and/or well-adopted) interface. All other interfaces (north-/east-/westbound) depend on applications and use cases and will hardly get universally standardised.

Let's wait and see what ONF NBI WG comes up with... But I remain sceptical that anyone is able to find a common denominator.

T.

Reply | Post Message | MESSAGES LIST | START A BOARD

wildcard22 1/23/2014 | 4:26:18 PM

Re: my favourite SDN definition T,

This defintion is close

SDN focuses on four key features:
• Separation of the control plane from the data plane
• A centralized controller and view of the network
• Open interfaces between the devices in the control plane (controllers) and those in the data plane
• Programmability of the network by external applications

But a couple of issues:

1) Instead of "centralized" controller, we have consistently said "logically-centralized". In other words, the controllers are actually distributed, but they give the appearence of being 'one', so that applications can be written on top, as though they were being written for a centralized platform.

2) Open interfaces between control and data planes is a no-brainer. But there has never been a requirement for the interface(s) between controllers to be open (i.e standardized). This is because, achieving point number (1) above is hard, and is critically dependent on these controller<->controller interfaces. Moverover, the needs of these interfaces to achieve (1) change, given the networking segment (enterpise, DC, WAN, etc) and the mode-of-operation (proactive, reactive, hybrid). So it is too early to "standardize" these interfaces

Reply | Post Message | MESSAGES LIST | START A BOARD

Eddie_ 1/23/2014 | 11:15:18 AM

Re: SDN is open! Sterling,

from my perspective "interoperability across vendors and layers" is not strong enough. An Ecosystem controlled by a single vendor could also claim some (internal) interoperabilty. "Interoperability" (especially between layers) could even be used as an argument for proprietary solutions.

Eddie

Reply | Post Message | MESSAGES LIST | START A BOARD

sterlingperrin 1/23/2014 | 9:19:32 AM

Re: SDN is open! Eduard0,

Good point about the requirement for openness. This is a recurring theme in my operator discusisons as well. In the HR definition, the wording about "interoperability across vendors and layers" is meant to address the opennes requirement as a proprietary technology would not achieve these needs.

Sterling

Reply | Post Message | MESSAGES LIST | START A BOARD

t.bogataj 1/23/2014 | 9:01:43 AM

Re: my favourite SDN definition Eddie,

Quoting definitions of of alliances and forums inevitably leads to their specific views (see my comment below). Citing ONF means adopting its view that SDN SBI equals OF.

I suggest reading an article by Sezer et al. in the July 2013 edition of IEEE Communications Magazine. Quoting from the article:

SDN focuses on four key features:
• Separation of the control plane from the data plane
• A centralized controller and view of the network
• Open interfaces between the devices in the control plane (controllers) and those in the data plane
• Programmability of the network by external applications

The definition can hardly get any better than that. And it also addresses your comment on 'openness'.

T.

Reply | Post Message | MESSAGES LIST | START A BOARD

Eddie_ 1/23/2014 | 7:48:28 AM

my favourite SDN definition includes all aspects discussed:

https://www.opennetworking.org/sdn-resources/sdn-definition

Reply | Post Message | MESSAGES LIST | START A BOARD

t.bogataj 1/23/2014 | 2:38:01 AM

Re: Defining SDN SDN is an architectural approach that separates applications from control layer, and (a part of) control layer from the forwarding plane. SDN is much wider than particular views that some vendors or industry alliances promote. Reducing SDN to OpenFlow SBI only, and saying that the two are equal is just too narrow.

A car is a four+ wheel transportation device. Gasoline-based internal-combustion engines are just one way to implement (a part of) a car. Saying that SDN=OF is like saying that a car is a gasoline-engine based device – a definition too narrow and thus false.

Time will tell what the operators (not vendors) will recognise as SDN and what they will adopt as SDN. Maybe OF will become a de facto standard within SDN, while other SBIs will not be used for SDN. But even then OF SBI will still be just one way to implement (a part of) SDN.

T.

Reply | Post Message | MESSAGES LIST | START A BOARD

TomNolle 1/22/2014 | 1:09:48 PM

Re: Defining SDN I have to disagree, Sam, unless "the industry" means something other than the broad community of buyers. In three consecutive surveys of 277enterprises I've never found the "separation of control plane" standard as their baseline for SDN. In the latest survey the majority did not even require that OpenFlow be a part of SDN. I would agree with those who will say that this view is in part the result of vendor merchandizing but whatever the cause it is what it is, and so it's fair to try to come up with a definition of SDN (and NFV) that fits current usage.

Keats said something like "Beauty is truth; truth beauty. That is all ye know on earth and all ye need to know" and it's a really nice sentiment but not a particularly helpful way of addressing the real world.

Reply | Post Message | MESSAGES LIST | START A BOARD

Eddie_ 1/22/2014 | 11:05:23 AM

SDN is open! What I am missing in the discussion about the definition of SDN s the word 'open'. From my perspective, openness is the most important architectural aspect of SDN as well as it is it's biggest benefit. A definition of SDN should takt that aspect into account. Otherwise any proprietary approach could also sail under the flag of SDN.

Eddie Beier

Reply | Post Message | MESSAGES LIST | START A BOARD

Page 1 / 2 > >>