Software-defined networking has generated lots of noise in recent months, thanks in no small part to the fact that VMware and Cisco don't know whether it's okay to send one another Christmas cards any more. That pair, and plenty of others, are talking up software-defined networking (SDN) as a must-have technology for anyone so …
SDN is horse shit
for anyone other that those ultra large service providers and massive enterprises.
It's all hype. You can google "techopsguys sdn" and the first link will go to my 4,300 in massive detail talk as to why SDN is over rated(I confirmed my suspicions with the person who created SDN which was what inspired the article). In a nutshell networking companies are trying to make L2/L3 networking exciting (it's not). They failed with DCB/DCE/FCoE 5-6 years ago and they are failing again with SDN today.
The network is not the bottleneck, I cover that in depth, obviously far more than I could include in a comment on el reg (especially because I can't have pictures and diagrams etc etc).
Maybe some day it will be for mere mortals, but by then it will be so automated you won't even know it's there. You won't have to "prepare" for it, it will just be there. No special training, no special products, it'll be basically be transparent and make the network easier to manage.
If you are not an ultra large enterprise or service provider and you think you need SDN, more likely you did an absolutely terrible job building your network.
I think that is still at least say 5 if not 10 years out.
Re: SDN is horse shit
Methinks you do complain too much. 4300 words dissing SDN?!
I notice you dont say you cover Layer1 where out of band control has been the way it's done, for ever. So nothing new here with this SDN, you might say. Likewise Layer 2, more than likely, as that's also now the domain of carriers large and small.
It's not the network per se that's the problem, it's the ops that go with scaling and managing it
In the networking space, for Microsoft, read Cisco. Anyone else is 'dangerous', until the alternatives arrive.
Wire (layer one) physically connects networks (for values of wire that include fiber, wireless, etc.). That "link" (technical term, look it up if you are clueless) cares not a whit about what that data actually is.
It is code at either end of the link (so-called "software" to the ignorant masses) that encodes/decodes the data that your intraweb-browser-thingie sends from you or presents to you.
The physical layer needs to be in place, and functionally happy on a ones & zeros level, long before you can expect the Session Layer to (several layers up) to provide so-called "software defined networking".
The whole SDN thing is marketing twaddle. The network is defined by wire.
Re: Uh ...
"The network is defined by wire".
Except where it's wireless, of course. :-)
Re: Uh ...
Yeah...you clearly have no idea what SDN is, let alone have the ability to make an informed decision about it.
@AC "14 hours ago (whatever that means, ElReg) Was: Re: Uh ...
What part of "for values of wire that include fiber, wireless, etc." do you not understand?
@astrax (was: Re: Uh ...)
Care to inform me, exactly, where I am misinformed? Marketing twaddle is not a valid answer.
Re: @astrax (was: Uh ...)
One of the main attractions of SDN is the ability to perform sweeping topological changes at layer 2 and layer 3 from a centralised position...why do you keep referring to the 'session' layer as if that is only where SDN operates? If you have a network that has multivendor equipment, wouldn't it be nice to make a few VLAN changes without having to go to each switch independently AND know the several different CLI's to implement the changes? It's exactly the same principle as using Puppet/Chef for servers, and apart from the odd nuance case, why the hell would you configure each server separately when you can do it centrally and know the configurations will be consistent?
Also, a network is most certainly not defined by the wire alone, unless you have a crossover cable from A to B. You could easily have a finite number of network configuration permutations that exist over the same bit of wire, even in a simple topology. That's why it's called a network *stack*.
Re: @AC "14 hours ago (whatever that means, ElReg) Was: Uh ...
@Jake; "What part of "for values of wire that include fiber, wireless, etc." do you not understand?"
Fair point, it was an off-the-cuff tongue-in-cheek response I made after skimming the comments, and I probably should have read the whole thing... (*)
(*) Before intentionally ignoring it and making a smartass comment anyway ;-)
“As of January 2014, Gartner estimates there are less than 500 global production implementations worldwide."
Global and worldwide? These implementations sound impressive.
Nah...I reckon a company called "global production" has 500 implementations of SDN worldwide.
Not there yet...
I think the biggest gripe people have with SDN is the obscurity associated with it. There exists a plethora of technologies that purport to be, in some capacity at least, related to SDN. This makes the subject quite difficult to approach from an administrative perspective, as there is still no definitive standard we should adopt (or at least pushing to formalise). OK, there is OpenFlow, however the impression I get is that the big boys like Cisco and Brocade are using OF as a marketing tool, citing "OpenFlow 1.3 compatibility with this range of switches...". Three times more expensive than your average switch.
SDN is supposed to remove the differentiation between network hardware vendors. Lots of little white boxes, happily configured via a centralised SDN controller. One day we will get there, and when we do, networking will change for a lot of people.
Not all, granted, but certainly a significant number of people.
Re: Not there yet...
I think also SDN is deliberately being obfuscated. Like networking was segmented wayback into WAN/MAN/LAN so SDN needs to be similarly subdivided.
Plus it is difficult to see and sell the benefits to a traditional smaller enterprise running mainstream n-tier enterprise applications in a "server room" rather than a data centre et al.
Re: Not there yet...
I think you are right on the button. SDN for small-medium sized businesses isn't encouraging at all until network equipment manufacturers start providing some sort of SDN provisioning in their firmware, or provide suitable API's. That way, you could say "Here's some really cheap switches that'll do everything you need. Here's the program to control your entire network. Oh, and have fun!".
Eliminate any traffic heading to NSA controlled IP's.
Think of the amount of bandwidth we'll save with that simple firewall rule. Then SDN will be unnecessary.
Olivia Newton John
What I know about networking is dangerous, so can someone explain this to me: how does SDN help you if what you're doing in software doesn't match your topology? And if you have to design, maintain and change your topology, which you do, isn't what you have hardware defined networking?
From my position of sublime ignorance, it looks like for SDN to be transparent you'd have to wire everything to everything else at backplane speeds. Even I can see that that's a lot of wire.
Someone help me out here.
Re: Olivia Newton John
You are not wrong. That is a lot of wire. Look up "mesh networking". Interesting in the lab, and for educational purposes, and in small-scale RealLife scenarios ... but in reality? No.
 "wire" here means the physical layer, which may not actually use physical media for communications.
Re: Olivia Newton John
The (very) general concept here is that a SDN controller will essentially talk to the network equipment directly. As a result, the SDN controller will always know the full network topology at any given time. Example: if a link goes down in a switching topology, the usual course of action would be for STP to kick and recalculate a suitable route. Due to that recalculation, STP is considerably slower then if you already had an overview of the entire network and knew the alternative routes. In theory, a SDN controller would be able to implement that topology change immediately because it already knows what to do (or if you're a smart netmin you would have a couple of contingency plans wrapped up in your SDN implementation).
Topology changes, as you correctly asserted are necessary, but the major benefit of SDN is that you can do that from a single place. There are tonnes of other interesting stuff too, like pseudo-mac addressing for layer 2 meshing and network service provisioning (counter DDoS etc). Worth having a look at, even if it is a few years away.
Hope that helps!
@astrax (was: Re: Olivia Newton John)
This is nothing new. We were doing what you describe in the early 1980s with N.E.T.'s IDNX systems (I was a sub-#20 employee of N.E.T. ...).
It's been industry standard ever since. Don't believe me? Ask Cisco ;-)
Apparently, the marketards are starting to figure out that there is something happening underneath what they see on the screen that makes it all work. Good thing? Bad thing? You decide.
(Hint: "Cloud" it's a marketing term used by people who don't understand that it's a catch-all for the word "network", as used by the '80s and early '90s textbooks describing packet switching, which were trying to teach upper-level protocols (OSI's so-called "presentation layer", mainly), without confusing people (management & marketing, mostly) with info about the lower layers that actually transport the data. Seriously. Think about it.)
Re: Olivia Newton John
 I beg to differ, all communication uses a physical medium of some sort. Be it atoms, EM waves or quantum entanglement. What other types of media are there? Metaphysical?
- Product round-up Ten excellent FREE PC apps to brighten your Windows
- Review Tough Banana Pi: a Raspberry Pi for colour-blind diehards
- Product round-up Ten Mac freeware apps for your new Apple baby
- Analysis Pity the poor Windows developer: The tools for desktop development are in disarray
- Chromecast video on UK, Euro TVs hertz so badly it makes us judder – but Google 'won't fix'