The hybrid cloud is becoming a notable thing. Companies such as Microsoft are pushing visions like CloudOS while work patterns change to allow more flexibility in worker location. Traditional networks focused on situations where workers perform all duties on premises owned by the company. How will that change? I sat down with …
Too much to write here, wrote on this topic a couple months ago. I know el reg doesn't like links to other sites but I can't duplicate the writing here with images etc..
basically sdn is a crock of shit. I see it useful for the hyper scale players out there, but the number of organizations where SDN will be really useful is quite limited. Network vendors haven't gotten the attention that the storage and server virtualization folks have gotten over the years. Networking is viewed as boring, it's basically a utility -- how often do you see people getting excited about a new UPS or power strip..? There have been some interesting things happening at layer 7 over the past decade but that's layer 7.
So networking companies have over hyped this SDN concept to hell and back to try to make things sound exciting again. A few years ago it was FCoE (and DCB/DCE) -- obviously that flopped!
Now it's SDN..
When in reality -- the people that really need this stuff (and I admit there are customers that do) -- already know about it. In a lot of cases I bet (Google, Amazon etc) already have a been doing a sort of SDN long before it was formalized as a term, because they got to that scale where they had to do something.
But in the link above I address more directly the flaws I find in the hype of SDN in general and how it's really just that - hype.
I did not feel comfortable attacking SDN dead on before recently because until recently I could not get an informed opinion of WHAT THE HELL SDN IS. I had the opportunity to ask the question to the inventor of SDN himself personally, which confirmed all my original thoughts and expectations and allowed me to write a good blog post on the topic.
I feel sorry for the networking companies I really do. I do wish they would spend more time on ease of use rather than adding ever more complexity. I was quite disappointed when I learned a few years back that TRILL was a layer 2 only protocol (I fully expected it to have a layer 3 component, because you know, we do this thing called ROUTING now and need redundant routers). I've really liked the protocol ESRP from Extreme which combines layer 2+3 (you can also do layer 2 OR layer 3). And Extreme in general has stuff that is easy to use, with a config language that more or less reads like english (configure vlan my_vlan add ports 1,2,3 etc).
But I suppose day to day ease of use is not flashy, not a catchy thing because you can just hire yourself an expensive network engineer and/or enroll in some complex training classes to learn how to use the equipment.
Look how easy (for the most part) modern storage is today(certainly are exceptions), or modern server virtualization. Most any idiot can fire up a vmware system and build a cluster and create vms with a few mouse clicks. sometimes that causes problems, but it's still easy to use.
Now some things in networking can be complex still, take routing protocols and stuff like that, that's fine. But basic L2 and L3 stuff should be dead simple easy to understand and build.
Perhaps in 3-5 years, maybe more, SDN will be integrated to the point where it's in a similar state. I don't know. If that's the case it's sad it took the industry 10-15+ years to get to a state of ease of use.
I have no doubt a decent part of the problem is the network admins themselves who pride themselves with their advanced certification and training courses on the overly complicated network equipment and they consistently over design things and just make life in general more complicated.
linking to other sites is fine, so long as it is not pimping...
Nate, your comments rock. We are always looking for independent IT practitioners who can write. Fancy joining our roster of sysadmin bloggers? We spell check and vet for libel, but that's all!
Re: my comments
I feel the need to debate you somewhat here Nate.
Point for your argument: 95% of my deployments don't even use VLANs, let alone anything more complicated. (Though Trunking and 802.3ad/LACP see widespread use.)
Point against your argument: if the "advanced features" were easier to use, at least half those same clients would be on them like white on rice.
The issue - at least at the SMB end of things - isn't that SDN-like features wouldn't make lives easier, reduce OpEx costs and so forth...it's that these companies don't have "network administrators." CCIE cost muchos dineros. Even if you have the money, you have to deal with the egos...and most SMB owners I know of just don't have time for the sorts of Prima Donnas that CCXX seems to attract.
But the do use virtualisation. They are leaping headfirst into storage virtualisation. They'd dearly love to have all the promised functionality of SDN, but with a nice UI and none of those nasty attendant network admins.
Some network vendors claim they have a solution that can meet these needs. Some go on about vendors but refuse point blank to discuss ease of use with me.
Nobody talks cost, not in hard numbers, not ever.
So to an extent you're right: what's on the table today just flat out doesn't apply to a lot of companies. Where you're wrong is that this isn't because the features aren't in demand...it's that those features have to bring simplicity with them in the form of ease of use and the ability to jettison the network admin from the payroll.
That day will come. 5 years, 10....15? Who knows! But virtualisation did away with a lot of dedicated application and hardware cluster admins. It made backups and disaster recovery easier and collapsed those specialties into generic admins in all but the largest organisations. I am seeing the same thing happening to Storage today; Tintri goes in and a storage admin goes out. (Hell, Nutanix goes in and they start culling storage and virtualisation admins, but that's another story for another day...)
Somewhere in the past 10 years vendors of all sizes and in all areas of IT forgot about ease of use. Ease of use isn't sexy. Everyone at every size scripts, right? Everyone can remember every single powershell command for every single application they use, right? Everyone knows ios by heart, right?
What do you mean, you can't afford 15 dedicated admins for each area? What kind of Mickey Mouse company are you?!?
It's interesting to see you pooh-poohing SDN because the fabric portion of the exercise is inherently a layer 2 activity. As far as I'm concerned that's a good thing. Routing is inherently north-south. It's a bottleneck and SMBs like me and mine sure as hell can't afford routers that fling around multiple 10 gigabit links. We can't keep going up the aggregation stack to the top in order to go out to the edge.
I don't even understand why I should ever have to worry about that stuff. Why the hell can't I just connect switch A up to switch B and have the things figure out how to make the bandwidth work? I care about the workloads that run on top of the network, not getting into the thing and writing a script to make it go.
Routing should be something that connects the heavy lifting to the users. I shouldn't need expensive bottlenecks to connect one big-ass high bandwidth device to another so they can play nicely. I shouldn't need expensive equipment or CCwhatevers just to make the damned switches work.
This is where SDN comes in, even in the smallest of businesses. Someone please explain to me why home routers, wifi devices and switches are even capable of layer-2 broadcast loops? We've had spanning tree (and alternatives) for well over a decade, but grandma still has to worry about how many cables are plugged where?
Accounting still can't just plug another cable between switch A and switch B and they'll "just go faster"? Why is this shit still an issue?
Auto MDIx was something we could all agree on, and I haven't needed a crossover cable in at least 6 years. Innovation seems to have stopped there. Protectionism and douchbaggery have completely stalled any advancement in networking and they hold everything else back.
Openflow – or more specifically OpenDaylight – looks like it is going to be the only way out of the morass of asshattedness we find ourselves in on this.
Who wants SDN? In my experience damned near everyone. What they don't want is the protectionist charlatanery that seems wrapped up in most attempts to sell it to the hoi polloi.
Anyways, that's my $0.02. Also: listen to Drew. More sysadmin bloggers are a good thing. The world needs more than my voice (gods know that's true!) and you're a bright chap. Join in and share your wisdom with the crowds. We have cookies.
P.S. One of us. One of us. One of us…
What Network Virtualization is?
A software emulator running on networked hardware emulating a virtual network, running all your virtual machines in the virtual cloud on a virtual ...
Re: What Network Virtualization is?
Abstraction layers provide easy of use at the cost of efficiency. When the cost of what's lost to inefficiency of the abstraction layers drops below the cost labour cost required to run things without the abstraction layers then the abstraction layers see widespread adoption.
That's business 101-class stuff...
- Product round-up Six of the best gaming keyboard and mouse combos
- China building SUPERSONIC SUBMARINE that travels in a BUBBLE
- Boffins attempt to prove the UNIVERSE IS JUST A HOLOGRAM
- Review Raspberry Pi B+: PHWOAR, get a load of those pins
- Linux turns 23 and Linus Torvalds celebrates as only he can