Facebook is leading a charge to displace traditional proprietary networking hardware and software in all of its data centers – potentially threatening the livelihood of large incumbents such as Cisco, Juniper, and Brocade in the wider market. The Open Compute Project Networking scheme was announced by the social network at …
"They do it by creating a real or perceived lock-in to their technology. [....]. They do that because more often than not they have shareholders they are beholden to"
A comment /from/ Facebook, about others. Hmmm....
OSI model FAIL
The predictable maintainability of the OSI model is the exact reason why you don't run Facebook applications on your switches and routers. The last thing we need is to have those whiny twerps running everything at the wrong layer so nothing can be maintained properly. Oh look, my router sent me a friend request. And the firewall is posting embarrassing photos of one of the network engineers vomiting nachos and schnapps on that hooker at the bachelor party last year.
Facebook is trying to make fundamental advances in technology when what Facebook ought to be doing is what it does best: junior high school drama.
Re: OSI model FAIL
Didn't we all laugh when Google (just a search engine at the time) announced it was going to release an operating system?
Now they have custom hardware and move a shed load of the internet's traffic around the world on their systems.
For businesses to survive, they have to move to other markets. Facebook, under its current form, may not be around for ever, so its wise not to keep all your eggs in that one basket.
I assume this is built on kit they've tested and used themselves
Anyone know how well it would play with Nicira?
Might this approach give one more confidence there aren't backdoors in one's network kit?
Only if you can read thousands of lines of code for every version. Open source != secure. It is only secure when security experts, who are also infallible, can review the code and give the thumbs up. Do you know of any experts that fit that profile, I don't.
@AC, Mr Wood only asked for "more confidence", not "absolute confidence", so it shouldn't be to difficult. Today we have zero confidence, all you need is one independent security expert to look at the latest version of the open source code and say, "I didn't find any backdoors" and you have greater than zero confidence. It's a (small) improvement.
The possibilities are endless... for trouble.
I certainly see some cool application potential here, but man, think about your average corporate developer writing code (not you ninja coders reading this, the $19/hr dork they brought in because some HR drone's cousin's sister works for this slave trader that cut your company a "deal" and whose code looks like something The Daily WTF would reject as "quit trolling us; this is obviously too bad to be real" - you know, average) for the control plane of your network. What could *possibly* go wrong?
OTOH, if this stuff makes it out of the datacenter and permeates the industry to the point where we can have lots of cool features all way down to the SMB segment without Cisco's "if it compiles then ship it" policy for lower-end gear, that would be cool too.
Sue ball approaching
So Cisco et al won't protect their numerous and essential patents Not
Re: Sue ball approaching
I'm not sure that Cisco will have any relevant patents - the majority of the pure packet moving functionality relies on RFC compliance and silicon so the software layer is unlikely to provide any areas that.
Cisco-only features (i.e. HSRP) would cause issues but there are already alternatives or probably don't apply in the markets Facebook are targeting.
My usual SDN scepticism
I can see where this applies in data centre networks, but the scale where it becomes useful reduces the potential customers to a few thousand and consolidation within the large data centre space reducing that further. White boxes instead of expensive Cisco or Juniper kit will make a huge difference for an environment with 100's of racks where the top-of-rack switch may be a little cheaper but the "core" switches are massively cheaper (i.e. <US$100k vs >US$500k for a fully loaded Cisco or Juniper).
For SME's there are already a lot of cheap, effective switches out there so I don't see SDN providing anything useful other than the additional cost of a SDN management tool/licences.
So that leaves the Enterprise market - do they stick with the current management tools and a small number of vendors who can offer all their switching requirements or do they move to an SDN architecture that saves them money on the hardware and costs them more on the management/licensing side.
For the amount of press SDN gets, I'm still trying to figure out what benefit that it would provide to convince me to move from an existing network vendor to a SDN network for the next building move or (small) data centre build/re-fit.
People tend to forget what happened in the telephony world...who remember the almighty Nortel?
- SMASH the Bash bug! Red Hat, Apple scramble for patch batches
- A BENDY iPhone 6, you say? Pah, warp claims are bent out of shape: Consumer Reports
- eXpat Files 'Could we please not have naked developers running around the office BEFORE 10pm?'
- CoTW Emma Watson should SHUT UP, all this abuse is HER OWN FAULT
- Vulture at the Wheel Renault Twingo: Small, sporty(ish), safe ... and it's a BACK-ENDER