Some thoughts on Openflow
It seems that everywhere i look, be it on twitter, google+ or my regular reading list of blogs, Openflow comes up alot.
What is Openflow?
Openflow is an initiative to create a protocol and an API infrastructure to improve overall functionality of your network.
I think thats the best way i can put it. I strongly suggest you take a look at Ivan’s posts about Openflow here and here.
How will it improve your network?
Openflow works by having a centralized controller(s) that is in contact with your network routers and switches. Since the controller(s) will have information from every device, it is able to make decisions on how to get your data from point A to point B through the best possible path.
The controller will make this happen by programming the data plane in your devices. Either directly or indirectly (i am not sure which yet).
What are the requirements?
Openflow is a vendor independent effort. That means that a general consensus will need to happen in order to make the protocol work correctly on multiple vendor platforms. Believe it or not, this actually appears to be happening.
Some modification on almost all platforms will need to take place. Hardware ASICS will need to be modified to support the Openflow specification for a flow. Software will need to be written to implement the protocol. All of this takes time and money.
My stance on Openflow
Openflow is a specification that, if implemented correctly, has the potential of disrupting the way we design networks. Imagine that you have some centralized controllers (servers with software) and some “dumb” piece of hardware. All you really need in this perfect world is some physical ports to connect to. The controllers will do the rest (i know this is very simplified).
This means your can buy “ports” from any vendor that supports Openflow and use your existing controllers to manage them. The competition between network vendors will then be in the controller space and what sort of application and services they can offer.
In the real world however, I feel that Openflow will take a lot of commitment from multiple vendors which dont play well together. I understand that some vendors are very keen on going Openflow, as it will create a totally new playing field for them to enter. Big vendors like Juniper and Cisco will be more reluctant to join the party, since they have a very large investment in current network designs. Trill and Qfabric have already been presented as the Next-Generation-Network for these companies.
Publicly, both Juniper and Cisco support the Openflow initiative. They kind of have to, in order to be considered to “playing nice” with the industry.
So currently, im hopeful, but skeptical. Only time will tell.
To get a better understanding of Openflow, you should visit two different sites, each one in the opposite site of the opinion scale. Greg at Etherealmind is very possitive about Openflow while Ivan is more skeptical.
Go check out what they have to say: