Re: [irs-discuss] IRS Problem Statement Posted

[James Kempf <[e-mail]>]

 
> [ off-topic ]
> 
> Today in IDR, Microsoft folks presented on running data 
> centers using BGP. So I guess we should put OSPF & ISIS to 
> rest. We can use BGP for everything.
> 

Yeah, right. Then when it breaks due to being used for something it wasn't designed, we can spend the next 10 years in IETF working groups coming up with fixes. And have to hire high priced BFG gurus to configure and keep it running in the meantime. :-) 

> [ / off-topic ]
> 
> Openflow abstractions are at the forwarding level. IRS does 
> not compete with that.
> 
> [ rant ]
> Did I just hear that you've convinced your company to replace 
> all of its networking gear with OF gear and a uber powerful 
> controller.
> [ / rant ]

Wollen Sie dann weiter die Unterhaltung auf Deutsch haben?
> 

Anyway, I've been working with OpenFlow on and off since 2006. You can do stuff like:

- make an access/aggregation network with a couple hundred switches look like a gigantic router to the outside world by aggregating the interfaces (and thereby make it easier to manage)

- do application specific flow steering for chained processing in middleboxes (complaints about DPI please to /dev/null)

- run a 3GPP/LTE mobile core with the control plane including routing completely in a cloud instead of only the mobility mangement part

- and finally, as Google has demonstrated, run your network at 90% capacity as opposed to 30% (and, yes, I know that the network was rather special and perhaps not characteristic of a provider network but it still shows what you can do)


But it is not magic, there are lots of problems that need work, and it could still go badly wrong. Right now, it is at the peak of the Gardner hype curve, give it another year, and everybody will be trashing it, unless of course the hype curve moves up another notch.

			jak

> Thanks
> Nitin
> 
> On Jul 30, 2012, at 6:48 PM, "Robert Raszuk" 
> <[e-mail]> wrote:
> 
> > hi Nitin,
> > 
> > 
> >> So IRS needs to define interfaces for both the lower layer 
> and the upper layer.
> > 
> > That's precisely what openflow protocol does. Thx for 
> confirming clearly the intentions ;).
> > 
> > Best regards,
> > R.
>