[AusNOG] Narelle with the first ever web server

[Matthew Moyle-Croft]

Daniel,
I don't doubt you can do other things with it, but the driving force is to allow "better" traffic engineering than current mechanisms.  Coupling OpenFlow with BGP or other information is about doing better internal TE.  Also, people like Google, Amazon, etc have large internal flows of traffic which don't conform to what things like MPLS AutoBandwidth do and want to build better ways of handling that.  

OpenFlow seems to be a normal part of the cycle between centralised and non-centralised systems that IT&T goes through.  


MMC

On 26/04/2012, at 8:54 AM, Daniel Griggs wrote:

> No I don't believe that is what OpenFlow is trying to solve at all.
> That is what Google have demonstrated it solving though for them though.
> 
> The point of open flow is to separate the control plane from the
> forwarding hardware. Allowing network operators to decide how much
> control they want over routing decisions in their network. If they
> want to put money into developing their own control plane and RIB to
> FIB engine they can, or they can just buy more Cisco/Juniper/Alcatel
> equipment. Perhaps one way to look at it is that it compliments the
> open source routing engines (Quagga/Bird) with forwarding hardware.
> 
> There is no reason why you can't have an Openflow component that
> speaks BGP to neighbours and uses a combination of what of learns from
> BGP and local policy to program the forwarding plane to route packets
> as required. Openflow just defines the protocol for communicating
> between the control plane and forwarding plane hardware, it isn't
> designed to replace BGP or network management tools.
> 
> On 24 April 2012 18:10, Matthew Moyle-Croft <mmc at mmc.com.au> wrote:
>> I don't think OpenFlow is trying to solve NNI issues - just internal TE
>> optimisation.
>> 
>> MMC
>> 
>> On 24/04/2012, at 3:38 PM, Narelle wrote:
>> 
>> 
>> 
>> On Tue, Apr 24, 2012 at 1:44 PM, Simon Knight <simon.knight at gmail.com>
>> wrote:
>>> 
>>>> [Openflow is a long way from being deployable for Internet scale stuff,
>>>> imho: no network management concepts at all!! Applications demanding
>>>> network
>>>> paths is not new, but there needs to be standards. I've got no issue
>>>> with
>>>> losing IP either, but they want to keep addressing so you'll still need
>>>> to
>>>> learn IPv6 folks. He's keen to develop it more and accepts all these
>>>> things.]
>>>> 
>>> 
>>> Off-topic, but OpenFlow isn't far away:
>>> 
>>> http://www.wired.com/wiredenterprise/2012/04/going-with-the-flow-google/all/1
>> 
>> 
>> That's another thing he was trying to convince me: that it is in large scale
>> data centres and private backbones but nowhere near deployable across
>> boarders. No BGP like function...
>> 
>> 
>> 
>> --
>> 
>> 
>> Narelle
>> narellec at gmail.com
>> _______________________________________________
>> AusNOG mailing list
>> AusNOG at lists.ausnog.net
>> http://lists.ausnog.net/mailman/listinfo/ausnog
>> 
>> 
>> 
>> _______________________________________________
>> AusNOG mailing list
>> AusNOG at lists.ausnog.net
>> http://lists.ausnog.net/mailman/listinfo/ausnog
>> 
> 
> 
> 
> -- 
> Daniel Griggs
> Network Operations
> e: daniel at fx.net.nz
> d: +64 4 4989567