There are many reasons, MTU, Architecture, flexibility. Many DC's do similar 
work by doing interISID or interVlan routing and extend the ISID or VLAN under 
the IP layer for VM moves without needing to jump through hoops to make sure 
they can provide optimum routing flows in and out of the DC as well as cross 
app tier subnets. 

Overlays can be very useful, but aren't a panacea in of themselves.

--
Paul Unbehagen


Sent from my iPad

On Jul 27, 2012, at 10:27 AM, Robert Raszuk <[email protected]> wrote:

> Hi Paul,
> 
> Clearly if you have bunch of legacy applications which require L2 and are not 
> capable of crossing L3 you need to provide them the L2 network.
> 
> But beyond that I really do not see a need for VLANs. Subnet - of course - 
> but within a given tier ... more ... withing given TOR.
> 
> What would be any other reasons to be stack with L2 for DC transport ?
> 
> Best,
> R.
> 
> 
> > Not sure you can avoid it completely, nor do I know that everyone else in 
> > the industry wants to avoid it either.   As avoiding VLANs doesn't address 
> > the LANs needed to support clusters of servers acting as any single tier of 
> > the application model described below.  Web clusters, app clusters, etc...
>> 
>> It's a fine balance that is flexible today because we can use the VLANs and 
>> IP subnets appropriately depending on the need and scale.  We should be sure 
>> not to box ourselves into a one size fits all mentality. InterVLAN routing 
>> inside a VRF is very appropriate for many DC's for example and used today.
>> 
>> --
>> Paul Unbehagen
> 
_______________________________________________
nvo3 mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/nvo3

Reply via email to