NTAC Peering Working Group

[Darrell Newcomb <>]

On Apr 16, 2013, at 11:16 AM, Dale Finkelson wrote:
> The NTAC Peering and Routing Working Group call is scheduled for Tuesday  
> April 16, at 4:00 EDT. Here are the call details:
> 734- 615- 7474 or 866- 411- 0013
> access code 0133542#
> 
> NTAC Peering and Routing Working group
> 4/16/13, 4:00 EDT


> 4.  Discussion of Connector <-> TR-CPS paths with respect to L3 
> Architecture discussions.


To help with that discussion Jeff Bartig and David Crowe asked for an 
outlined strawman to aide in discussion.  The below (very) roughly outlines 
one possible approach.



Using as a starting point the Layer3 Architecture layout/details as outlined 
and shared with Internet2's l3-architecture group for review.

A key diagram from that effort is attached below.  In it you'll see pencilled 
in the existing links (solid lines) connecting Exchange point sites(penciled 
in circles) to the IP routers.  For this strawman the AL2S devices, their 
links, and FIBs have been described as being out of scope for use in any 
innovative way; this outlines using AL2S devices for vanilla Layer2 transport.

Over the history of CPS and now TR-CPS this area of the network (IP routers 
to exchange point sites) has been the most consistently under-provisioned.  
It continues to be a challenge that limits growth/usage at times and adds 
overhead to manage risk there regularly.  In the recent week there has been 
connectors with actual usage ~ 120Gbps and ~318Gbps of connector capacity 
over the 180Gbps of transmission capacity(IP routers to exchange point 
sites).  It's worth noting a number of connectors obtain TR-CPS service by 
connection to the exchange point routers; that usage isn't in those figures.

This strawman shows adding links between AL2S and the exchange point router 
to support growth and the AL2S-switch arrival point for connectors.
        As the l3-architecture plans have connectors establishing connections 
to AL2S in replacement of links to the IP-core routers(formerly T-series 
Junipers); the psuedo-wire/vlans connecting through to TR-CPS exchange point 
sites could then be delivered direct from AL2S to exchange point sites.  This 
would potentially avoiding the cost of an additional pair of interfaces on 
the IP-core routers, remove another point of failure, and relieve the legacy 
IP backbone from carrying that traffic for connectors.

AL2S would not be used as backbone between exchange point sites per the above 
constraints.  Additionally FWIW generally it has been most economical to use 
dedicated waves between exchange point sites to avoid interfaces costs on 
intermediate devices, retain headroom along links used for high speed 
research application, among other factors.  Though this is not part of the 
AL2S access being focused on the TR-CPS backbone links can be seen here: 
http://atlas.grnoc.iu.edu/atlas.cgi?map_name=Internet2%20TR-CPS

A breakdown of the 10G and 100G links to support this strawman initially and 
then at logical conclusion with all connectors are migrated is in the 
attached spreadsheet.

Attachment: 2013-l3-arch-withtrcpslinks.pdf
Description: Adobe PDF document

Attachment: TR-CPS - AL2S strawman.xlsx
Description: application/vnd.openxmlformats-officedocument.spreadsheetml.sheet