NTAC Peering Working Group

[Darrell Newcomb <>]

On Nov 19, 2013, at 7:25 AM, Dale Finkelson wrote:

> The NTAC Peering and Routing WG is scheduled to meet on Nov 19 at 4:00PM 
> EST.
> 
> 734-615-7474 or 866-411-0013
> Code 0133542#
> 
> 3.   Connector <-> TR-CPS paths

3. Connector < -- > Exchange Point Sites  (Aggregation Interconnects)

Relative to a year ago much progress has been made in this area and it  
continues to be an area of focus for improvement.  

Please find attached an updated spreadsheet regarding the Aggregation 
Interconnects; which is that part of the network within the national 
footprint that reaches out into exchange point location.  Or stated as an 
example the functional equivalent of capacity from connector into 
Starlight/710NLSD facility for onward interconnection with partners.



Discuss today 3 additional upgrades to enhance this area of the 
infrastructure:

i) Refresh Chicago exchange point site
ii) Refresh New York exchange point site
        Initial estimates are that refreshing these two locations would:
        a) Reduce monthly recurring costs from today's baseline at those 
locations
        b) Have a net-cost of 0.01/Mbps over 3 years for the additional 
Capacity added through this change.  
        With increased capacity&usage beyond the initial baseline those 
figures would decline and together help lower the ongoing profile of costs.  
After refresh, though also possible today, the increased capabilities would 
be expected to better support having that equipment be used for multiple 
purposes beyond TR-CPS.

        Details would be similar at both sites.  To talk through Chicago for 
this example would be an upgrade of legacy/1st-generation MX960 chassis; 
these originally had been removed from the R&E IP network and would have 
headed off to a warehouse or trade-in credit but were used for TR-CPS as a 
dated, cash-friendly, but viable for the moment chassis.   An initial 100G 
link to AL2S which would augment four 10GE's which would remain for 
redundancy.  The upgrade path for additional 100GE link is quite economical.


Helpful conceptual point to keep in mind for the above additional uses and 
below virtualization:
        The valuable parts of TR-CPS is the service enabled; the 
infrastructure bits needed to enable that service are important, but have not 
been essential requirements to be dedicated hardware.  Specific hardware at 
exchange point sites happened to be a very practical instantiation within the 
environment present at one moment in time.




iii) Routing virtualization in key aggregation points
        Proposal is for some cities/locations where aggregation and synergies 
exist to establish virtual forwarding instances in the common national 
infrastructure.  Seattle, McClean Virginia, Chicago, New York 32AOA, Los 
Angeles, Houston.  With only endorsed customers attached to these points any 
of the capacity to/from this would be part of the .  This will .  Suggesting 
this is undertaken as trial in New York to reduce complexity and increase 
efficiency and to surface any pain-points.  

        Some brief history about why this doesn't exist from many years 
previously:  Due to constraints present in the operating environment during 
initial integration of the networks that became TR-CPS the layer2 transport 
all the way to exchange point sites was the only viable option to deliver 
service.  While that isn't theoretically optimal; it was a viable approach 
and most of the sub-optimal ("drag") can be managed if the transport 
infrastructure reaches a scale and quality of operational outcomes which more 
than make up for that drag.  This enjoined a period of success, but 
trajectory now and near-future look to not be gaining the offsetting 
efficiencies.   The hurdles which were previously present no longer appear to 
be present and wide range of contributors to the Internet2 infrastructure 
believe an approach like this is now viable.






Regarding deployment status of previously pending Aggregation Interconnects:
        i) Dallas's additional 10GE circuit into Houston AL2S is now complete 
(Nov 1st).  
        Thanks are due to the LEARN team in helping overcome areas of limited 
confidence in certain parts of this deployment.  As well as the numerous NOC 
contributors for executing field-work and activation in advance of the LEARN 
activities.


        ii) New York
        No changes from last month; this is the last of the summer 2013 
upgrades.  Activity in New York is continuing with the Internet2 fiber 
technicians.  I'm told the best folks are on it; this has been the case since 
early August 2013 when the environment for this was explored in-depth.  
Headway is being made in recent weeks and there is prospect for completion, 
but no visibility on committed time-frame or rough expectations setting for 
completion date (expectation setting only around next date of action).

Attachment: TR-CPS - Internet2 Aggregation interconnectsNov2013-draft.xlsx
Description: application/vnd.openxmlformats-officedocument.spreadsheetml.sheet