NTAC Peering Working Group

[Steven Wallace <>]

From the very beginning of the Commercial Peering Service (now TR-CPS), it was envisioned as providing better-than-commodty connectivity to commercial cloud providers. The current TR-CPS service description on the I2 web site says: “...offers a low cost path with higher performance goals than commercial alternatives…”. That’s a large part of the initial motivation for CPS. In the beginning there was no difference in the underlying infrastructure between R&E and CPS, however that’s changed substantially over the years.

Regardless of how Net+ providers are connected, I suggest it’s not correct, and potentially harmful, to depict TR-CPS as commodity connectivity, unless that’s really how the service is managed now.

ssw

On Nov 5, 2013, at 2:37 PM, David Farmer <> wrote:

On 11/5/13 12:49 , Michael H Lambert wrote:

On 18 Sep 2013, at 12:42, Linda Roos <> wrote:

Dear NTAC Routing and Peering Committee,
Attached, please find, a document on NET+ traffic on R&E routes.  This document has been reviewed by the Network Architecture, Operations and Policy Program Advisory Group (NAOPpag).  Should you have any questions or comments on the document, please let me know.  Thanks.

It's been a while and I've seen no comments, so I will do so now.  I have a couple of problems with this sentence:

"While these service providers may transport some “non-research” traffic to members over the network, reaching the providers over commodity paths or TR-CPS paths which are provisioned like commodity paths will not achieve what the community wants with NET+."

I would revise that to ..."not achieve what *some of* the community wants with *for some* NET+ *providers*."

1) Where are the data that support this claim about community expectations for network performance for Net+?

I will give you an anecdote, or an existence proof;  We (NLG) want NET+ available through R&E, we don't provide TR-CPS to all our participants and think it is important to provide access to NET+ through R&E.

But, I respect your right to think differently and not do so, as long as you respect our right to do so, hence my revisions above.

2) I view TR-CPS as being closer in performance to the R&E network than to the commodity Internet.  If there is a prevailing view that TR-CPS has become too "commodity-like", then perhaps what is needed is a strategic review of the peer selection process, followed by a review of individual peers in that context.

As I said, we don't even provide TR-CPS to all of our participants, but even if we did, I don't agree.  We are limited in the amount of TR-CPS traffic we can pull across our links from I2, this may or may not be an issue for everyone, but it is a difference.

I respect that there are limits.  If it make sense for to prefer (some) Net+ services (for some connectors) via TR/CPS over R&E then we should make that known to Internet2.  Perhaps we can get a report from the TR/CPS study group at 4PM today?

-dave

Also, I disagree that we should ever limit who we peer with for TR-CPS, it serves a completely different purpose than R&E or even NET+.  Besides getting us good access to content we need for our network users, it serves the outreach mission of our institutions, ensuring the best experience for our content to the broadest possible set of users.

Limiting who we peer would result in curtailing the usefulness of one of these two important uses of TR-CPS.

Michael

-----
Michael H Lambert, GigaPoP Coordinator         Phone: +1 412 268-4960
Pittsburgh Supercomputing Center/3ROX          FAX:   +1 412 268-5832
300 S Craig St, Pittsburgh, PA  15213 USA      

-- 
================================================
David Farmer               Email:
Office of Information Technology
University of Minnesota
2218 University Ave SE     Phone: 1-612-626-0815
Minneapolis, MN 55414-3029  Cell: 1-612-812-9952
================================================