Chris started the call with an intro and some background:
He started by saying that "We" in the context of his remarks referred to the entirety of the community, not just the staff at Internet2.
For the background his main points were:
- The RFP was a snapshot of ideas at a certain point in time.
- It was representative of a static point in a very fluid and on-going discussion.
- Overall the goal of the RFP/rethink of the network is to make clear to everyone that I2's network is not another commodity network. It is not another Cogent or AT&T, it needs to be differentiated from those.
- The current funding opportunities give us a chance to be transformative once again.
- The goal is to provide the community with a framework to allow them to be transformative once again as well.
Grover then provided more about his perspective:
- He was asked who he represented in this discussion. He responded that he was at IU but working on I2 activities.
- He sees this activity as creating Building Blocks. A toolbox out of which many services can be created. The hardware that goes in place should be agnostic. What matters is that we can get hardware that is production quality with maximum availability.
- The idea is that a new service would only be introduced once it has been established that it will neither harm nor impact any other production services that are already in use. There needs to be protocols developed for testing/vetting new services. Then they will only be merged into the actual network when they are proven safe.
- The two main points he was making were: 1) The boxes that are put in place are intended to support a "toolbox" view of the network. 2) The goal is to try to increase the overall availability as much as we can. The network needs to move forward and manage risk and maintain availability all the the same time.
Bill Owens asked: Does the current proposed solution really have the characteristics that Grover discussed?
Akbar pointed out that the traditional model was a fiber infrastructure with boxes providing IP services on top of that. The new model seems a departure from that to a fiber infrastructure with SDN enabled boxes delivering IP services.
- Grover responded that this was not what is planned. We will deploy something that works just like what we have but has some SDN like capabilities. Those capabilities are vetting via a well defined process. Perhaps they are then run in parallel with the traditional approach. But they are only put in place when they are proven ready.
Akbar: It seems that ultimately we will make this change. How do we get from here to there?
-Grover: We set up a process where people can do new stuff in a way we can understand what they are wanting to do, that gets tested and vetted, once proven it can be put on the live network backbone.
Cort: Something seems confusing here. I2 has for a while now put a lot of effort into developing "products" with a name (DCN, ION etc). Really what I2 should do is look overall at a broader scope. The message should be something like: 1) I2 does circuits etc (layer 2). 2) These things can meet peoples needs. 3) Focus on matching capability to need and thus meeting needs and don't be concerned about productizing some brand.
The boxes the vendors bid will all do a lot, they have many capabilities. How those capabilities get used to deliver service to the community is really up to the community to work out not for I2 to define with a specific set of products.
Matt Davy: Again he reiterated that this process really needs to create a Next Generation Platform that is available to all for any new activity they might conceive.
The gear purchased needs to: 1) Do what we do now 2) be from a vendor or vendors that share our overall vision of where the I2 network is headed.
John Moore: Differentiation of the I2 network from the carrier networks is very important. He shares Bill's perception that the vendors responses were somewhat disappointing. Perhaps too much of the RFP trying to force Open Flow into the carrier market when its really stronger and better developed in the Data Center Market.
- Akbar noted that this was probably from our being out ahead of the market.
Matt Davy: Johns perception is overall probably correct. We are certainly trying to drive the market. In a fundamental sense this is about providing a platform with a programmable API that can be used by the community.
Grover: There has maybe been too much emphasis placed on Open Flow/SDN in this process. He agreed that these devices as proposed by the vendors were not ready to run Open Flow in a production environment with the next year.
Bill Owens: Really the point is that there has not been sufficient input into this process.
- There was some discussion in general at this point about whether the RFP process really did in fact show us that the decisions on hardware did in fact establish if the hardware was able to adequately support the current set of services and activities of the network.
- The general view was that based on the responses and subsequent questions and answers any of the bid hardware would be able to do what the I2 network now does just fine. The only differentiator in the bids might be OPEX and the state of SDN development.
Chris Robb: What does this group (the NTAC) want its role and the process for its involvement to be?
- Put another way, "What would the NTAC want the AOAC to charge them to do?"
- In response to a question from Bill O. , Yes the AOAC has been aware of this process and the activities and the discussions.
Bill Owens: What will happen now? The context here was that the RFP was out, the review team has gone through it and a recommendation made, is there a reason for further architectural discussion or is this really a done deal?
Chris Robb: Its really better for the NTAC to tell I2 shat they would propose they do, I2 needs to hear how and to what degree they want to participate.
Akbar: 1) Yes the NTAC should continue to participate. 2) Use the next NTAC meeting to work though exactly what the role of the NTAC should be.
There is certainly some push to move in certain directions, there are some assumptions about what this new infrastructure would be like. But there are substantial and many questions about how to get from where we are now to realizing some of that vision. These are areas the NTAC would be very useful in addressing.
Jeff Bartig: There is currently a L3 backbone. There was a lot of activity around SC getting 100G interfaces put in. Is that 100G activity moving forward?
Chris Robb/Grover: For the moment we are not investing in any more 100 L3 interfaces. The build of the optical system continues to take place but ordering more 100G router interfaces will be on hold till final purchasing decisions are made.
Dale
In attendance, in no particular order except when they announced themselves:
Michael Lambert
Linda Roos
Chris Robb
Akbar Kara
Ryan Vaughn
David Crowe
Eric Boyd
James Deaton
Dale Finkelson
Joe St. Sauver
Bill Owens
John Moore
Dave Pokorney
Matt Zekauskas
Steve Wolff
Jon-Paul Heron
Brian Cashman
Shumon Huque
Denz Gurkan
Chris Griffin
Grover Browning
Jimmy Kyriannis
Wendy Huntoon
Cort Buffington
Jeff Bartig
Roger Hess
Matt Davy
Hans Adleman
Andrew Lee
Edward Aractingi