NTAC Meeting 7-2-13
IPv6 WG report:
There is a NetGuru meeting planned for GEC. Alan will suggest some v6 related topic and questions for that group.
A question has come up about whether it makes sense to continue to offer v6 transit through 11537. Opinions should be offered to the WG mailing list.
The chairs are planning on having a WG call in Aug.
Peering and Routing:
There has been discussion in the WG about the process of upgrading some of the paths between the TR-CPS devices and the backbone.
SDN:
At GEC 17 there will be both an SDN WG meeting and a NetGuru's meeting.
Agenda items being considered:
- Panel on success/horror stories related to SDN deployment.
- Discuss potential agenda for a Focused Technical Workshop.
- Operations supporting researchers, they should have far more researchers at the meeting then usual.
- Interdomain SDN
- Spanning tree and Open Flow
- There has been discussion about Open Daylight.
GEC 17 will be in Madison, WI on July 21-23. See GENI.net for more information.
I2 Update:
Chris talked about the TSB from Brocade. The new code releases have been tested and it is believed a stable build has been identified that corrects the problem discussed in the TSB. This code should be available Mid-Week (week of July 1). The NOC will spend a week or so doing burn in tests and assuming its good deploy after that. This is a very high priority activity.
Grover discussed the AL2S-AL3S plans:
The plans are as discussed in the document sent to the NTAC recently.
Basic features of the original plan:
- Use the AL2S 100G transport for the IP network.
- There will be 2 100G paths between the routers and the AL2S nodes.
- Use traditional vlans.
- Use Open Flow configured vlans.
During operational testing it was discovered that: - Traditional vlans do not work on the same interface as an Open Flow configured vlan.
- This is not a bug, it is a feature that has not been implemented as of yet.
- Going forward instead of a path based on traditional L2 vlans use the inter router L1 links.
- Those paths will be augmented as needed.
The idea is still to get to using the Open Flow based vlans as the primary paths. This move will be based on understanding the stability of that path.
Jay Ford asked if all of the connections from end sites are subject to the same restrictions? They will not be able to use traditional vlans and Open Flow based vlans either. He also asked if the SDN vlan numbers were sufficient to scale to this scenario. The answer to this was yes.
The Brocades will still do traditional vlans, just not on the ports that are configured to do SDN.
Darrell asked about the traditional 10G connections to the routers?
- These will be available for at least the next year. They are in place to allow for building confidence in the Open Flow paths and to provide a backup path for the end sites.
Jay Ford asked if this came about because the testing was not sufficient?
- Grover answered that the belief was the testing was fine. However the testing mistakenly covered a slightly different scenario then intended. For the tests the traditional vlans on the ports also had an IP address associated with them. That works just fine. What was not tested was pure layer 2 between ports in a box with Open Flow also configured.
There are already a sure of tools set up to monitor the IS-IS adjacencies between the routers over the AL2S path. This can be used to monitor the availability of those paths. Chris pointed out that the metrics on those adjacencies was set very high, such that they would never come into use.
- The NOC is collecting data
- However simply monitoring the adjacencies might not be fully sufficient for determining availability. The NOC is looking a some other options to make sure the reporting is accurate.
Darrell asked how long the IS-IS adjacencies have been in place?
Chris answered that it has just been a few days. All of the physical interfaces were only recently installed. Once they were done IS-IS was set up.
A question was raised about the Junipers. The Juniper AL2S devices do not have the specific problems in the Brocade TSB. New code for the Junipers is expected next week. This code will make the junipers much more stable then they have been to date.
There was then a discussion about timing.
Grover stated that it would be good to have all the backbone capacity in use by Aug 15. That was based on the restart of the fall semester soon after that date. He also pointed out that having the data to inform the decision about using the AL2S paths would be important. And that the connectors require the data sooner rather then later so that they can make any needed decisions.
Given these goals what is the way to move forward?
The general view was try to get any data possible out ASAP, even if its not the best data. Over the summer continue to refine and harden the data.
Node installs:
Chris reported that the BTOP build was finished. The last sites to go up were:
Phoenix, Albany, Boston and Minneapolis.
Sites coming up are (in no particular order):
Portland
Jackson, MS
Pittsburgh
Columbia, MO.
Grover Browning, Cas d'Angelo, Andrew Lee, Hans Addleman, Matt Z, Chris Robb, Chris Spears, Brian Cashman, Linda Roos, Ryan Harden, Lina WInkler, Alan Whinery, Darrell Newcomb, Alison, Michael Lambert, David Crowe, Joe St Sauver, Dan Schmidt, Dave Pokorney, Tom Johnson, David Wood, Jay Ford, Dale Finkelson