NTAC call
Tuesday, January 16, 2007
In the call:
Ana Preston, Rob Vietzke, Matt Zekauskas, Paul Schopis, David Wood, Marla Meehl, Jerry Sobieski, John Streck, Dave Pokorney, Jose Dominguez, John Moore, Jeff Schiller, Joe St Sauver, Leo Donelly, Dan Magorian, Mark Johnson, Matt Davy, Michael Lambert
New and pending ACTION ITEMS:
1. Dave Farmer, Paul Schopis, Dave Reese offered to write an initial recommendation/write up on membership within this group, roles and expectations.
STATUS: Paul will ping them and check on status.
2. Eric Boyd to create new list (consolidated on measurement - see notes below); expect an email from Eric
STATUS: Matt Zekauskas offered to check with Eric. Will report back via email or at next call
3. Paul to check with Jerry Sobieski on Layer 2 working group
4. Dan Magorian to report back at next NTAC call status and updates on commodity/peering working group and when at next Joint Techs group will be convened.
Next call: February 6, 2007 at 4 p.m. ET
NOTES from the call:
Reminder that NTAC calls are 1st and 3rd Tuesday of each month. Focus of calls will be on NTAC (1st Tuesday) and suggested working group updates (3rd Tuesday).
1st call of year was cancelled due to it happening on Jan 2, 2007. This is the first call of the group after Dec face to face meeting.
Internet2 Network update provided by Rob Vietzke:
- Connectors starting to cut over in Northeast - in addition to first leg from NY to Chicago. Working on Boston, Washington and Chicago.
- Coordination and working with connectors to get ready and getting transition (see NOC/Chris Robb's tracking at http://i2net.blogspot.com/ )
- Atlanta coming up on March 2nd. See http://www.internet2.edu/network/deployment.html for latest deployment schedule
- Expect ring through Houston and Kansas City to be ready by April
- Alluded to series of calls with connectors. If any of you have any issues, please send email. If you haven't heard from us and think you should, let us know.
- On schedule so far.
- On dynamic services: working a lot, alluded to joint work with Europeans on services trial.
- Trying to get control plane integrated with CIENA. Some lab testing going on with NOC to get this.
- Reported too doing some work to expedite new wave turn-ups and cross-connects.
Question on routers being pulled out. Rob reported on router swap in NY to get loaner Juniper out off there and get loaner to Indianapolis. Hoping to happen by 2/1/07.
Indy router by 1st week of Feb.
Internet2 Network Service description Document:
Input on draft sent on Internet2 network Service descriptions. No input so far. Ana will resend to the group.
NTAC issues:
NTAC composition
Paul provided background and alluded to NTAC "makeup" and membership requirements. Dave Farmer, Paul Schopis, Dave Reese offered to write an initial recommendation/write up on membership within this group, roles and expectations.
ACTION: Paul will ping them to check on status
Follow up on Measurement
Paul asked about status re. Measurement and per NTAC face to face, Eric Boyd was going to create new list; matt Z to check
Follow up on layer 2 working group:
ACTION: Paul to talk to Jerry about this.
Commodity and Peering working group.
Alluded to including this during Joint Techs. Dan and Ana will talk offline to see what can be done.
ACTION: for next call, summarize where discussion may be taking place while at Joint Techs
Discussion about what type of questions to pose to group. Suggested use same AS or different AS?
Policy/distinguishability from Abilene. Does peering need to be on separate AS?
Per Dan?
Had good discussion on that in December. Since current is RFC2547. 11537.
Paul alluded to OARNET: today have two internal ASes at OSU because originally (way back) (early Abilene) way asked to do it. 1998. concerned that since 2 ASes talk
internally, not cause confusion when same groups of routes from same AS.
Discussion around combining routes. taking one, moving to other, and way to move back.
if figured out how to do that for existing routes, should still work.
Non-internet2 eligible, eligible to use peering service?
Know its being that way for trial. don't know if definitive stuff beyond that.
In Oregon, if maintain MPLS internally, and change service model so don't join two routing tables which we don't do today at Oregon gigapop.
Maintaining same AS, will end up with just one routing table, which we would have to have way to differentiate prefix origins. can do communities, but makes it harder.
Someone from call: don't mix. OGIG and ron are two separate entities w/separate devices.
So although same AS, come in separate
Someone said doing MPLS VPNs. combine into one routing table, peer with RON and any other connector.
Not an issue for directly connected participants. passing to RON is different game. either separate device, and make ron peer with that one. or..
Suggestion that group talks with as many people as possible. Make recommendations.
Someone: Have same issues at GigaPop. Statewide network gets one table.
looking at doing VRFs to separate to see both.
But it's a transport problem. if no cost issue,
Another port off of Abilene router, dump into another, but only have one pipe, and cost issue.
unless 2547 and have way to keep separate, lots of fancy plumbing
to get them off.
Rob Vietzke: my understanding is that under new AUP, much less restrictive on what routes pass to R&E
Dan Magorian: envisioning reverse fish problem. solved policy issue, since this is more relaxed, undo what did for fish.
Bill Owens: w/in NYSERNET had to play weird games with multi-level route maps to prevent violations with NYSERNET AUP, because of unintentional bending of old conditions of use.
Hope capability continues to exist in new network and that there is no commercial traffic across NYSERNET network. That is my biggest concern. Want to make sure none of those routes leak into network. Likewise, that mine don't leak into peering.
Dan re the MAX: 2547 works fairly well. same as used everywhere in max.
no leakage.
Gigapops running 2547 that intend to accept internet2's handoff on this from their VRF into your VRF no issue.
but even people that don't want to run VRFs, and mix as choice, that's not different than what we do to customers right now.
if choose to keep separate, never see routes or traffic.
Folks that are using separate routers, not vrfs, could potentially have issue that paul/jose were saying.
If single connection does two things, have to get out of router w/o mixing in, do fancy x-c or pass through switch first or something.
Topology or l2 separation issue, not what AS should be used.
Oregon: could potentially change way RON works. they run commodity for everyone
in the state. They also do some Internet2 traffic – also do some I2 traffic for SEGPs - for that not that big of an issue. Did have fish problem for some ELTs. Got creative. By adding commodity or peering, changes a little bit. troubleshooting could be harder.
Tell us what you see the advantage as bb provider of single AS?
Matt Davy: tech wise, option to not running same AS to run logical routers on Juniper. Our feeling was that 2547 VPNs were (1) standard, so can swap vendor choice and (logical routers are proprietary) (2) widely deployed in commercial world, so code stable & bug free than logical routers or equivalent.
Weird feature, enterprise in couple of locations, to run same AS on customer sites, and pass through middle w/o as loop detection.
Dan: as override. Use that for Internet2 DC office. Had long discussion with SEs at Juniper, not work for what we want to do.
trial is running with 2547. Mature technology, reasonably happy with.
Paul: using logical routers in our net for 3 yrs, no problem but would by arg standard solution
Would hate to do another transition with un-interoperable technology
Question: Roadmap for replacing routers at this point? Nope
Jose suggesting polling gigapops on this - or has this been done? He could not make the first meeting - don't know if all represented in working group?
Good idea to do anyway, people need to know what's coming up.
Paul/Dan: process question. Could write a 1-2 page summary, share and say agree/disagree. Would be good to lay out what the recommendations should be - send maybe to working group first before sending to overall NTAC.
More discussion will be taking place at the Joint Techs; expect also some more discussion to take place via email.
Call adjourned.
NTAC roster as of Jan 2007:
Brian Court
Cas D'Angelo
Cort Buffington
Dave Reese
Dave Pokorney
David Wood
Dave Jent
Dale Finkelson
Dale Smith
David Farmer
George Loftus
Jose Dominguez
Jerry Sobieski
John Moore
Jeff Schiller
Joe St Sauver
John Streck
Michael Lambert
Leo Donnelly
Dan Magorian
Matt Zekauskas
Mark Johnson
Matt Davy
Mike Schlicht
Bill Owens
Paul Schopis
Ron Hutchins
Tim Lance
Linda Winkler