ClusterB: 2-13-09-cluster-b-invite.txt

Message-ID: <4963C7A6.1030401@bbn.com>
Date: Tue, 06 Jan 2009 16:05:42 -0500
From: Aaron Falk <falk@bbn.com>
User-Agent: Thunderbird 2.0.0.19 (Windows/20081209)
MIME-Version: 1.0
To: Larry Peterson <llp@cs.princeton.edu>, 
 Nick McKeown <nickm@stanford.edu>,
 Guido Appenzeller <appenz@cs.stanford.edu>, 
 Guru Parulkar <parulkar@stanford.edu>,
 Rob Ricci <ricci@cs.utah.edu>, Jon Turner <jon.turner@wustl.edu>, 
 Peter O'Neil <poneil@maxgigapop.net>,
 James Sterbenz <jpgs@ittc.ku.edu>
CC: Chip Elliott <celliott@bbn.com>, gpo-tech@bbn.com
Subject: Invite to Cluster B Networking meeting - Feb 13, 2009
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

Folks,

You are invited to a review of how the Spiral 1 control plane for Cluster B will interface with network aggregates and components. The Enterprise GENI project has examined the PlanetLab code and expressed concerns to us about the difficulty integrating OpenFlow equipment in Spiral 1 given the current state of the code and interfaces.

We feel this raises a broader concern about a critical technical issue.  Spiral 1 of the GENI control plane *must* demonstrate support for network aggregates and components, as this is an item of very high technical risk.

The plan has been to encourage aggregates to integrate using the PlanetLab-provided Aggregate Manager (AM). In other words, rather than implement an AM on their own to meet an interface spec, each project would adapt the aggregate-facing side of the PlanetLab reference AM.  The Enterprise GENI project has raised two specific concerns: First, if an aggregate is substantially different from a PlanetLab node, how hard is it to adapt PLC? Second, how hard will it be to evolve the over-the-wire interface once the initial cluster is integrated? We feel it is worth examining these two issues and develop consensus on integration strategy early on.

Again, demonstrating control plane functionality is one of the two central goals for Spiral 1. This meeting is a chance to have a small, focused f2f discussion about the specifics of the interface and its rationale. The objective of the meeting is to come to agreement on how to demonstrate in Spiral 1 that networking components and aggregates will be supported by the GENI control plane.  Key technical contributors for the PlanetLab and Enterprise GENI projects are invited along with representatives from other networking projects in Cluster B.  The ProtoGENI project is also invited since it shares the problem of supporting networking and interoperates with PlanetLab.

In preparation for the meeting, I'd like to ask each of network projects in Cluster B to review the current PlanetLab code and be prepared to talk about how you plan to integrate your aggregate.  If you like, you can prepare a few slides.

Strawman Agenda

  1. Review of PlanetLab code architecture
  2. Discussion of how to integrate PLC and network aggregates
  3. Develop a plan for demonstrating networking support in Spiral 1

Invited:

  PlanetLab: Larry Peterson & staff
  Enterprise GENI: Nick McKeown, Guru Parulkar, Guido Appenzeller
  ProtoGENI: Rob Ricci
  GpENI: James Sterbenz
  Mid-Atlantic Crossroads: Peter O'Neil
  Internet Scale Hosting: Jon Turner
  GPO staff

Logistics

  Time:     8:30am - 3:00pm
  Date:     February 13th, 2009
            plan to arrive the night before
  Location: Denver Airport Marriott at Gateway Park
            http://www.marriott.com/hotels/travel/dengp
  Expenses: paid by GPO
  Rooms:    RSVP to ekeohane@bbn.com
            GPO will reserve a room; you pay, we will reimburse