Automated GOLEs and Fenius: Pragmatic interoperabilityWinter Joint Techs 2011

Clemson, SC

Evangelos Chaniotakis, ESnet Network Engineer

Lawrence Berkeley National Lab

2Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science

Introductions

• GLIF: Global Lambda Integrated Facility– Many participants: NRENs, exchanges, sites, companies

• GOLE: GLIF Open Lightpath Exchange – Many GLIF participants operate a GOLE

• Automated GOLE Task Force– Founded in 2010 with the objective of automating lightpath

provisioning across multiple GOLEs– Current chair: Jerry Sobieski– Close collaboration with GNI API Task Force, OGF NSI, NML

• GNI API Task Force:– Founded in 2008 with the objective of developing a common API

between circuit provisioning services– Chair: Evangelos Chaniotakis

3Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science

GLIF map

4Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science

Motivation

• The research community has a global, well connected “network of networks”

• Many networks provide some circuit-oriented service,• And, it is becoming increasingly common for these circuits

to be “stitched” across multiple networks,SC09: ~20 L2 circuits. SC10: ~50, SC11: ?• Stitching these “by hand” is s l o w – and cuasesss errors.• Software does exist to automate circuit provisioning....• But it generally doesn’t work when stitching,• Because we don’t have a common provisioning toolkit

5Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science

Fenius

A simple common provisioning API and toolkit + Just good enough to support common stitching case today,+ Software that implements it,+ That works with existing provisioning software,+ Motivation for network operators to deploy & operate,

=(the potential for) Global Interoperability

6Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science

Automating the GOLEs

• Several GOLEs committed network and human resources for a one-year test drive of automation,

• The Fenius software project by GNI API Task Force was decided be used as a thin interoperability layer.

• Fenius instances were deployed to Automated GOLE sites• The various NRMs were configured to work with Fenius,• Topology information was collected,• PerfSONAR PingER nodes are deployed to test the data

plane• Weekly calls to keep the project alive & kicking

7Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science

Automated GOLE Deployment

• Asia:– JGN2+, AIST, KDDI

• USA: – ESnet SDN, Internet2 ION & MANLAN, – StarLight, USLHCnet, Caltech

• Europe: – Netherlight, CERNLight, University of Amsterdam,– NorthernLight, PSNC, CERNLight,

8Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science

Global Deployment

9Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science

PingER Demonstration

JGN2+PingER

KDDIPingER

UvAPingER

PSNCPingER

JGN2+

KDDI

StarLight,USLHCnet,

Internet2 ION,NetherLight,

PSNC

10Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science

Demonstration control plane

JGN2+PingER

KDDIPingER

UvAPingER

PSNCPingER

JGN2+IDC

KDDIG-Lambda

StarLight Argia,USLHCnet IDC,

Internet2 ION IDC,NetherLight DRAC,

PSNC,

FeniusSuperAgent

11Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science

Demonstration

12Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science

Fenius Components

• Common provisioning API over web services,• Translation framework for Java (optional)• Translators for IDC, G-lambda, Harmony, dynamicKL• Native implementation in OpenDRAC,• Scriptable command-line client • Web UI “Superagent” with topology and pathfinding• Monitoring and visualization (in progress)

13Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science

Translation Framework

CommonWeb UIClient

G-lambda

Common API

G-lambdaTranslator

IDC

Common API

IDCTranslator

CommonCommand-Line

Client

CommonMonitoring

Client

DRAC

Common API

14Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science

Challenges and lessons learned

• Be enthusiastic and believe in your project• Getting collaborators interested can be difficult

– Get buy-in from their management• Resources can be unreliable

– Get buy-in from their management (and yours)• Remember to account for differences in time zones,

language, culture• Once project gains momentum, keep the ball rolling• Keep away from politics • Make sure everyone has a chance to contribute• You will need to work long hours

15Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science

Future directions

• Deploy in more networks,• Implement missing features,• Extend API and toolkit (keeping it simple!),• Gain support from more network management software,• Harden software & make it operational,• Improve user interface• Better documentation• Take over the world! :)

16Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science

Special Thanks

• Frank Blankman• Scott Campbell• Joan Garcia Espin• John Graham• Jeroen van der Ham• Takatoshi Ikeda• Nils Jacobson• Gigi Karmous-Edwards• Tomohiro Kudoh• Kavitha Kumar• Ali Lahlou• Andrew Lake• Tom Lehman• Mathieu Lemay• John MacAuley

• Gerben van Malenstein• Edoardo Martelli• Takahiro Miyamoto• Azher Mughal• Fumihiro Okazaki• Jan Radil• Jordi Ferrer Riera• Sandor Rosza• Ryousei Takano• Thomas Tam• Jin Tanaka• Alan Verlo• John Vollbrecht• Fred Wan• Xi Yang