Multihoming with Locator/ID Separation Protocol: An ... ljilja/cnl/presentations/haeri/ManFI_2015_ ¢ 

Multihoming with Locator/ID Separation Protocol: An ... ljilja/cnl/presentations/haeri/ManFI_2015_ ¢ 

  • View
    0

  • Download
    0

Embed Size (px)

Text of Multihoming with Locator/ID Separation Protocol: An ... ljilja/cnl/presentations/haeri/ManFI_2015_...

  • Soroush Haeri, Rajvir Gill, Marilyn Hay, Toby Wong, and Ljiljana Trajković Communication Networks Laboratory, Simon Fraser University, Vancouver, British Columbia, Canada

    BCNET Vancouver, British Columbia, Canada

    Communication Networks Laboratory http://www.ensc.sfu.ca/~ljilja/cnl/

    References •  BCNET [Online]. Available: http://www.bc.net. •  D. Farinacci, V. Fuller, D. Meyer, and D. Lewis, “The locator/ID separation protocol

    (LISP),” IETF, RFC 6830, Jan. 2013. •  G. Huston. (April 1, 2015) Analysis of BGP routing table dynamics [Online].

    Available: http://bgp.potaroo.net/. •  LISP pilot network [Online]. Available: http://www.lisp4.net/. •  D. Meyer, L. Zhang, and K. Fall, “Report from the IAB workshop on routing and

    addressing,” IETF, RFC 4984, Sept. 2007.

    Multihoming with Locator/ID Separation Protocol: An Experimental Testbed

    •  LISP employs the map-and-encap technique •  It operates in the data and control planes and introduces two new

    namespaces:

    •  Endpoint Identifiers (EIDs) for connection endpoints •  Routing Locators (RLOCs) for the connection point of a node to the

    network

    •  Ingress Tunnel Routers (ITRs): perform EID-to-RLOC mappings and are responsible for encapsulating the outgoing packets

    •  Egress Tunnel Routers (ETRs): accept LISP encapsulated packets, de-capsulate the packets, and forward them to end devices

    •  British Columbia's network extends over 1,400 km and connects Kamloops, Kelowna, Prince George, Vancouver, and Victoria

    •  Shown is the traffic bound for CANARIE (Canada’s Advanced Research and Innovation Network), the commercial Internet (Transits), and the Vancouver Internet Exchange (VANIX)

    ManFI 2015 The Seventh IFIP/IEEE International Workshop on

    Management of the Future Internet May 2015, Ottawa, ON, Canada

    BCNET Scalability of the Internet Architecture •  The Internet Protocol (IP) namespace is overloaded with information

    about the location and the identity of network devices

    •  Locator/identifier overload is a recognized cause of the exponential growth of the routing tables in the Internet's Default-Free Zone (DFZ)

    •  Two main techniques to achieve the locator/ID split: address rewriting and map-and-encap

    Growth of the BGP FIB table from 1997 to 2015

    Locator/ID Separation Protocol (LISP)

    Data Plane

    Contol Plane

    ITR ETR

    MR MS

    M ap

    -R eq

    ue st

    M ap

    -R ep

    ly

    Ne g.

    M ap

    -R ep

    ly

    M ap

    -R eg

    ist er

    M ap

    -N ot

    ify

    LISP control and data planes and the exchanged messages

    Multihomed LISP Testbed

    EID Space: A.B.C.16/28

    EID: A.B.C.30 Subnet A

    Subnet B

    BCNET Internet

    LISP Pilot Network

    P-xTR

    MS/MR

    Virtual Communication Channel

    xTR

    Simon Fraser University Burnaby Campus

    RLOC: X.Y.Z.9

    RLOC: X.Y.Z.13

    •  Location: Communication Networks Laboratory at Simon Fraser University Burnaby campus in British Columbia, Canada

    •  Network architecture: a work-station connected via a dedicated fiber link to a LISP router (xTR) located at the BCNET transit exchange center in downtown Vancouver

    •  LISP router (xTR) is connected to two BCNET subnets •  LISP pilot network infrastructure is used for control plane operations •  EID address space A.B.C.16/28 is assigned by the LISP pilot network

    Testbed Performance Evaluation •  Traffic: peer-to-peer BitTorrent file transfer of “The Zeitgeist: Moving

    Forward” motion picture

    •  Data transfer: 6.7 GB of data exchanged within 28 min and 15 s •  Distributions of traffic on the site connections was periodically changed •  Various link failures were tested by shutting down the network interfaces

    of the LISP router

    •  Traffic captured: inbound and outbound traffic to and from the multihomed site

    •  Traffic traces contain only the headers of the transmitted packets

    00:00 02:00 04:00 06:00 08:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 24:00 26:00 28:00 30:000

    10

    20

    30

    40

    50

    60

    70

    80

    90

    100

    Time (mm:ss)

    Ra tio

    o f p

    ac ke

    ts tra

    ns im

    itte d

    via S

    ub ne

    t A (%

    )

    Link to Subnet A

    Measured ratio Assigned ratio

    00:00 02:00 04:00 06:00 08:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 24:00 26:00 28:00 30:00 0

    5

    10

    15

    20

    25

    30

    35

    40

    45

    50

    Time (mm:ss)

    Th ro

    ug hp

    ut (M

    bp s)

    Subnet A Subnet B

    00:00 02:00 04:00 06:00 08:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 24:00 26:00 28:00 30:000

    10

    20

    30

    40

    50

    60

    70

    80

    90

    100

    Time (mm:ss)

    Ra tio

    o f p

    ac ke

    ts tra

    ns im

    itte d

    via S

    ub ne

    t B (%

    )

    Link to Subnet B

    Measured ratio Assigned ratio

    BCNET traffic map: March 31 2015 18:03:04