MPLS switching and IP routing - TCP-IP

This is a discussion on MPLS switching and IP routing - TCP-IP ; Bonsoir, I heard that a MPLS switching table can't be deduced without the help of an IP routing table. So it means that a LSR is necessarily an IP router. Do you agreee with that ? Best regards, Michelot...

+ Reply to Thread
Results 1 to 16 of 16

Thread: MPLS switching and IP routing

  1. MPLS switching and IP routing

    Bonsoir,

    I heard that a MPLS switching table can't be deduced without the help
    of an IP routing table. So it means that a LSR is necessarily an IP
    router.

    Do you agreee with that ?
    Best regards,
    Michelot


  2. Re: MPLS switching and IP routing

    > I heard that a MPLS switching table can't be deduced without the help
    > of an IP routing table. So it means that a LSR is necessarily an IP
    > router.


    The labels that are carried by an LSR (as opposed to an LER) normally
    correspond to router loopbacks. You need *something* to populate that
    MPLS switching table with labels, and this "something" is normally LDP
    or RSVP, based on IP addresses.

    In theory, an LSR could could have its MPLS switching table populated
    by another protocol and would not have to be an IP router. It practice
    I believe you would have a hard time finding an LSR that isn't also an
    IP router (the market for a such a box is probably vanishingly small).

    Steinar Haug, Nethelp consulting, sthaug@nethelp.no



  3. Re: MPLS switching and IP routing

    "Michelot" wrote:

    > Bonsoir,
    >
    > I heard that a MPLS switching table can't be deduced without the help
    > of an IP routing table. So it means that a LSR is necessarily an IP
    > router.
    >
    > Do you agreee with that ?
    > Best regards,


    Bonsoir Michelot,

    That doesn't sound right. MPLS switching tables should only operate on
    labels. The whole concept began with ATM switches and ATM VCs, after
    all.

    The label-switched path is supposed to be independent of any IP routing
    rules, so I don't see why IP routing tables would help deduce anything
    about the LSP. RFC 3031 says:

    MPLS forwarding can be done by switches which are capable of
    doing label lookup and replacement, but are either not capable
    of analyzing the network layer headers, or are not capable of
    analyzing the network layer headers at adequate speed.

    Sometimes it is desirable to force a packet to follow a
    particular route which is explicitly chosen at or before the
    time the packet enters the network, rather than being chosen by
    the normal dynamic routing algorithm as the packet travels
    through the network. This may be done as a matter of policy,
    or to support traffic engineering. In conventional forwarding,
    this requires the packet to carry an encoding of its route
    along with it ("source routing"). In MPLS, a label can be used
    to represent the route, so that the identity of the explicit
    route need not be carried with the packet.

    These decisions are not based on IP routing tables.

    Also:

    MPLS stands for "Multiprotocol" Label Switching, multiprotocol
    because its techniques are applicable to ANY network layer protocol.
    In this document, however, we focus on the use of IP as the network
    layer protocol.

    RFC 3031 does explain how the "forwarding class" CAN be based on IP
    header information, so the MPLS switch assigning the label may indeed
    have to be also an IP router. But the point is, this is not mandatory.

    Albert


  4. Re: MPLS switching and IP routing

    Bonjour Steinar,

    Thanks for your reply.

    > The labels that are carried by an LSR (as opposed to an LER) normally
    > correspond to router loopbacks.


    Why your word of loopback, could you give more?


    > You need *something* to populate that
    > MPLS switching table with labels, and this "something" is normally LDP
    > or RSVP, based on IP addresses.


    Yes, I understand that, and LDP seems the more commun use in today
    networks.

    > I believe you would have a hard time finding an LSR that isn't also an
    > IP router (the market for a such a box is probably vanishingly small).


    The issue is that configuration we meet today: IP over Ethernet over
    MPLS. In this case, it seems that MPLS doesn't refer to IP. About the
    box, for example, the 1660 SM from Alcatel treats MPLS, and it never
    refers to routing in its description, but only Ethernet switching.

    Thanks for your advice,
    Michelot


  5. Re: MPLS switching and IP routing

    Bonjour Albert,

    Thanks also for your text.

    > That doesn't sound right. MPLS switching tables should only operate on
    > labels.


    Yes, you're right. During the connection maintening phase, MPLS works
    only on labels. But we have to populate these labels in the switching
    table.

    > RFC 3031 says:
    >
    > MPLS forwarding can be done by switches which are capable of
    > doing label lookup and replacement, but are either not capable
    > of analyzing the network layer headers, or are not capable of
    > analyzing the network layer headers at adequate speed.
    >
    > These decisions are not based on IP routing tables.


    That concerns the connection maintening phase and not the
    establishment of the switching table in each LSR/LER.

    > MPLS stands for "Multiprotocol" Label Switching, multiprotocol
    > because its techniques are applicable to ANY network layer protocol.
    > In this document, however, we focus on the use of IP as the network
    > layer protocol.
    >
    > RFC 3031 does explain how the "forwarding class" CAN be based on IP
    > header information, so the MPLS switch assigning the label may indeed
    > have to be also an IP router. But the point is, this is not mandatory.


    Interresting, I need to read again that.
    IETF is only focused on IP and, perhaps, to see if a LSR could be an
    Ethernet switch we have to consider IEEE or ITU-T.

    Best regards,
    Michelot


  6. Re: MPLS switching and IP routing

    Michelot writes:
    > Interresting, I need to read again that.
    > IETF is only focused on IP and, perhaps, to see if a LSR could be an
    > Ethernet switch we have to consider IEEE or ITU-T.


    If you're interested in layer two forwarding ("Ethernet switch"), you
    might also want to look at TRILL in addition to MPLS.

    --
    James Carlson, Solaris Networking
    Sun Microsystems / 1 Network Drive 71.232W Vox +1 781 442 2084
    MS UBUR02-212 / Burlington MA 01803-2757 42.496N Fax +1 781 442 1677

  7. Re: MPLS switching and IP routing

    Hi James,

    > If you're interested in layer two forwarding ("Ethernet switch"), you
    > might also want to look at TRILL in addition to MPLS.


    Thanks for that concept.

    I saw that the more common TRILL encapsulation is TRILL over Ethernet:
    http://www.tools.ietf.org/html/draft...trill-encap-00

    My question is:

    When we have Ethernet over MPLS (and not the contrary) could the LSR/
    LER be only Ethernet switches, and not IP routers?

    Best regards,
    Michelot


  8. Re: MPLS switching and IP routing

    > When we have Ethernet over MPLS (and not the contrary) could the LSR/
    > LER be only Ethernet switches, and not IP routers?


    They would have to be more than plain Ethernet switches, since a normal
    Ethernet switch knows nothing about MPLS labels.

    Steinar Haug, Nethelp consulting, sthaug@nethelp.no

  9. Re: MPLS switching and IP routing

    On Jun 22, 10:16 am, Michelot wrote:
    > When we have Ethernet over MPLS (and not the contrary) could the LSR/
    > LER be only Ethernet switches, and not IP routers?



    No, an LER functioning as an ingress or egress router pretty much has
    to be an IP router (otherwise what happens to the packet at that
    point). An LSR need only look at the flow labels. Even in the case
    of an LSR, that's not the same function as an 802.1 bridge (it's
    actually somewhat simpler - you only have to deal with a modest number
    of flow labels rather than the entire set of MAC addresses on the LAN).


  10. Re: MPLS switching and IP routing

    Bonsoir Steinar,

    > > When we have Ethernet over MPLS (and not the contrary) could the LSR/
    > > LER be only Ethernet switches, and not IP routers?

    >
    > They would have to be more than plain Ethernet switches, since a normal
    > Ethernet switch knows nothing about MPLS labels.


    Yes, they are not plain switches bought at the supermarker, under
    plastic... We're talking about PE and P equipments with a possibly
    Ethernet switching function, located in the MAN or in the backbone.

    Best regards,
    Michelot


  11. Re: MPLS switching and IP routing

    Bonsoir Robert,

    Thanks to help me arguing and clarify the ideas. I think I have a
    beginning of solution.

    We know IP-MPLS which is IP over MPLS over Ethernet.
    The talking not concerns IP-MPLS but Ethernet-MPLS which is IP over
    Ethernet over MPLS.

    Ethernet-MPLS is called VPLS by IETF, and described e.g. in RFC 4762.

    > No, an LER functioning as an ingress or egress router pretty much has
    > to be an IP router


    In IP-MPLS or Ethernet-MPLS the LER/LSR include necessarily an IP
    routing function. The LDP signaling protocol is running on UDP for
    hello and TCP for exchanging tables.

    > (otherwise what happens to the packet at that point).


    In the Ethernet-MPLS system, an established connexion between 2 PE is
    called a pseudowire (PW), it is a MPLS label that points to a specific
    customer. These connexions can be prepared by LDP almost like in the
    IP-MPLS system.

    But, in the data plan (after establishing) the both systems are
    totally different. The PE (LER) is running like a switch. It learns
    MAC addresses on both sides, from customer side and from pseudowire
    side and put the MAC frame in the associated PW. Thus, for the
    transport, it is MAC over MPLS.

    Finally, in Ethernet-MPLS, the PE (or LER) include a IP routing
    function (in the control plan, for the establishing phase) and an
    Ethernet switching function in the data plan.

    The LSR behaviour is like in IP-MPLS: IP routing in the control plan,
    and MPLS switching in the data plan.

    Sorry if it is not clear.
    Thanks for your comments.
    Michelot


  12. Re: MPLS switching and IP routing

    > We know IP-MPLS which is IP over MPLS over Ethernet.
    > The talking not concerns IP-MPLS but Ethernet-MPLS which is IP over
    > Ethernet over MPLS.
    >
    > Ethernet-MPLS is called VPLS by IETF, and described e.g. in RFC 4762.


    Ethernet over MPLS comes in several forms. VPLS is *multipoint*, i.e.
    you simulate an Ethernet segment over MPLS. You can also have Martini
    tunnels which are *point to point*. There are significant differences.
    VPLS (multipoint) implies that the MPLS box needs to do MAC address
    learning, which is *not* necessary for an MPLS box which terminates a
    Martini tunnel.

    Also, since you can put basically anything on top of Ethernet, it need
    not be only IP over Ethernet over MPLS.

    Steinar Haug, Nethelp consulting, sthaug@nethelp.no

  13. Re: MPLS switching and IP routing

    Bonsoir Steinar,

    Thanks to correct me about the wrong association between VPLS and PW.

    > Ethernet over MPLS comes in several forms.


    In these services the customer Ethernet is carried end to end, from
    the CE source to the CE sink. We can quote the 3 service types of
    Ethernet over MPLS, as indicated in RFC 4026.

    (1) VPWS (virtual private wire service): the PE are connected in point
    to point, through PW
    (2) IPLS (IP only LAN-like service): the PE are connected in point to
    multipoint, the customer Ethernet has to contain only IP or ARP. It's
    a subset of VPLS
    (3) VPLS (virtual private LAN service): like IPLS but Ethernet can
    support everything.

    > VPLS (multipoint) implies that the MPLS box needs to do MAC address
    > learning, which is *not* necessary for an MPLS box which terminates a
    > Martini tunnel.


    You're right.

    > Also, since you can put basically anything on top of Ethernet, it need
    > not be only IP over Ethernet over MPLS.


    Except for the IPLS case if I understand correctly RFC 4086. You're
    right for VPWS and VPLS where Ethernet (802.3 or 802.1Q) can support
    PPP, PPPoE besides IP or ARP. I don't see anything else.

    Best regards,
    Michelot


  14. Re: MPLS switching and IP routing

    Bonjour Michelot,

    > Thanks to correct me about the wrong association between VPLS and PW.


    Euh... no! VPLS is using PW. The point to multipoint VPLS is made of
    multiple points to points PW.

    "[G.8110.1 (11/2006)] Pseudowires can be used as a packet transport
    mechanism for emulating LAN Services over a Transport MPLS network.
    Like in Virtual Private LAN Service (VPLS) a full mesh of T-MPLS
    trails (PW trails) is used to interconnect split-horizon ETH Flow
    Domains".

    Best regards,
    Michelot


  15. Re: MPLS switching and IP routing

    > > Also, since you can put basically anything on top of Ethernet, it need
    > > not be only IP over Ethernet over MPLS.

    >
    > Except for the IPLS case if I understand correctly RFC 4086. You're
    > right for VPWS and VPLS where Ethernet (802.3 or 802.1Q) can support
    > PPP, PPPoE besides IP or ARP. I don't see anything else.


    Martini tunnels and VPLS can certainly support much more than just PPP,
    PPPoE, IP and ARP :-) Should you desire to use it for DECnet, Novell
    IPX, NetBEUI or other strange things - it'll work just fine.

    Steinar Haug, Nethelp consulting, sthaug@nethelp.no

  16. Re: MPLS switching and IP routing

    Bonjour Steinar,

    > Martini tunnels and VPLS can certainly support much more than just PPP,
    > PPPoE, IP and ARP :-) Should you desire to use it for DECnet, Novell
    > IPX, NetBEUI or other strange things - it'll work just fine.


    Ok, I'm not very familiar of legacy strange-thing protocols. Thanks to
    put my feet on earth!

    Best regards,
    Michelot


+ Reply to Thread