MPLS etc.. 9 May 2017 AN

Transcription

1 MPLS etc.. 9 May 2017 AN

2 Multi-Protocol Label Switching MPLS-TP FEC VPLS PBB-TE LDP MPLS-TE LABEL MP-BGP MPLS is not alone LSP TAG H-VPLS GMPLS ISIS-TE MPƛS RSVP-TE SR OSPF-TE T-MPLS PCEP

3 Multi-Protocol Label Switching...because its techniques are applicable to ANY network layer protocol. It was designed to provide a unified data-carrying service for both circuit-based clients and packet-switching clients which provide a datagram service model; (converged AToM networks) It can be used to carry many different kinds of traffic, including IP packets, as well as native ATM, SONET, Frame Relay and Ethernet frames, but also Pseudo Wires (PW), VPLS, IP VPNs; Label Switched Paths are similar to circuit-switched paths (virtual circuit) in ATM or FR networks, except not dependent on particular Layer-2 technology;

4 Multi-Protocol Label Switching Label switching idea brilliant innovation? Back then ATM was popular, only encapsulations for flow over ATM were defined ATM did not become the big hit and is mostly replaced now by IP/Ethernet devices

5 Multi-Protocol Label Switching Next Cisco came with Tag Switching (1997); This was brought to the IETF for open standardisation; IETF working group involved other vendors and MPLS was defined; Tag renamed to Label; Really: Back then traffic was growing faster than router vendor and service providers could keep up with; Existing routing equipment was very expensive; performance was not enough (no in hardware forwarding of packets); Fixed length label lookup was faster; RFC-3031: Multiprotocol Label Switching Architecture [June 2001]

6 ATM switches offered higher-speed interfaces and faster forwarding; ISPs were building backbones with ATM switches and routers as edge devices; That required building full-mesh networks and that's a lot of configuration and difficult management; Vendors were trying to implement tighter integration between router and ATM switch control planes; Label idea comes from "label swapping" in Frame Relay, ATM. Basically: edge device of MPLS network applies tag tag switch forwards according to label swapping table (pre-established) edge device removes tag and forwards packet

7 It was invented for fast(er) routing" (more like L2 switching); IDEA: Each flow might be special but numerous flows share the same forwarding behaviour; All packets with same label follow the same path; Fixed length label lookup is faster than longest match lookups; Applications TCP UDP SCTP, MPTCP, DCCP, MPLS is sometimes Ethernet IP MPLS ATM FR PPP called: Layer 2.5 protocol Physical

8 Routing vs Switching Switches switch and Routers route. Switching makes packet (frame) forwarding decisions based on layer 2 data [MAC address] Routing makes packet forwarding decision based on layer 3 data [IP address] Switch switches within the subnet/within the network Router routes between the networks Gateways?? Bridges?? Forwarding?? recursive lookup - to determine the next hop + outbound interface

9 Topology Database IP Routing table Forwarding table OSPF BGP RIB Routing Information Base FIB Forwarding Information Base Not all entries in RIB have next hops that are directly connected; OSPF route has an outgoing interface; it s computed by the SPF algorithm and transferred into the IP routing table; BGP route has no outgoing interface and its next hop is not directly connected; the router has to perform recursive lookups to find the outgoing interface; IP routes are copied to Forwarding Information Base (FIB) and their next hops are resolved, outgoing interfaces are computed and multiple entries are created when the next-hop resolution results in multiple paths to the same destination; Simplified view

10 Routing vs Switching Switches switch and Routers route. Juniper specific

11 Routing vs Switching Switches switch and Routers route. Bridging Switching makes packet (frame) forwarding decisions based on layer 2 data [MAC address] Routing makes packet forwarding decision based on layer 3 data [IP address] MPLS is somewhere in between??? (Layer 2.5) Regardless of how you decide to call the physical (or virtual) device that forwards the data across your network, it s important to understand whether it forwards the data based on physical layer-2 addresses (we called that bridging) or based on logical, hierarchical layer-3 addresses (what we called routing 20 years ago). Ivan Pepelnjak Blog:

12 Routing vs Switching MPLS-based networks scale better than those using ATM or Frame Relay because of two major improvements: Automatic setup of virtual circuits based on network topology (core IP routing information), both between the core switches and between the core (P-routers) and edge (PE-routers) devices. Unless configured otherwise, IP routing protocol performs topology autodiscovery and LDP establishes a full mesh of virtual circuits across the core. VC merge: Virtual circuits from multiple ingress points to the same egress point can merge within the network. VC merge significantly reduces the overall number of VCs (and the amount of state the core switches have to keep) in fully meshed networks. Important purpose of MPLS: reduces routes in routers and forwarding table entries Multiple IP prefixes into one tag

13 Multi-Protocol Label Switching Protocol used in the core of networks LER = Label Edge Router (or PE = Provider Edge router) LSR = Label Switching Router (or P = Provider router) Single domain (ISP) Can be used for Traffic Engineering (TE) LSP R3 R6 LER R10 CE Label push LSR Label swap R9 R1 CE Ingress R2 LER R5 LSR Transit R8 LER CE Egress R4 LSR R7 LSR Label pop LSP = Label Switched Path: unidirectional path between LERs

14 Multi-Protocol Label Switching MPLS header applied to packet [This header is put between layer-2 and layer-3 header (shim header) in IP] Label EXP S TTL Label field= 20 bits EXP field = 3 bits S field = 1 bit TTL field = 8 bits Label = number, picked by the router (local) EXP = experimental bits, for Class of Service (*) S = Bottom of Stack TTL = Time-to-Live (to detect loops) Protocol to establish an end-to-end path through the MPLS network Still hop-by-hop forwarding mechanism Builds a connection-oriented service on the IP network (*) Renamed to Traffic Class field RFC-5462

15 Multi-Protocol Label Switching Forwarding Equivalent Class - all IP packets follow same path and receive same treatment at each node FEC can be destination prefix, FEC = Forwarding Equivalence Class or source address, DSCP, The ingress router receives packet and determines to which FEC it belongs; Packets which should be forwarded in the same manner belong to same FEC; Forwarded with the same label (over the same LSP); Label Forwarding Information Base - forwarding table mapping between labels to outgoing interfaces LIB = Label Information Base Mapping between previous hop (incoming port, label) and FEC; Mapping between FEC and next hop (outgoing port, label); Each router has its own LIB, generates LFIB (Label Forwarding Information Base);

16 Prefixes distributed with OSPF/ISIS LSR/transit LSR/transit LSR/transit FEC = loopback address PE1 ingress PE2 Z Y X PE1 egress (PE1,L1) PE1 assigns a Label (L1) to its own L0 address (FEC) and advermses that to its LDP peer X

17 Prefixes distributed with OSPF/ISIS LSR/transit LSR/transit LSR/transit FEC = loopback address PE1 ingress PE2 Z Y X PE1 egress (PE1,L2) (PE1,L1) X evaluates whether PE1 is on the IGP shortest path for that FEC. If successful X assigns L2 for FEC PE1, installs forwarding state swapping L2 and L1 and advermses a binding for L2 and FEC PE1 to Y. PE1 assigns a Label (L1) to its own L0 address (FEC) and advermses that to its LDP peer X

18 Prefixes distributed with OSPF/ISIS LSR/transit LSR/transit LSR/transit FEC = loopback address PE1 ingress PE2 Z Y X PE1 egress (PE1,L3) (PE1,L2) (PE1,L1) Y will do similar processing. The LSP setup proceeds from egress to ingress. X evaluates whether PE1 is on the IGP shortest path for that FEC. If successful X assigns L2 for FEC PE1, installs forwarding state swapping L2 and L1 and advermses a binding for L2 and FEC PE1 to Y. PE1 assigns a Label (L1) to its own L0 address (FEC) and advermses that to its LDP peer X

19 Prefixes distributed with OSPF/ISIS LSR/transit LSR/transit LSR/transit FEC = loopback address PE1 ingress PE2 Z Y X PE1 egress (PE1,L4) (PE1,L3) (PE1,L2) (PE1,L1) Y will do similar processing. The LSP setup proceeds from egress to ingress. X evaluates whether PE1 is on the IGP shortest path for that FEC. If successful X assigns L2 for FEC PE1, installs forwarding state swapping L2 and L1 and advermses a binding for L2 and FEC PE1 to Y. PE1 assigns a Label (L1) to its own L0 address (FEC) and advermses that to its LDP peer X

20 Prefixes distributed with OSPF/ISIS LSR/transit LSR/transit LSR/transit FEC = loopback address PE1 ingress PE2 Z Y X PE1 egress PE1-push L4 (PE1,L4) (PE1,L3) (PE1,L2) (PE1,L1) Y will do similar processing. The LSP setup proceeds from egress to ingress. X evaluates whether PE1 is on the IGP shortest path for that FEC. If successful X assigns L2 for FEC PE1, installs forwarding state swapping L2 and L1 and advermses a binding for L2 and FEC PE1 to Y. PE1 assigns a Label (L1) to its own L0 address (FEC) and advermses that to its LDP peer X

21 Prefixes distributed with OSPF/ISIS LSR/transit LSR/transit LSR/transit FEC = loopback address PE1 ingress PE2 Z Y X PE1 egress PE1-push L4 (PE1,L4) (PE1,L3) (PE1,L2) (PE1,L1) swap (L4, L3) Y will do similar processing. The LSP setup proceeds from egress to ingress. X evaluates whether PE1 is on the IGP shortest path for that FEC. If successful X assigns L2 for FEC PE1, installs forwarding state swapping L2 and L1 and advermses a binding for L2 and FEC PE1 to Y. PE1 assigns a Label (L1) to its own L0 address (FEC) and advermses that to its LDP peer X

22 Prefixes distributed with OSPF/ISIS LSR/transit LSR/transit LSR/transit FEC = loopback address PE1 ingress PE2 Z Y X PE1 egress PE1-push L4 (PE1,L4) (PE1,L3) (PE1,L2) swap (L4, L3) swap (L3, L2) (PE1,L1) Y will do similar processing. The LSP setup proceeds from egress to ingress. X evaluates whether PE1 is on the IGP shortest path for that FEC. If successful X assigns L2 for FEC PE1, installs forwarding state swapping L2 and L1 and advermses a binding for L2 and FEC PE1 to Y. PE1 assigns a Label (L1) to its own L0 address (FEC) and advermses that to its LDP peer X

23 Prefixes distributed with OSPF/ISIS LSR/transit LSR/transit LSR/transit FEC = loopback address PE1 ingress PE2 Z Y X PE1 egress PE1-push L4 (PE1,L4) (PE1,L3) (PE1,L2) swap (L4, L3) swap (L3, L2) (PE1,L1) swap (L2, L1) Y will do similar processing. The LSP setup proceeds from egress to ingress. X evaluates whether PE1 is on the IGP shortest path for that FEC. If successful X assigns L2 for FEC PE1, installs forwarding state swapping L2 and L1 and advermses a binding for L2 and FEC PE1 to Y. PE1 assigns a Label (L1) to its own L0 address (FEC) and advermses that to its LDP peer X

24 LSP LSR/transit LSR/transit LSR/transit Prefixes distributed with OSPF/ISIS FEC = loopback address PE1 ingress PE2 Z Y X PE1 egress PE1-push L4 (PE1,L4) (PE1,L3) (PE1,L2) swap (L4, L3) swap (L3, L2) (PE1,L1) swap (L2, L1) Y will do similar processing. The LSP setup proceeds from egress to ingress. X evaluates whether PE1 is on the IGP shortest path for that FEC. If successful X assigns L2 for FEC PE1, installs forwarding state swapping L2 and L1 and advermses a binding for L2 and FEC PE1 to Y. PE1 assigns a Label (L1) to its own L0 address (FEC) and advermses that to its LDP peer X Label actions: Push to the stack, Swap top label, PoP from the stack; S-bit is set to 1 in MPLS header if label is last label on the stack;

25 A 10 B C F D E 5 5 G 28 = label 5 = IGP metric Label 3 is announced by router B to its neighbor, 3 is a special value, Implicit NULL label [RFC-3032]; This triggers Penultimate Hop Popping (PHP) the LSR (E) before the LER (B) pops the label and forwards normal IP packet to LER (B); simplifies processing at LER (saves one lookup); default behaviour of most implementations, not mandatory;

26 LSR A receives mapping for Label N for FEC X from peer LSR B; LSR A will use Label N for forwarding if and only if B is on the IGP shortest path for destination X from A s point of view; OR: LSPs set up via LDP follow the IGP shortest path and LDP uses IGP to avoid loops; LSPs shift with IGP path changes; - Danger of blackholing/looping during reconvergence; But who/what assigns the labels? Goal is to build a forwarding table with mapping between incoming label and outgoing label; Routers pick the label values (local significance only) The MPLS architecture uses downstream label assignment: router expects to receive the traffic with label it picked locally. Called downstream because label assigned to traffic at point X was picked by a router who is one hop further down in the direction of the traffic flow from X.

27 Multi-Protocol Label Switching Summary MPLS - set up virtual circuits across switched architecture Paths are called LSPs LSPs are unidirectional Traditional MPLS follows the traditional routing (shortest path) Every router assigns label to every prefix in its routing table (simplified, depends on implementation, independent or ordered label distribution control) Label is local to the router LER = Label Edge Router (or PE = Provider Edge router) LSR = Label Switching Router (or P = Provider router) FEC = Forwarding Equivalence Class LIB = Label Information Base

28 Topology Database LDP Database OSPF/IS-IS BGP LDP Routing table (inet.0) RIB Routing Information Base FEC mapping table (inet.3) LIB Label Information Base (mpls.0) Label Routing table FIB Forwarding Information Base LFIB Label Forwarding Information Base Forwarding table Juniper specific See also:

29 Multi-Protocol Label Switching Summary With MPLS, routers (LSRs) never look at IP addresses, only at labels you can encapsulate anything within MPLS (eg carry IPv6 traffic over IPv4 network) In MPLS: switching traffic based on labels advertised by LDP (or RSVP, BGP) In IP routing: routing based on destination address for which the entries in the routing table have to exist (configured statically or installed by routing protocol) (Traditional) IP routing: Bound to the IP topology, that's what comes from the routing protocols With MPLS: You can create (with use of different labels and label stacks) different topologies & services (MPLS-TE, MPLS VPNs)

30 Label Distribution - Control Plane

31 Label Distribution - Control Plane How are bindings between labels and FECs distributed through network? You need routing and signalling; Manual configuration not an option, need protocol; 2 options: invent new protocol or extend existing protocol to carry labels; Both were done: New protocol: LDP Label Distribution Protocol Two existing protocols: RSVP and BGP

32 LDP - made by IETF [RFC-5036] Fundamental concept in MPLS is that 2 LSRs must agree on the meaning of labels used to forward traffic. Protocol used where one LSR informs another of the LABEL BINDINGS it has made. Specifically designed for label distribution - does nothing else but that, no routing, in fact it relies on an IGP for all routing decisions; UDP discovery and TCP session with peers; Adjacent LSRs inform each other of the label bindings; An IGP protocol is configured on all LSRs; New IGP routes (prefixed in routing table) lead to new label bindings; Labels can be withdrawn when IGP routes are no longer valid; Hard-state; Expected to work until explicitly torn down

33 Label Distribution - Control Plane LDP works between directly connected neighbors or peers; Peers are automatically discovered (via multicast to well-known UDP port); Initialization: exchange information regarding features and modes supported; Next: information regarding binding Labels and FECs exchanged; After discovery a TCP session is established and LDP session is set up; [why chosen to use TCP? Reliable delivery and incremental updates, not periodic refreshes] To keep session up keepalive messages are sent. Label messages: advertise new labels, withdraw labels

34 RSVP-TE Resource ReserVation Protocol RSVP was developed before MPLS; To create bandwidth reservations for individual traffic flows in network as part of the int-serv model; Its mechanism is to reserve bandwidth along each hop of a network for an end-to-end session; Doesn't scale (create, maintain, tear-down state for each traffic flow!), so it is not/hardly used. RSVP extensions for MPLS to create and maintain LSPs and to create associated bandwidth reservations [RFC-3209]; Better scaling (single LSP can carry all traffic between ingress and egress router pair, not per flow);

35 RSVP-TE RSVP-signaled LSP does not necessarily follow IGP shortest path; Extensions allow for explicit routing (specify entire path or specific transit nodes) Explicitly Routed LSP: An LSP whose path is established by a means other than normal IP routing. Bandwidth reservation is optional Creation of RSVP-signaled LSP is initiated by ingress router by sending RSVP Path message; Destination is the egress router; Transit routers inspect the message and make modifications (define label, check and reserve bandwidth); Path message: label request object, Explicit Route Object (ERO), Record Route Object, Sender Tspec ERO contains addresses of nodes through which the LSP must pass;

36 RSVP-TE in response the egress router sends an RSVP Resv message, this follows the reverse path back to ingress; establishes the LSP (send label in Resv message); With Record Route Object Resv message: label object, Record Route Object routers can check if the path is loop-free Path and Resv messages travel hop-by-hop through network - establish state at each node; Periodic exchange of messages after establishment to refresh the state (if missed LSP is torn down); RSVP-signaled LSPs follow single path from ingress to egress (even in case of multiple available paths); LSP still unidirectional!

37 MP-BGP Multiprotocol Extensions to BGP It supports multiple address families, easy to define and carry new types of reachability information and associated attributes; Advertise prefix and label(s) associated with it; [RFC-3107] Carrying Label Information in BGP-4: The label mapping information for a particular route is piggybacked in the same BGP Update message that is used to distribute the route itself. Can be used inter-domain (between AS-es BGP is used); Often BGP is already used so no need for another protocol; This is used for Layer3 VPN between sites interconnected by MPLS (provider) core network; Each VPN has its own VRF (Virtual Routing and Forwarding instance); MPLS forwarding uses stacked labels: outer label for LSP forwarding inner label to differentiate between different VPNs

38 LSP FEC LABEL TAG OSPF-TE ISIS-TE Now back to some friends of MPLS LDP RSVP-TE MP-BGP MPLS-TE MPLS-TP T-MPLS VPLS H-VPLS MPƛS GMPLS PBB-TE PCEP SR Due to Moore s Law lookup speed is no longer the biggest problem, but since 1997 a lot of new ways to use MPLS and Family have been found...

39 Traffic engineering What is it? Process of manipulating traffic on an (IP) network to make better use of capacity; Is not network engineering, but linked; Reduce overall cost of operations by more efficient use of bandwidth resources; With traditional IP and IP routing protocols difficult: tweaking link cost or weight to influence IGP behaviour. availability of resources (e.g. bandwidth) not taken into account. IGPs distribute network topology information through network; Can be used to calculate the routes of LSP automatically; When required to establish LSPs not following IGP routes, with guaranteed QoS characteristics and backup LSPs that avoid single points of failure you need more: Traffic Engineering extensions -TE RFC-3272: Overview and Principles of Internet Traffic Engineering

40 Traffic engineering Cost optimisation (better utilisation of network resources); congestion management; dynamic services & traffic profiles; Efficient routing (predictable, deterministic paths); Availability/ resilience / fast restoration; QoS / separate realtime latency-critical services from other traffic; MORE CONTROL MPLS-TE: set of extensions to MPLS explicit or constraint based routing; use RSVP-TE to set up explicit paths; bandwidth reservation;

41 MPLS-TE use RSVP-TE to set up explicit paths; bandwidth requirement; RSVP sets up TE-tunnel (VC) to endpoint using Path and Resv messages, reserving the bandwidth on path. explicit or constraint based routing; If no external NMS is used routers need to figure out where BW is available -TE extensions to OSPF (new LSA) and ISIS (new TLV) make it possible to tell how much BW available on every link. Router does a Constrained SPF (CSPF) calculation. Initiated by config on router interface or external third party application (NMS/PCEP) More extensions: fast rerouting, restoration, QoS, Shared Risk Link Groups, link coloring, make-before-break, pre-emption, auto-bandwidth, etc..

42 MPLS-TP Transport Profile In 2006 the ITU-T started with its own MPLS-like technology: T-MPLS or Transport MPLS Continued as joint effort of IETF together with ITU; Now called: MPLS-TP; MPLS-TP is set of extensions to IP MPLS feature set that fulfills packet transport requirements; MPLS MPLS-TP is the subset of MPLS MPLS-TP MPLS extensions tailored for transport networks; + Added functionality based on Transport requirements

43 MPLS-TP Transport Profile Result of the joint effort had a list of 115 requirements [RFC5654/2009] Some differences between MPLS and MPLS-TP: MPLS-TP is bi-directional LSPs (follow same path both ways) No LSP merging, no PHP No ECMP (Equal-cost multi-path routing) Explicit support for ring topologies Static configuration via Network Management System (NMS) or using control plane MPLS-TP does not assume IP connectivity between devices Differences are mainly in the control plane, management, and OAM MPLS-TP applies additional constraints, eliminates some complex functions that make networks unpredictable and non-deterministic

44 VPLS Virtual Private LAN Service A VPLS is (provider) service that emulates the full functionality of traditional LAN a Layer 2 Virtual Private Network (VPN) VPLS is "private" in that CE devices that belong to different VPLSs cannot interact VPLS is "virtual" in that multiple VPLSs can be offered over a common packet switched network (over IP or MPLS network) PPVPN Layer 2 Layer 3 P2P P2M PE-based CE-based (VPWS) VPLS IPLS BGP/MPLS Virtual IPsec IP VPNs Router [RFC-4026: Provider Provisioned Virtual Private Network (VPN) Terminology]

45 VPLS Virtual Private LAN Service Ethernet service: frames sent to broadcast addresses and unknown dest MAC addresses are flooded to all ports; all unknown unicast, broadcast and multicast frames are flooded over the corresponding PWs to all PE nodes participating in the VPLS Responsibility of service provider to create loop-free topology; Full-mesh of pseudo-wires connecting the edge sites; Using LDP for Signaling [RFC-5461] Using BGP for Auto-Discovery and Signaling [RFC-5462] Number of limitations in redundancy, multicast, multihoming, provisioning simplicity New RFC on defining the requirements for a new solution: Ethernet VPN (EVPN) [RFC-7290 (May 2014)]

46 Segment Routing (SR) Alternative for LDP and RSVP-TE in MPLS networks; Segment routing is a new forwarding paradigm that provides source routing; The source can define the path a packet will take; SR uses MPLS to forward packets (in IPv4), but labels are distributed by IGP (OSPF, ISIS); Source Routing the source chooses a path and encodes it in the packet header as an ordered list of segments the rest of the network executes the encoded instructions Segment: an identifier for any type of instruction forwarding or service Example segments: particular node in network, network link, prefix Source Routing: a node steers packets through an ordered list of instructions (segments) that are encoded in the packet header

47 Segment Routing (SR) Segment Routing Forwarding Plane MPLS: ordered list of segments is represented as a stack of labels Segment Routing re-uses MPLS data plane without any change Segment is encoded as MPLS label Applicable to IPv4 and IPv6 address families IPv6: an ordered list of segments is encoded in a routing extension header SR allows to enforce a flow through any topological path and service chain while maintaining per-flow state only at the ingress node to the SR domain; Application for SR enable some kind of application controller that can steer traffic over different paths, depending on different requirements and the current state of the network; Doing this now requires MPLS-TE, as well as keeping state in many device, with SR, there is no need to keep state in intermediary devices The state is in the packet, not in the network

48 Segment Routing (SR) segment routing with central optimization (PCE - path computation element) 9105 Next Header Length Type Segments Left First Segment Flags Reserved Segment A B C D Z Segment 1 Segment 2 Segment n (1) N O P Optional Type Length Value Objects (variable) routing along any explicit path segment encoded as IPv6 address ordered list of segments is encoded as ordered list of IPv6 addresses in the routing header active segment is indicated by the Destination Address of the packet next active segment is indicated by a pointer in the routing header (1) Picture from: See for more info: IETF SPRING WG and

49 MPLS et cetera MPLS over 15 years old but Still lot of activities on standardisation; Lot of new activities using or extending MPLS Just look at the number of RFCs after RFC-3013 (> 130) and the drafts in the MPLS WG of the IETF: Skipped MPLS VPNs, GMPLS, seamless MPLS, Pseudo Wires, and much more.

50 Thanks for your attention! Questions?