Title: Deploying MPLS
1Deploying MPLS
Ross Chandler
2MPLS Multiprotocol Label Switching
- Context Packet Switched Networks
- E.g. IP Networks, The Internet.
- Frame Relay ATM too.
- Data is chopped up into frames or cells and
passed from router to router along a stable path
based on info in an attached header. -
- The header is added by the sending host.
- The packets arrive and are forwarded
intermittently (asynchronously).
3What is MPLS? Original Motivation
- Consider the Internet
- A typical service provider core router will have
about 150,000 prefixes path information in its
routing tables. - Each prefix is announced by one of about 18314
routing domains - BGP Autonomous Systems. - Core Router Default-free router. It knows how
to get everywhere on the Internet.
4What is MPLS? Original Motiviation
- Any individual router in this network will only
have a comparatively small number of
interfaces/ports linking it to other routers. - E.g. A router with 6 interfaces.
- When a packet arrives at this router it can
continue its journey by going 6 ways (it could be
redirected out the port it came in on). - But to make its forwarding decision it has to
look at a forwarding table with 150,000 prefixes.
5What is MPLS? Original Motiviation
- This is where the Labels come in.
- If the router agrees labels to use with its
neighbouring routers and attaches these to
packets entering and leaving its ports (label
swapping) then the forwarding decision process
only has to look through a much shorter
label-based routing table. - The label values used have local significance
only so they are only 20 bits in size.
6What is MPLS? Original Motiviation
There used to be a speed advantage to doing
things this way but now it is just as fast to
look up against the full global routing table
because routers have ASICs built to do this!
MPLS lives on though because of its other
applications.
7What is MPLS? Terminology
Labels are inserted between L3 header and L2
header
Label Switch Router
Label Edge Router
8What is MPLS? Terminolgy
- Label Edge Routers convert IP packets into
Labeled IP Packets -
There are only a limited number of possible paths
through any network. Routers along these paths
agree the labels that will be used by packets
traveling along them. The paths are called LSRs
Label Switched Paths. The destination prefixes in
each LSR belong to a FEC forwarding equivalence
class. A label corresponds to each FEC.
Label distribution protocol (LDP) running between
the routers distributes the Labels to use for
each FEC. Labels can only have local significance
although the routers will try to agree the same
label along a path to make it easier for people
to follow.
9Example Label FIB
- edge5-crash tag-switching forwarding-table
- Local Outgoing Prefix Bytes
tag Outgoing Next Hop - tag tag or VC or Tunnel Id switched
interface - 16 88 62.77.188.209/32 0
Gi0/2 159.134.125.34 - 88 62.77.188.209/32 0
Gi0/1 159.134.125.2 - 17 20 159.134.191.48/28 0
Gi0/1 159.134.125.2 - 20 159.134.191.48/28 0
Gi0/2 159.134.125.34 - 19 17 159.134.124.160/27 0
Gi0/1 159.134.125.2 - 17 159.134.124.160/27 0
Gi0/2 159.134.125.34 - 166 159.134.124.160/27 0
Gi0/1 159.134.125.1 - 166 159.134.124.160/27 0
Gi0/2 159.134.125.33 - 44 Pop tag 159.134.68.80/30 0
Gi0/1 159.134.125.2 - Pop tag 159.134.68.80/30 0
Gi0/2 159.134.125.34 - 132 Untagged 159.134.127.151/32 0
Gi0/1 159.134.125.6 - Untagged 159.134.127.151/32 0
Gi0/2 159.134.125.59 - 163 Untagged 192.168.3.0/24V 0
Se6/22 point2point - 1177 Untagged 0.0.0.0/0V
222945684 Se6/13 point2point - 689 Untagged 0.0.0.0/0 0
Gi0/1 159.134.125.2 - Untagged 0.0.0.0/0 0
Gi0/1 159.134.125.1
10MPLS Label Distribution Protocol
- The router at the Head-end of an LSP picks a
label for each FEC that it pops labeled IP
packets for. - Poping a labelled packet returns it to being an
IP packet and it then continues its journey via
IP switching. It is an LER for the FEC. - If a router isnt popping a packet it swaps a new
label on before forwarding the packet. It is an
LSR for the FEC. -
11MPLS Traffic Engineering QoS
- By default the LSPs are the same as the paths
through the network that the Interior Rounting
protocol (OSPF or IS-IS) would select. - Additional contraints can be applied to set-up
LSPs that route traffic along a different path. - e.g. IP packets with their precedence bits set to
indicate high priority might belong to a FEC that
has a LSP with a higher priority for bandwidth. - e.g. An LSP can be set-up to use multiple paths
between locations to fully utilise the links and
spread out the traffic load. -
12Label Stacking
- IP packets can have more than one label attached
to them. Extra Labels are popped or pushed onto a
label stack at LERs . New services are provided
using label stacking. - e.g. VPNs - both L3 (IP) and L2 (Ethernet).
- IPv6 packets may be labelled at a
dual-stack IPv4/v6 LER and swapped across an IPv4
core - no need for a separate buggy IPv6 Interior
Gateway Protocol.
13RFC2547bis Layer 3 IP VPNs
- This is the most common service provided using
MPLS. It is a peer-to-peer L3 VPN. .I.e. Customer
router talks an IP routing protocol to the
provider router. - Customer routers are connected to the Provider
Edge routers where their packets arrive into
separate VRFs on the PEs. Here the packets are
labelled according to the distination prefix in
the customer network. The packets are then
labelled again to put them into an LSP that will
carry them to the PE router connecting to the CE
with the destination prefix. The labels are
popped at the remote end.
14RFC2547bis Layer 3 IP VPNs
- The Inner Labels to use are distributed between
PE routers using Multiprotocol BGP. - MP-BGP is used in the same way to communicate the
labels used to label swap IPv6 packets across and
IPv4 only core. It is also used by Kompella Layer
2 (Ethernet) VPNs. - MP-BGP is so useful it might out-last MPLS!
- Balance L2TPv3 is being used as an alternative
to MPLS for RFC2547 VPNs by some providers.
15IPv6 over MPLS
- IPv6 Packets are labelled and swapped across core
as they are with IPv4 - Two methods
- Native IPv6 over MPLS
- or
- 6PE where IPv4/MPLS core infrastructure is
IPv6-unaware.
16IPv6 over MPLS
- 6PE
- PEs are IPv4/IPv6 dual-stacked and support 6PE.
- IPv6 reachability info exchanged between 6PEs
using MP-iBGP. - Advantage No need to run a separate IPv6 IGP on
core.
17Example Commands for Cisco
- Global Conf.
- tag-switching tdp router-id Loopback0
- OR
- mpls ldp router-id Loopback0
- interface POS0/0
- description STM4-c to coreb.cra.dublin
- bandwidth 622000
- ip address 159.134.191.234 255.255.255.252
- tag-switching ip OR mpls ip
- pos framing sdh
- pos scramble-atm
- pos flag s1s0 2
18Handling of routes on Provider Networks.
- Interior Routing Protocol (e.g. OSPF or IS-IS)
- Manages reachability for internal network
elements. - Router Loopback Interfaces
- Internal Link Addresses
- Objective is to minimise number of entries in
the database so that the SPF algorithm is run as
little as possible.
19Handling of routes on Provider Networks.
- Exterior Routing Protocol (BGP4)
- Manages reachability for everything else outside
the core network. - All customer blocks, pool addresses, hosting
LANs. - Including dial/DSL users.
- BGP scales much better for large numbers of
prefixes.
20Possible Future GMPLS - Generalised MPLS
- With conventional MPLS the Label Distribution
control plane messages (TCP and UDP IP packets)
share the data plane with the labelled traffic. - GMPLS supports TDM (e.g. SDH), wavelength
- (? /DWDM), and spatial (fibre switching)
extending control done to the transport (fibre)
network. - A single control plane for all layers of the
network.
21References
- MPLS Forum http//mplsforum.org/
- The CIDR Report http//bgp.potaroo.net/cidr/
- MPLS Notes http//www.opalsoft.net/qos/MPLS.h
tm - RFC 2283 Multiprotocol Extensions for BGP4
- RFC 2547 BGP/MPLS VPNs
- IPv6 over MPLS Deployment http//aristote1.aristo
te.asso.fr/Presentations/IPv62002/P/Lourdelet-MPLS
/6pe-renater.pdf -
22The End