1. ATM and Multi-Protocol Label Switching (MPLS)By
Behzad Akbari
Spring 2011
These slides are based in parts on the slides of J. Kurose (UMASS) and Shivkumar (RPI)

2. Outline ATM basics IP over ATM MPLS basics MPLS VPN
MPLS traffic engineering

3. Asynchronous Transfer Mode: ATM
1990’s/00 standard for high-speed (155Mbps to 622 Mbps and higher) Broadband Integrated Service Digital Network architecture
Goal: integrated, end-end transport of carry voice, video, data
meeting timing/QoS requirements of voice, video (versus Internet best-effort model)
“next generation” telephony: technical roots in telephone world
packet-switching (fixed length packets, called “cells”) using virtual circuits

4. ATM architecture adaptation layer: only at edge of ATM network
physical
ATM
AAL
end system
switch
adaptation layer: only at edge of ATM network
data segmentation/reassembly
roughly analagous to Internet transport layer
ATM layer: “network” layer
cell switching, routing
physical layer

5. ATM: network or link layer?
Vision: end-to-end transport: “ATM from desktop to desktop”
ATM is a network technology
Reality: used to connect IP backbone routers
“IP over ATM”
ATM as switched link layer, connecting IP routers IP
network
ATM
network

6. ATM Adaptation Layer (AAL)
ATM Adaptation Layer (AAL): “adapts” upper layers (IP or native ATM applications) to ATM layer below
AAL present only in end systems, not in switches
AAL layer segment (header/trailer fields, data) fragmented across multiple ATM cells
analogy: TCP segment in many IP packets
physical
ATM
AAL
end system
switch

7. ATM Adaptation Layer (AAL) [more]
Different versions of AAL layers, depending on ATM service class:
AAL1: for CBR (Constant Bit Rate) services, e.g. circuit emulation
AAL2: for VBR (Variable Bit Rate) services, e.g., MPEG video
AAL5: for data (eg, IP datagrams)
User data
AAL PDU
ATM cell

8. ATM Layer Service: transport cells across ATM network
analogous to IP network layer
very different services than IP network layer
Guarantees ?
Network
Architecture
Internet
ATM
Service
Model
best effort
CBR
VBR
ABR
UBR
Congestion
feedback
no (inferred
via loss) no
congestion
yes
Bandwidth
none
constant
rate
guaranteed
minimum
Loss no
yes
Order no
yes
Timing no
yes

9. ATM Layer: Virtual Circuits
VC transport: cells carried on VC from source to dest
call setup, teardown for each call before data can flow
each packet carries VC identifier (not destination ID)
every switch on source-dest path maintain “state” for each passing connection
link,switch resources (bandwidth, buffers) may be allocated to VC: to get circuit-like perf.
Permanent VCs (PVCs)
long lasting connections
typically: “permanent” route between to IP routers
Switched VCs (SVC):
dynamically set up on per-call basis

10. ATM VCs Advantages of ATM VC approach:
QoS performance guarantee for connection mapped to VC (bandwidth, delay, delay jitter)
Drawbacks of ATM VC approach:
Inefficient support of datagram traffic
one PVC between each source/dest pair) does not scale (N*2 connections needed)
SVC introduces call setup latency, processing overhead for short lived connections

11. ATM Layer: ATM cell 5-byte ATM cell header 48-byte payload
Why?: small payload -> short cell-creation delay for digitized voice
halfway between 32 and 64 (compromise!)
Cell header
Cell format

12. ATM cell header VCI: virtual channel ID
will change from link to link thru net
PT: Payload type (e.g. RM cell versus data cell)
CLP: Cell Loss Priority bit
CLP = 1 implies low priority cell, can be discarded if congestion
HEC: Header Error Checksum
cyclic redundancy check

13. ATM Physical Layer (more)
Two pieces (sublayers) of physical layer:
Transmission Convergence Sublayer (TCS): adapts ATM layer above to PMD sublayer below
Physical Medium Dependent: depends on physical medium being used
TCS Functions:
Header checksum generation: 8 bits CRC
Cell delineation
With “unstructured” PMD sublayer, transmission of idle cells when no data cells to send

14. ATM Physical Layer Physical Medium Dependent (PMD) sublayer
SONET/SDH: transmission frame structure (like a container carrying bits);
bit synchronization;
bandwidth partitions (TDM);
several speeds: OC3 = Mbps; OC12 = Mbps; OC48 = 2.45 Gbps, OC192 = 9.6 Gbps
TI/T3: transmission frame structure (old telephone hierarchy): 1.5 Mbps/ 45 Mbps
unstructured: just cells (busy/idle)

15. IP-Over-ATM IP over ATM
replace “network” (e.g., LAN segment) with ATM network
ATM addresses, IP addresses
Classic IP only
3 “networks” (e.g., LAN segments)
MAC (802.3) and IP addresses
ATM
network
Ethernet
LANs
Ethernet
LANs

16. IP-Over-ATM
AAL
ATM
phy
Eth IP
app
transport

17. Datagram Journey in IP-over-ATM Network
at Source Host:
IP layer maps between IP, ATM dest address (using ARP)
passes datagram to AAL5
AAL5 encapsulates data, segments cells, passes to ATM layer
ATM network: moves cell along VC to destination
at Destination Host:
AAL5 reassembles cells into original datagram
if CRC OK, datagram is passed to IP

18. IP-Over-ATM Issues: IP datagrams into ATM AAL5 PDUs
from IP addresses to ATM addresses
just like IP addresses to MAC addresses!
ATM
network
Ethernet
LANs

19. Re-examining Basics: Routing vs Switching

20. IP Routing vs IP Switching

21. MPLS: Best of Both Worlds
PACKET ROUTING
CIRCUIT SWITCHING
HYBRID IP
MPLS+IP
ATM
TDM
Caveat: one cares about combining the best of both worlds
only for large ISP networks that need both features!
Note: the “hybrid” also happens to be a solution that bypasses IP-over-ATM mapping woes!

22. History: Ipsilon’s IP Switching: Concept
Hybrid: IP routing (control plane) +
ATM switching (data plane)

23. Ipsilon’s IP Switching
ATM VCs setup when new IP “flows” seen, I.e.,
“data-driven” VC setup

24. Issues with Ipsilon’s IP switching

25. Tag Switching Key difference: tags can be setup in the background
using IP routing protocols (I.e. control-driven VC setup)

26. Multi-Protocol Label Switching (MPLS)

27. Background It was meant to improve routing performance on the Internet
Routing is difficult using CIDR (longest prefix matching)
Using the label-swapping paradigm to optimize network performance
MPLS is similar to virtual circuits
Only a fixed-sized label is used (like a VCID) with local scope
It is very datagram oriented though
It uses IP addressing and IP routing protocols

28. Goals of MPLS
To enable IP capability on devices that cannot handle IP traffic
Making cell switches behave as routers
Increased performance
Using the label-swapping paradigm to optimize network performance
Forward packets along “explicit routes” (pre-calculated routes not used in “regular” routing)
MPLS also permits explicit backbone routing, which specifies in advance the hops that a packet will take across the network.
This should allow more deterministic, or predictable, performance that can be used to guarantee QoS
To support certain virtual private network services

29. IP Regular Destination Based Forwarding
Address
Prefix
Address
Prefix
Address
Prefix
I/F
I/F
I/F
128.89 1
128.89
128.89
171.69 1
171.69 1 … … … …
128.89 1
Data
Now when a packet comes into a router a look up is done based on the IP address in the packet, at match is obtained, and the packet is forwarded out the appropriate interface.
The packet follows the same process on a hop-by-hop bases through the network until it reaches its destination.
Data 1
Data
Data
Packets Forwarded Based on IP Address
171.69

30. MPLS Example: Routing Information
Label
Address
Prefix
Out
I’face
Out Label In
Label
Address
Prefix
Out
I’face
Out Label In
Label
Address
Prefix
Out
I’face
Out Label
128.89 1
128.89
128.89
171.69 1
171.69 1 … … … … … …
128.89 1
You Can Reach Thru Me
Tag edge routers and tag switches use standard IP routing protocols to identify routes through the network. Theses fully interoperate with non-tag switching routers
So what tag switching does is it extends the forwarding table by adding a tag field. One for the incoming tag and one for the outgoing tag. Note the topology of the network is discovered using unmodified layer 3 protocols such as OSPF.
You Can Reach and Thru Me 1
Routing Updates (OSPF, EIGRP, …)
You Can Reach Thru Me
171.69

31. Labels for Destination-Based Forwarding
A label is allocated for each prefix in its table
The label is chosen locally
Think of them as indices into the routing table
Router advertises this to its neighbors
“label distribution protocol” (LDP)
Packets addressed to the prefix should, for efficiency, be tagged with the label.
The label of an incoming packet is “swapped” before being forwarded to the next router.

32. MPLS Example: Assigning Labels
Address
Prefix
Out
I’face
Out Label In
Label
Address
Prefix
Out
I’face
Out Label In
Label
Address
Prefix
Out
I’face
Out Label -
128.89 1 4 4
128.89 9 9
128.89 - -
171.69 1 5 5
171.69 1 7 … … … … … … … … … … … …
128.89 1
Tag routers and switches use the tables generated by the standard routing protocols to assign and distribute tag information via the tag distribution protocol (TDP). Tag routers receive the TDP information and build a forwarding database, which makes use of the tags.
TDP is then used to bind tags to routes and distribute this information to each routers upstream neighbor.
Use Label 9 for
Use Label 4 for and Use Label 5 for 1
Label Distribution Protocol (LDP)
(downstream allocation)
171.69
Use Label 7 for

33. MPLS Example: Forwarding Packets
Label
Address
Prefix
Out
I’face
Out Label In
Label
Address
Prefix
Out
I’face
Out Label In
Label
Address
Prefix
Out
I’face
Out Label -
128.89 1 4 4
128.89 9 9
128.89 - -
171.69 1 5 5
171.69 1 7 … … … … … … … … … … … …
128.89 1
Data
When we get to the first router, the one performing the tag imposition, there’s an IP look-up based on the IP prefix. It finds the forwarding table entry and it discovers that to get to the destination it should use tag x. It sticks that tag on the front of the packet and forwards it along to the next hop tag switch.
At this point the router can just do pure tag forwarding, gets in the packet with tag x, figures our that the outgoing interface is y, and the outgoing tag replaces the incoming tag. Note packet is forwarded based solely on the tag without re-analyzing the network layer header. This provides the essential separation of routing and forwarding referred to earlier.
The packet reaches the tag edge router at the egress point of the network ,where the tag is stripped off and the packet delivered. 9
Data 1
Data 4
Data
Label Switch Forwards Based on Label

34. MPLS Operation
1a. Existing routing protocols (e.g. OSPF, IS-IS) establish reachability to destination networks.
4. Edge LSR at egress removes(POP) label and delivers packet.
1b. Label Distribution Protocol (LDP)
establishes label to destination network mappings.
Here is how MPLS actually works.
Step 1A: a routing protocol such as OSPF, EIGRP, or IS-IS determines the layer 3 topology. A router builds a routing table as it “listens” to the network. A Cisco router or IP+ATM switch can have a routing function inside that does this. All devices in the network are building the layer 3 topology.
Step 1B: The Label Distribution Protocol establishes label values for each device according to the routing topology, to pre-configure maps to destination points. Unlike ATM PVCs where the VPI/VCIs are manually assigned, labels are assigned automatically by LDP.
Step 2: An ingress packet enters the Edge LSR. The LSR labels it, does all the layer 3 value-added services, including QoS, Bandwidth management, and so forth. It then applies a label to it based on the information in the forwarding tables. (This also reflects QoS, which we’ll discuss in detail in the next section).
Step 3: the core LSR read the labels on each packet on the ingress interface, and based on what the label says, sends the packet out the appropriate egress interface with a new label.
Step 4: the egress Edge LSR strips the label and sends the packet to its destination.
2. Ingress Edge LSR receives packet, performs Layer 3 value-added services, and labels(PUSH) packets.
3. LSR switches packets using label swapping(SWAP) .

35. Remarks Rather than longest prefix-matching we use label matching
Labels can be very efficient, simply an index into the routing table
Regular IP routing is still used
E.g., we could use OSPF to determine the routes
Then we use labels for efficiency in per-hop routing
Note that a “Setup” phase (like in VC’s) is not used

36. Placement of “labels”
For Ethernet, the “protocol number used” is 0x8847 for MPLS
I.e., the “protocol number” of IP is not used.
Thus, IP never sees the message!

37. Label Header Header= 4 bytes, Label = 20 bits.
Label
EXP S
TTL
Label = 20 bits EXP = Class of Service, 3 bits
S = Bottom of Stack, 1 bit TTL = Time to Live, 8 bits
Header= 4 bytes, Label = 20 bits.
Can be used over Ethernet, 802.3, or PPP links
Contains everything needed at forwarding time
The tag frame encapsulation uses what’s called a shim header. It’s a header that sits between the MAC layer header and the layer 3 header in the packet. It consists of one or more entries that look like this.
Its a 32-bit word per entry which contains:
the tag your forwarding on
a 3-bit class of service field
an 8-bit time to live field, and
a 1-bit end of stack.
The end of stack is what allows you to determine when you’re popping a tag, whether this is the last tag on the packet, or whether there are further tags.

38. Some Definitions
Forwarding Equivalence Class (FEC): a group of IP packets which are forwarded in the same manner (e.g., over the same path, with the same forwarding treatment)
Labeled Switched Router (LSR): A router capable of supporting MPLS labels.
Labeled Switched Path: a sequence of LSR’s so that data can traverse the entire path using labels.

39. Traffic Aggregates: Forwarding Equivalence Classes
Packets are destined for different address prefixes, but can be
mapped to common path
IP1
IP2
LSR
LER
LSP
#L1
#L2
#L3
The “Forwarding Equivalence Class” is an important concept in MPLS. An FEC is any subset of packets that are treated the same way by a router. By “treated” this can mean, forwarded out the same interface with the same next hop and label. It can also mean given the same class of service, output on same queue, given same drop preference, and any other option available to the network operator.
When a packet enters the MPLS network at the ingress node, the packet is mapped into an FEC. The mapping can also be done on a wide variety of parameters, address prefix (or host), source/destination address pair, or ingress interface. This greater flexibility adds functionality to MPLS that is not available in traditional IP routing.
FECs also allow for greater scalability in MPLS. In Ipsilon’s implementation of IP Switching or in MPOA, their equivalent to an FEC maps to a data flow (source/destination address pair, or source/destination address plus port no.). The limited flexibility and large numbers of (short lived) flows in the Internet limits the applicability of both IP Switching and MPOA. With MPLS, the aggregation of flows into FECs of variable granularity provides scalability that meets the demands of the public Internet as well as enterprise applications.
In the current Label Distribution Protocol specification, only three types of FECs are specified:
- IP Address Prefix
- Router ID
- Flow (port, dest-addr, src-addr etc.)
The spec. states that new elements can be added as required.
FEC = “A subset of packets that are all treated the same way by a router”
The concept of FECs provides for a great deal of flexibility and scalability
In conventional routing, a packet is assigned to a FEC at each hop (i.e. L3 look-up), in MPLS it is only done once at the network ingress

40. Label Switched Path (LSP)
47.1
47.2
47.3 1 2 3 IP

41. Label Merging
When multiple input streams corresponding to the same FEC exit using the same MPLS label.
InLabel NextHop Label
Port
Port
Netw D
Dest NextHop Label
D Port R2 R4 R1
Port 3
Port 1
Port 5 R3
Dest NextHop Label
D Port

42. Non-Label Merging
Each source-destination pair has its own label at each LSR router.
InLabel NextHop Label
Port
Port
Netw D
Dest NextHop Label
D Port R2 R4 R1
Port 3
Port 1
Port 5 R3
Dest NextHop Label
D Port

43. Pushing-Requesting Labels
R2 can “push” a label to R1, indicating which label to use to reach D
R1 can “request” a label from R2 to be used to reach D.
If using non-merging, usually R1 requests a label from R2
Netw D R2 R4 R1

44. ATM Most importantly, we can use ATM switches for IP
We can turn “ATM Cell switches” into “label switching routers” usually only by changing the software and not the hardware of the switch.

45. IP over ATM (Before MPLS)
We had every router with a VC over an ATM network to every other router
Known as an “overlay” network
Whole ATM network looked like a single “subnet” to the IP Routers
ATM switches are not aware that the payload is an IP packet

46. IP disassembly into ATM cells
IP becomes an “application” to the ATM layer.
IP packets have to be broken into small 48-byte pieces, and placed into ATM Cells
Cells are sent over the ATM circuit (e.g. from R1 to R6), the switches only see ATM Cells, not IP packet
At R6, the cells are regrouped and the IP packet restored

47. ATM switches as LSRs (using MPLS)
ATM switches are now “peers” of MPLS routers
No longer viewed as a single subnet, each link is now a subnet

48. Advantages of MPLS vs overlay
Each MPLS router has fewer “adjacencies” (i.e. neighbors)
This reduces the OSPF traffic to the router significantly
In OSPF you receive the topology of the entire network via each of your neighbors.
Each router now has a view of the entire topology
Not possible in overlay networks (ATM network “black box”)
Routers have better control of paths in case of link failures
In overlay networks, the ATM switches would do the rerouting
ATM switches may still support native ATM if desired.

49. How to route IP packets? Can we send IP messages to our neighbors?
We can use a special VCID (say 0) to send IP messages to our neighbor.
Each node has a VCID 0 with each of its neighbors (a “single hop” VCID
Thus, to send an IP message to a neighbor
Disassemble the IP packet into ATM Cells
Send them on VCID 0 of the link of the desired neighbor
The neighbor reassembles the IP packet
Since we can send an IP message to any neighbor
This implies ATM LSR’s can execute ANY Internet protocol based on IP (e.g., OSPF, RIP, etc) and forward IP datagrams

50. End-to-end VC’s Disassembly/reassembly at each hop is wasteful
It is better to establish an e-2-e VC for each source/destination pair, e.g., from R1 to R6
From OSPF (or other mechanism), each router knows which other router is ATM or regular router
R1 “requests” a label from LSR1 for destination R6
LSR1 requests a label from LSR3 for destination R6
LSR3 requests a label from R6

51. Explicit Routing Similar to “source routing” but done by a router
“Fish” network due to its shape
R1 -> R7 : R1 R3 R6 R7
R2 -> R7 : R2 R3 R4 R5 R7
Perhaps we want to balance the load somehow
Cannot be done with regular IP
IP routing does not look at the source of the message

52. Explicitly Routed (ER-) LSP
#216
#14
#462
ER-LSP follows route that source chooses. In other words, the control message to establish the LSP (label request) is source routed.
#972 A B C
Route= {A,B,C}

53. Explicitly Routed (ER-) LSP ContdIP 1
47.1 3 1 2 3 1 2
47.3 3
47.2 2 IP

54. Explicit Route Advantages
Traffic Engineering
You can control how much traffic travels through some point in the network
This is done by controlling the paths taken by traffic
Fast-rerouting
You can bypass broken links quickly with explicit routing.
No need to wait for a routing protocol (OSPF) to react.
How?
Keep track of two paths, regular path and backup path
If the regular path fails use the backup

55. Virtual Private Networks
We can do VPN’s with MPLS.
Virtual Private Network
A group of connected networks
Connections may be over multiple networks not belonging to the group (e.g. over the Internet)
E.g., joining the networks of several branches of a company into a “private internetwork”

56. Virtual Private NetworksC A B M K L C A B M K L

57. Tunneling
IP Tunnel
Virtual point-to-point link between an arbitrarily connected pair of nodes
IP Tunnel
Network 1
Network 2
Internetwork R1 R2
IP Dest = 2.x
IP Payload
IP Dest = 2.x
IP Payload
IP Dest =
IP Dest = 2.x
IP Payload

58. Tunneling Advantages of tunneling Disadvantages
Transparent transmission of packets over heterogeneous networks
The data carried may not even be IP messages!
Only need to change relevant routers (end points)
Coupled with encryption, gives you a secure private internetwork.
End-points of tunnels my have features not available in other Internet routers.
Multicast
Local Addresses
Disadvantages
Increases packet size
Processing time needed to encapsulate and decapsulate packets
Management at tunnel-aware routers

59. Virtual Private Networks with MPLS
We can do similarly with MPLS
We can connect different sites with an MPLS tunnel
We can send regular IP traffic through the tunnel, or any other type of traffic.

60. “Layer 2” tunnel Use MPLS to provide a tunnel between two
LANs (Ethernet, etc)
ATM points
Any data can be “wrapped” with a label
It need not be IP datagrams
LSR does not look “beyond” the label

61. Demultiplexing Label
What to do with the packet once it reaches the other side of the tunnel?
A “demultiplexing” label needs to be added to inform the end-point router what to do with the packet.

62. E.g., Emulate a VC
ATM cells with a specific VCID come in at the entrance of the tunnel
ATM cells at the end of the tunnel should have the appropriate VCID for the next switch after the router.

64. Emulate a VC (steps)
An ATM cell arrives to the input LSR with VCID 101
The head router attaches the demultiplexing label and identifies the emulated circuit
The head router attaches the tunnel label (to reach the tail router)
Routers in the middle forward as usual
The tail router removes the tunnel label, finds the demultiplexing label, and identifies the VC
The tail router modifies the VCID to the next ATM switch value (202) and sends it to the ATM switch.

65. Label Stacks The previous example has a stack of two labels
You can have larger stacks of labels in the header.
In the example
It enables to have a tunnel
And many types of traffic within the tunnel

66. “Layer 3” VPN’s The packet being carried is an IP packet
Hence the name “layer 3” VPNs
Service provider (see picture next ..)
Has many customers
Each customer has many sites
These sites are linked with tunnels to appear to be one large Internetwork
Each customer can only reach its own sites
The customer is isolated from the rest of the Internet and from other customers