1. Switching and Forwarding
3.2 Bridges and LAN Switches
3.3 Cell Switching (ATM)
3.4 Implementation and Performance

2. Two limitations on the directly connected networks
limit on how many hosts can be attached, examples
only two hosts can be attached to a point-to-point link
the Ethernet specification allows no more than 1,024 hosts

3. limit on how large of a geographic area a single network can serve, examples
an Ethernet can span only 2,500 m
wireless networks are limited by the ranges of their radios
point-to-point links can be quite long

4. Goal
build networks that can be global in scale
Problem
how to enable communication between hosts that are not directly connected
Solution
computer networks use packet switches to enable packets to travel from one host to another, even when no direct connection exists between those hosts

5. Packet switch
a device with several inputs and outputs leading to and from the hosts that the switch interconnects
Core job of a switch
take packets that arrive on an input and forward (or switch) them to the right output so that they will reach their appropriate destination

6. A key problem that a switch must deal with is the finite bandwidth of its outputs
if packets destined for a certain output arrive at a switch and their arrival rate exceeds the capacity of that output, then we have a problem of contention
the switch queues (buffers) packets until the contention subsides, but if it lasts too long, the switch will run out of buffer space and be forced to discard packets
when packets are discarded too frequently, the switch is said to be congested

7. 3.1 Switching and Forwarding
a multi-input, multi-output device, which transfers packets from an input to one or more outputs
star topology
switched networks are more scalable (i.e., growing to large numbers of nodes) than shared-media networks because of the ability to support many hosts at full speed

8. A switch provides a star topology

9. Scalable Networks
The figure shows the protocol graph that would run on a switch that is connected to two T3 links and one STS-1 SONET link
Example protocol graph running on a switch

10. A switch forwards packets from input port to output port
Port selected based on address in packet header
Advantages
cover large geographic area (tolerate latency)
support large numbers of hosts (scalable bandwidth)

11. Example switch with three input and output ports

12. How does the switch decide on which output port to place each packets?
general answer
it looks at the header of the packet for an identifier that it uses to make the decision
three common approaches
datagram (or connectionless) approach
virtual circuit (or connection-oriented approach)
source routing

13. 3.1.1 Datagram Switching Sometimes called connectionless model
Analogy: postal system
No connection setup phase
no round trip delay waiting for connection setup
a host can send data as soon as it is ready

14. Each packet is forwarded independently of previous packets that might have been sent to the same destination
two successive packets from host A to host B may follow completely different paths (perhaps because of a change in the forwarding table at some switch in the network)

15. A switch or link failure might not have any serious effect on communication if it is possible to find an alternate route around the failure and to update the forwarding table accordingly
Since every packet must carry the full address of the destination, the overhead per packet is higher than for the connection-oriented model

16. Source host has no way of knowing if the network is capable of delivering a packet or if the destination host is even up and running
Each switch maintains a forwarding (routing) table

17. Example the hosts have addresses A, B, C, and so on
a switch consults a forwarding table (routing table) to decide how to forward a packet

18. Datagram forwarding: an example network

19. The table shows the forwarding information that switch 2 needs to forward datagrams
Destination
Port A 3 B C D E 2 F 1 G H

20. 3.1.2 Virtual Circuit Switching
Sometimes called connection-oriented model
Analogy: phone call
Explicit connection setup (and tear-down) phase
it requires that a virtual connection from the source host to the destination host is set up before any data is sent
Typically wait full RTT (Round Trip Time) for connection setup before sending first data packet

21. If a switch or a link in a connection fails
the connection is broken and a new one needs to be established
Subsequence packets follow same circuit
Each switch maintains a Virtual Circuit (VC) table

22. Entry in the VC table on a single switch contains
a virtual circuit identifier (VCI)
uniquely identifies the connection at this switch
which will be carried inside the header of the packets that belong to this connection

23. an incoming interface
on which packets for this VC arrive at the switch
an outgoing interface
in which packets for this VC leave the switch
a potentially different VCI that will be used for outgoing packets

24. Two classes of approaches to establish connection state
Permanent Virtual Circuit (PVC)
Switched Virtual Circuit (SVC)

25. Permanent Virtual Circuit (PVC)
administrator configures the state, in which case the virtual circuit is “permanent”
administrator can also delete the state, so a permanent virtual circuit (PVC) might be thought of as a long-lived, or administratively configured VC

26. Switched Virtual Circuit (SVC)
a host may set up and delete a VC by sending messages without the involvement of a network administrator
this is referred to as signaling, and the resulting virtual circuits are said to be switched
an SVC should more accurately be called a “signaled” VC, since it uses signaling (not switching) to distinguish an SVC from a PVC

27. Example
assume that a network administrator wants to manually create a new virtual connection from host A to host B
two-stage process
connection setup
data transfer

28. An example of a virtual circuit network
(11)
(7)
(5)
(4)
An example of a virtual circuit network

29. The administrator picks a VCI value that is currently unused on each link for the connection
suppose
VCI = 5, the link from host A to switch 1
VCI = 11, the link from switch 1 to switch 2
VCI = 7, the link from switch 2 to switch 3
VCI = 4, the link from switch 3 to host B

30. VC table entry at switch 1
Incoming Interface
Incoming VCI
Outgoing Interface
Outgoing VCI 2 5 1 11
VC table entry at switch 1
Incoming Interface
Incoming VCI
Outgoing Interface
Outgoing VCI 3 11 2 7
VC table entry at switch 2
Incoming Interface
Incoming VCI
Outgoing Interface
Outgoing VCI 7 1 4
VC table entry at switch 3

31. A packet is sent into a virtual circuit network

32. A packet makes its way through a virtual circuit network

33. Hop-by-hop flow control
each node is ensured of having the buffers it needs to queue the packets that arrive on that circuit
example, an X.25 network-a packet-switched network that uses the connection-oriented model

34. X.25 network employs the following three-part strategy
buffers are allocated to each virtual circuit when the circuit is initialized
the sliding window protocol is run between each pair of nodes along the virtual circuit, and this protocol is augmented with flow control to keep the sending node from overrunning the buffers allocated at the receiving node

35. the circuit is rejected by a given node if not enough buffers are available at that node when the connection request message is processed

36. Examples of virtual circuit technologies
Asynchronous Transfer Mode (ATM)
Frame Relay, e.g., Virtual Private Network (VPN)
Frame Relay operates only at the physical and data link layers

37. 3.1.3 Source Routing
Neither virtual circuits nor conventional datagrams
All the information about network topology that is required to switch a packet across the network is provided by the source host

38. Various ways to implement source routing
method1
put an ordered list of switch ports in the header and to rotate the list so that the next switch in the path is always at the front of the list
for each packet that arrives on an input, the switch would read the port number in the header and transmit the packet on that output

39. Source routing in a switched network (where the switch reads the rightmost number)

40. method2
example, rather than rotate the header, each switch just strip the first element as it uses it
method3
have the header carry a pointer to the current “next port” entry, so that each switch just updates the pointer rather than rotating the header

41. and (c) pointer. The labels are read right to left
Three ways to handle headers for source routing: (a) rotation, (b) stripping,
and (c) pointer. The labels are read right to left

42. 3.2 Bridges and LAN Switches
LANs have physical limitations (e.g., 2500m)
Bridge (LAN switch)
connect two or more LANs
Extended LAN
a collection of LANs connected by one or more bridges
accept and forward strategy (accept all frames transmitted on either of the Ethernets, so it could forward them to the other)

43. 3.2.1 Learning Bridges Do not forward when unnecessary
whenever a frame from host A that is addressed to host B arrives on port 1, there is no need for the bridge to forward the frame out over port 2

44. Illustration of a learning bridge

45. Host
Port A 1 B C X 2 Y Z
How does a bridge come to learn on which port the various hosts reside?
each bridge inspects the source address in all the frames it receives
when host A sends a frame to a host on either side of the bridge, the bridge receives this frame and records the fact that a frame from host A was just received on port 1
in this way, the bridge can build a table just like the following table

46. Host
Port A 1 B C X 2 Y Z

47. 3.2.2 Spanning Tree Algorithm
Problem: extended LAN has a loop in it
frames potentially loop through the extended LAN forever
example
bridges B1, B4, and B6 form a loop

48. Extended LAN with loops

49. Solution: bridges run a distributed spanning tree algorithm
spanning tree is a subgraph of a graph that covers (spans) all the vertices, but contains no cycles

50. Example of (a) a cyclic graph; (b) a corresponding spanning tree

51. Spanning tree algorithm (developed by Radia Perlman)
each bridge has a unique identifier (e.g., B1, B2, B3)
the algorithm first elects the bridge with the smallest ID as the root of the spanning tree
the root bridge always forwards frames out over all of its ports

52. each bridge computes the shortest path to the root and notes which of its ports is on this path
this port is selected as the bridge’s preferred path to the root

53. finally, all the bridges connected to a given LAN elect a single designated bridge that will be responsible for forwarding frames toward the root bridge
each LAN’s designated bridge is the one that is closest to the root, and if two or more bridges are equally close to the root, then the bridges’ identifiers with the smallest ID wins

54. Spanning tree with some ports not selected

55. Bridges have to exchange configuration messages with each other and then decide whether or not they are the root or a designated bridge based on these messages
configuration messages contain
the ID for the bridge that is sending the message
the ID for what the sending bridge believes to be the root bridge
the distance, measured in hops, from the sending bridge to the root bridge

56. each bridge records current best configuration message for each port
initially, each bridge believes it is the root
when learn not root, stop generating config messages
in steady state, only root generates configuration messages
when learn not designated bridge, stop forwarding config messages
in steady state, only designated bridges forward config messages

57. root continues to periodically send config messages
if any bridge does not receive config message after a period of time, it starts generating config messages claiming to be the root
upon receiving a config message over a particular port
the bridge checks to see if that new message is better than the current best configuration message recorded for that

58. the new configuration message is considered “better” than the currently recorded information if
it identifies a root with a smaller ID or
it identifies a root with an equal ID but with a shorter distance or
the root ID and distance are equal, but the sending bridge has a smaller ID

59. Sequence of events
assume all the bridges boot at about the same time and all the bridges would start off by claiming to be the root
(Y, d, X) denotes a configuration message from node X in which it claims to be distance d from root node Y

60. Sequence of events on the activity at node B3
B3 receives (B2, 0, B2)
since 2 < 3, B3 accepts B2 as root [(B2, 1, B3)]
B3 adds one to the distance advertised by B2 (0) and thus sends (B2, 1, B3) toward B5 [(B2, 1, B3), (B2, 2, B5)]
meanwhile, B2 accepts B1 as root because it has the lower ID, and it sends (B1, 1, B2) toward B3 [(B1, 1, B2), (B1, 2, B3)]

61. B5 accepts B1 as root and sends (B1, 1, B5) toward B3 [(B1, 1, B5), (B1, 2, B3)]
B3 accepts B1 as root, and it notes that both B2 and B5 are closer to the root than it is [(B1, 2, B3), (B1, 1, B2), (B1, 1, B5)]
B3 stops forwarding messages on both its interfaces (this leaves B3 with both ports not selected) [(B1, 1, B2), (B1, 1, B5)]

62. Spanning tree with some ports not selected
(1)
(5b)
(6)
(2)
(7)
(4b)
(3)
(4a)
(5a)
Spanning tree with some ports not selected

63. 3.2.3 Broadcast and Multicast
Since most LANs support both broadcast and multicast, then bridges must also support these two features
Broadcast
each bridge forwards a frame with a destination broadcast address out on each active (selected) port other than the one on which the frame was received
Multicast
implemented in exactly the same way, with each host deciding itself whether or not to accept she message

64. 3.2.4 Limitations of Bridges
Do not scale
Do not accommodate heterogeneity

65. Do not Scale
It is not realistic to connect more than a few (tens of) LANs by means of bridges
the spanning tree algorithm scales linearly, i.e., there is no provision for imposing a hierarchy on the extended LAN
bridges forward all broadcast frames and broadcast does not scale

66. Virtual LAN (VLAN) used to increase the scalability of extended LANs
allows a single extended LAN to be partitioned into several seemingly separate LANs
each virtual LAN is assigned an identifier (sometimes called a color), and packets can only travel from one segment to another if both segments have the same identifier
this limits the number of segments in an extended LAN that will receive any given broadcast packet

67. Example four hosts (W, X, Y, Z) on four different LAN segments
in the absence of VLANs, any broadcast packet from any host will reach all the other hosts
suppose that we define the segments connected to hosts W and X as being in one LAN, VLAN 100
also define the segments that connect to hosts Y and Z as being in VLAN 200
to do his, we need to configure a VLAN ID on each port of bridges B1 and B2
the link between B1 and B2 is considered to be in both VLANs

68. Two virtual LANs share a common backbone

69. When a packet sent by host X arrives at bridge B2
the bridge observes that it came in a port that was configured as being in VLAN 100
it inserts a VLAN header between the Ethernet header and its payload
the bridge applies normal rules for forwarding to the packet, with the extra restriction that the packet may not be sent out an interface that is not part of VLAN 100
thus, even a broadcast packet can’t be sent out the interface to host Z, which is in VLAN 200
Ethernet header
VLAN header
Payload

70. An attractive feature of VLANs
it is possible to change the logical topology without moving any wires or changing any addresses
example
if we want to make the segment that connects to host Z be part of VLAN 100, and thus enable X, W and Z be on the same virtual LAN, we would just need to change one piece of configuration on bridge B2

71. Do not Accommodate Heterogeneity
Bridges are fairly limited in the kinds of networks they can interconnect
Bridges make use of the networks frame header and so can support only networks that have exactly the same format for addresses
Bridges can be used to connect Ethernets to Ethernets, (Token Ring) to 802.5, and Ethernets to rings, since both networks support the same 48-bit address format
Bridges do not readily generalize to other kinds of networks, such as ATM