Network Layer: HosttoHost Communication

1
Network LayerHost-to-Host Communication
2
Network Layer Motivation

• Can we built a global network such as Internet by
  extending LAN segments using bridges?
• No! Bridged networks do not scale
• 4 problems
• We can only bridge certain link-layer
  technologies together
• Link layers to be bridged must have similar MAC
  address structure
• Bridge table explosion
• Bridges use MAC addresses for forwarding and MAC
  addresses are flat, i.e., not hierarchical
• the bridge table needs to have an entry per host
  in the network? bridge table explosion!!!

3
Network Layer Motivation

• 3. Robustness
• Change of network topology requires a new
  spanning tree computation
• 4. Link-layer broadcast storms
• Notice that a bridged network is still a single
  LAN!
• A link-layer broadcast packet must still be
  delivered to ALL hosts in the network.
• Can you imagine receiving a link-layer broadcast
  packet from a host 5000 km away at your host?
• Bottom Line Bridged/Switched LANs dont scale!
• Whats the solution? --- Next

4
How to achieve scalable growth?

• Divide the network into separate LANs that are
  NOT part of the same LL broadcast domain
• Connect the LANs using routers
• Notice that we CANNOT use bridges to connect
  separate LANs as bridged LANs form a single LL
  broadcast domain, which is what we are trying to
  avoid to achieve scalability

5
Communication Issue

• How do two hosts on separate LANs, e.g., A and E,
  communicate?
• Recall that using the Link Layer (LL), only hosts
  that are neighbors, that is, hosts that are
  within the same LAN can communicate.
• Solution Design a new layer, called the network
  layer, that would provide host-to-host packet
  delivery for hosts that are in separate LANs

6
Network Layer Responsibility

• Responsibility Deliver a packet from a sending
  host, e.g., A, to one (unicast) or more
  (multicast) receiving host(s)

7
Network Layer Issues

• Addressing Each host/router interface must have
  a GLOBALLY unique network address
• Recall that within a LAN, this was achieved by
  globally unique MAC addresses at the LL
• We need a similar GLOBALLY unique address at the
  network layer e.g., IP address
• Packet Forwarding How does a packet sent from
  host A to host E make it to host E?
• Recall that within a LAN, this was achieved
  transparently by bridges/switches, with each
  bridge building a forwarding table with respect
  to MAC addresses on the fly
• We need a similar path determination algorithm
  with respect to the destination network address.
• Called path determination or the routing problem.

8
Network Layer Service Model

• Every Network Layer must export a service model
  (interface) to the layers on top of it
• guaranteed bandwidth?
• preservation of inter-packet timing (no jitter)?
• loss-free delivery reliable vs. unreliable?
• in-order delivery ordered vs. unordered?
• unicast vs. multicast delivery?
• congestion feedback to sender?

9
Network Layer Abstraction
The question that a network designer must answer
to satisfy the chosen service model Will the
network be based on

• That is, should we establish an end-to-end path
  through the network for the packets to flow?
• Yes Virtual-Circuit Networks (X.25, Frame-Relay,
  ATM)
• No Datagram Networks (the Internet)

10
Virtual Circuits Networks Signaling
C
R5
R2
R10
R3
A
R6
D
R9
R4
R1
R7
B
R8

• Virtual Circuit Networks (e.g., X.25, Frame
  Relay, ATM)
• Establish a path along which the packets will
  flow between the source and the destination. How?
• Use a signaling (virtual circuit establishment)
  protocol
• Ex B tells its router (R1) that it wants to talk
  to C
• The call establishment message is forwarded by
  the routers in the network until it reaches C.
  Then a reply comes back from C to B.
• Path established at call setup time remains fixed
  during packet exchange
• Routers maintain state information for ongoing
  connections

11
Virtual Circuits Networks Forwarding
C
R5
A
R2
R10
R3
3
3
1
45
R6
43
53
2
69
D
2
22
2
R9
1
9
R4
66
R1
77
12
R7
B
R8
VC table at R1
VC table at R2

• each packet carries tag (virtual circuit ID),
  which determines next hop
• Path established at call setup time remains fixed
  during packet exchange
• Routers maintain state information for ongoing
  connections

12
Datagram Networks Idea
C
R5
A
R2
R10
R3
C
C
C
D
R6
D
D
C
D
C
C
R9
D
R4
C
R1
D
C
C
D
R7
D
B
D
R8
D

• Datagram networks (e.g. the Internet)
• No call establishment before data exchange
• Simply put the destination address on top of the
  packet and submit it to the network for delivery
• Similar to postal service

13
Datagram Networks Forwarding
C
R5
A
R2
R10
R3
C
C
C
3
1
D
R6
D
D
2
C
D
2
C
C
R9
1
D
R4
C
R1
D
C
R7
B
R8
Forwarding table at R1
Forwarding table at R2

• Destination address is written on top of a packet
  and it is simply submitted to the network for
  delivery (like postal service)
• Routers look at destination address in packet to
  determine the next hop
• No connection-state information needed in the
  routers
• Routes may change during session

14
VC vs. Datagram network why?

• Internet (Datagram)
• data exchange among computers
• elastic service, no strict timing req.
• smart end systems (computers)
• can adapt, perform control, error recovery
• simple inside network, complexity at edge
• many link types
• different characteristics
• uniform service difficult

• X.25, Frame Relay, ATM (Virtual Circuit)
• evolved from telephony
• human conversation
• strict timing, reliability requirements
• need for guaranteed service
• dumb end systems
• telephones
• complexity inside network

15
Network layer service models
Guarantees ?
Network Architecture Internet ATM ATM ATM ATM
Service Model best effort CBR VBR ABR UBR
Congestion feedback no (inferred via
loss) no congestion no congestion yes no
Bandwidth none constant rate guaranteed rate gua
ranteed minimum none
Loss no yes yes no no
Order no yes yes yes yes
Timing no yes yes no no

• Internet model being extended Intserv, Diffserv