Delivery and Forwarding

Kae Won Choi

1
Connection, connection-less service
 datagram network provides network-layer
connectionless service
 virtual-circuit network provides network-layer
connection service
 analogous to TCP/UDP connection-oriented /
connectionless transport-layer services, but:
 service: host-to-host
 no choice: network provides one or the other
 implementation: in network core
Virtual circuits
“source-to-dest path behaves much like telephone
circuit”
 performance-wise
 network actions along source-to-dest path

 call setup, teardown for each call before data can flow
 each packet carries VC identifier (not destination host
address)
 every router on source-dest path maintains “state” for
each passing connection
 link, router resources (bandwidth, buffers) may be
allocated to VC (dedicated resources = predictable
service)
VC implementation
a VC consists of:
1. path from source to destination
2. VC numbers, one number for each link along path
3. entries in forwarding tables in routers along path
 packet belonging to VC carries VC number
(rather than dest address)
 VC number can be changed on each link.
 new VC number comes from forwarding table
 VC forwarding table
12 22 32

1 3
2
VC number
interface
forwarding table in number
northwest router:
Incoming interface Incoming VC # Outgoing interface Outgoing VC #

1 12 3 22
2 63 1 18
3 7 2 17
1 97 3 87
… … … …

VC routers maintain connection state information!

Virtual circuits: signaling protocols
 used to setup, maintain teardown VC
 used in ATM, frame-relay, X.25
 not used in today’s Internet

application application
5. data flow begins 6. receive data
transport transport
network 4. call connected 3. accept call
1. initiate call network
data link 2. incoming call
data link
physical physical
Datagram networks
 no call setup at network layer
 routers: no state about end-to-end connections
 no network-level concept of “connection”
 packets forwarded using destination host address

application application
transport transport
network 1. send datagrams 2. receive datagrams network
data link data link
physical physical
Datagram forwarding table
4 billion IP addresses, so
routing algorithm rather than list individual
destination address
local forwarding table
list range of addresses
dest address output link (aggregate table entries)
address-range 1 3
address-range 2 2
address-range 3 2
address-range 4 1

IP destination address in
arriving packet’s header
1
3 2
Route method versus next-hop method

9
Host-specific versus network specific method

10
Default method

11
Forwarding module

12
Route example

13
Routing table example

14
Address aggregation

15
Longest mask matching

16
Hierarchical routing

17
Common fields in a routing table

18
Example

19
Router architecture overview
two key router functions:
 run routing algorithms/protocol (RIP, OSPF, BGP)
 forwarding datagrams from incoming to outgoing link

forwarding tables computed, routing

pushed to input ports routing, management
processor
control plane (software)

forwarding data
plane (hardware)

high-seed
switching
fabric

router input ports router output ports

 Input port functions
lookup,
link forwarding
line layer switch
termination protocol fabric
(receive)
queueing

physical layer:
bit-level reception
data link layer: decentralized switching:
e.g., Ethernet  given datagram dest., lookup output port
see chapter 5 using forwarding table in input port
memory (“match plus action”)
 goal: complete input port processing at
‘line speed’
 queuing: if datagrams arrive faster than
forwarding rate into switch fabric
Switching fabrics
 transfer packet from input buffer to appropriate
output buffer
 switching rate: rate at which packets can be
transfer from inputs to outputs
 often measured as multiple of input/output line rate
 N inputs: switching rate N times line rate desirable
 three types of switching fabrics

memory

memory bus crossbar

Switching via memory
first generation routers:
 traditional computers with switching under direct control
of CPU
 packet copied to system’s memory
 speed limited by memory bandwidth (2 bus crossings per
datagram)

input output
port memory port
(e.g., (e.g.,
Ethernet) Ethernet)

system bus
Switching via a bus
 datagram from input port memory
to output port memory via a
shared bus
 bus contention: switching speed
limited by bus bandwidth
 32 Gbps bus, Cisco 5600: sufficient bus
speed for access and enterprise
routers
Switching via interconnection network
 overcome bus bandwidth limitations
 banyan networks, crossbar, other
interconnection nets initially
developed to connect processors in
multiprocessor
 advanced design: fragmenting
datagram into fixed length cells, crossbar
switch cells through the fabric.
 Cisco 12000: switches 60 Gbps
through the interconnection
network
Output ports

datagram
switch buffer link
fabric layer line
protocol termination
queueing (send)

 buffering required whenDatagram

datagrams arrive from
(packets) can be lost
fabric faster than the transmission rate lack of buffers
due to congestion,
 scheduling discipline chooses among queued
datagrams for transmission
Priority scheduling – who gets best
performance, network neutrality
Output port queueing

switch
switch
fabric
fabric

at t, packets more one packet time later

from input to output

 buffering when arrival rate via switch exceeds

output line speed
 queueing (delay) and loss due to output port buffer
overflow!
Input port queuing
 fabric slower than input ports combined -> queueing may
occur at input queues
 queueing delay and loss due to input buffer overflow!
 Head-of-the-Line (HOL) blocking: queued datagram at front
of queue prevents others in queue from moving forward

switch switch
fabric fabric

output port contention: one packet time later:

only one red datagram can be green packet
transferred. experiences HOL
lower red packet is blocked blocking