Download Routing

Link Layer Switching




Connecting local networks
Bridges
Repeaters, Hubs, Bridges, Switches,
Routers, Gateways
Virtual LANs
1
Ethernet
• 50  thick: 500 m
• 50  thinn: 185 m
• max 4 repeaters
• traffic on one segment
means traffic on all other
segments
2
CSMA/CD (IEEE 802.3)
Link
Physical
Logical Link Control
A-MAC
Phys. A
B-MAC
Phys. B
C-MAC
Phys. C
3
Bridges

Connection on link layer:


forwarding based on MAC addresses
self-learning bridges



operation
Advantages and limitations
Spanning-tree bridges


operation
Advantages and limitations
4
Self-learning Bridge
Bridge
routing table
LLC
MAC_1
Phys_1
Forwarder
MAC_1
MAC_2
Phys_1
Phys_2
Network 1
LLC
MAC_2
Phys_2
Network 2
5
Self-learning Bridge
Learning
&
routing
Driver interface
.
1
LAN 1
Driver interface
.
2
LAN 2
Routing table
MAC-adr. Interface
.
time
Mac-1
2
-------Mac-2
3
--
Driver interfaec
.
3
LAN 3
6
Self-learning Bridge
Learning phase
Start
Extract
Sender and receiver
MAC-adresser
Look up in
Routing table
Look up in
Routing table
Yes
Update
interface #
and timer
Sender
known?
Forwarding phase
Receiver
known?
No
New entry
MAC-addr
interface #
and timer
Put frame into
correct
outgoing queue
Broadcast
frame, except
on receiving
interface
End
7
Link Layer Switching
Multiple LANs connected by a backbone
to handle a total load higher than the
capacity of a single LAN.
8
Bridge from 802.x to 802.y
IEEE 802 frame formats
9
Bridges from 802.x to 802.y
Operation of a LAN bridge from 802.11 to
802.3.
10
Local “Internetworking”
A configuration with four LANs and two
bridges.
11
Problem with standard bridge
Two parallel transparent bridges.
12
Spanning tree


Goal: each bridge should identify the
interfaes for forwarding traffic
Build a spanning tree

From on root node




Self-configuring
To all nodes
Only these interfaces in the spanning
tree can forward traffic
Provides the shortest path for all traffic
13
Spanning Tree Algorithm
Configuration phase:

Each nodes sends out:






Its own identity (ID) (MAC-address)
ID to the root-bridge
Number of hops to root-bridge
In this way, building up a spanning
tree, bridge with lowest ID become
root node
Start forwarding frames
14
Spanning Tree Bridges
(a) Interconnected LANs. (b) A spanning tree
covering the LANs. The dotted lines are not
part of the spanning tree.
15
Remote Bridges
Bridges can be used to connection
physically distant local networks
16
Switches, Routers and
Gateways
(a) Which device is in which layer.
(b) Frames, packets, and headers.
17
Hub (Nav)
< 100 m
Hub
Hub
Transceiver
Hub
18
Repeaters, Hubs, Bridges,
Switches, Routers and Gateways
(a) A hub. (b) A bridge. (c) a switch.
19
Switched Ethernet
Switch:
•Switches on MAC-addr
•Buffers frames, therefor
no collision
•Competition only for
switch capacity
10, 100, 1000
Mb/s
Switch
Server
Server
20
Gigabit Ethernet
Central server
Gigabit
switch
1000 Mb/s
100 Mb/s
Central server
100/1000
Group server
100/1000
Switch
Switch
Working group 1
Working group 2
Group server
21
Virtual LANs
A building with centralized wiring using hubs and a switch.
22
Virtual LANs (2)
(a) Four physical LANs organized into two VLANs,
gray and white, by two bridges. (b) The same 15
machines organized into two VLANs by switches.
23
The IEEE 802.1Q Standard
Transition from legacy Ethernet to VLAN-aware Ethernet. The
shaded symbols are VLAN aware. The empty ones are not.
24
The IEEE 802.1Q Standard (2)
The 802.3 (legacy) and 802.1Q Ethernet
frame formats.
25
Conclusion

Bridges:





efficient connection alternative
Limits/isolates collision domains
Can be used for traffic isolation
Do not consume IP addresses
Switches:


High use degree, no danger of collisions
Used for establishing virtual LANs
26
Routing and Packet Switching

Goal


Overview of how routing fits into the
Internet architecture
Principles for typical routing protocols


Strengths and weaknesses
Structure




Primary tasks of the network layer
Datagram and virtual line
Some performance considerations
Routing and forwarding
27
Network layer
Server
Client
Disk
Disk
link
28
Tasks of the Network Layer

Responsible for end-to-end transport


Addressing of machines
Forwarding

Connectionless


Connection-oriented (e.g. MPLS or ATM)




datagram; no fixed path through the network
Three phases: connection establishment, data
transmission, teardown
Fixed path through the network
Relatively reliable and ordered transmission
Flow control
29
Forwarding
A
R
R
B
LAN-A
LAN-B
30
Routing and lookup


Mail: [email protected]
Name to address conversion:

ifi.uio.no til IP address: 129.240.64.2


Forward through the network w.r.t. the
network address


Find MAC-address to router and send packet(s)
Based on lookup in routing tables
At the destination router


Convert machines IP address to a MAC address
Send packet to the receiving machine
31
Place of Routing in the
architecture

Structured

Network dimensioning



Traffic directioning




Where should lines be established?
Capacity of lines
Mapping of connections down to paths through the net
Routing to choose paths
Routing of individual packets
Best effort

Routers choose the next hops separately for each
packet
32
Routing


Routing tables can be computed based on
state information about the network
Data exchanged between nodes:


Between neighbour nodes (distance vector
routing; RIP)
Between all nodes in the network (link state
routing; OSPF, IS-IS)
33
Routing types

Static vs. dynamic


Centralized vs. distributed




Dynamic with error handling, new links, changes
of the load
Distributed when routes are computed at all nodes
Global vs. local topology knowledge
Source routing vs. routing
Kilde ruting vs. ruting


In source routing the source chooses the routing
In routing each router choose the next hop
34
Routing Parameters
Performance parameters
•Number of hops
•Price
•Delay
•capacity
Sources of routing information
•None
•Local to the node
•Neighbour nodes
•Nodes along the path
•All nodes in the network
Routing decisions made
•In each node (distributed)
•In a central node (routing
center)
•At the sender (source
routing)
Update interval
•Continously
•Periodic
•In case of large load variations
•In case of topology changes
35
Routing hierarchy

In large networks


Hierarchically structured
Link state



On campus or in companies



Open Shortest Path (OSPF)
Intermediate System to Intermediate System
(IS-IS)
Distance vector, RIP
Static routing
Ad-hoc networks, stationary or mobile
wireless networks

Many different protocols depending on
scenarios
36
Router model
Routing prosess
1
Routing
Route
computationTopology
database
table
2
Pre-
2
1
process
3
Forwarding
3
process
In
Out
e
Principle structure of a router with three incoming and three outgoing connections
37
Routing alternatives



Flooding
Static routing
Adaptive routing should handle




Loss of a link (error, e.g. cable is broken)
Loss of a node (error, e.g. power loss, OS crash)
High traffic load (persistant of transient
congestion, bottleneck)
Disadvantages




Complex, distributed, and not always correct
Adaptivity must be balanced against additional overhead
Can lead to oscillations (route flapping) if reactions are
too fast
Can be unattractive if reactions are too slow
38
Demands on a routing strategy







Shall give correct routes
Shall demand minimal load on nodes
Shall be stable and converge quickly
Fair towards different data streams
Provide optimal routes
Scale with the size of the network
Size with the number of destinations
39
Plug-and-play capabilities





Find neighbour nodes and routers
Detect when neighbours go up and
down
Detect capacity of own links
Send and receive topology information
Send after timer or major changes to
the network
40
Distance vector characteristics

Nodes exchange a vector with their
shortest distance to all destinations

Periodic exchange


Advantage


Convergence is ensured
Simple
Disadvantages


Vulnerable to errors
Slow dissemination in case of problems
41
Distance Vector
5
Di ==
di1
.
.
diN
Distance vector
Si ==
si1
.
.
siN
Next node vector
3
B
2
A
2 9
1
D
C 5
1
F
2
E
Dest. delay Next node
A
0
B
2
B
C
5
C
D
1
D
E
6
C
F
8
C
Node A before change
42
Router model
Routing prosess
1
Routing
Route
computationTopology
database
table
2
Pre-
2
1
process
3
Forwarding
3
process
In
Out
e
Principle structure of a router with three incoming and three outgoing connections
44
Link state



Routing database
Routing table
Periodical and in case of changes



Nodes flood their state onto the link to all other nodes
At start, new nodes downlink the database from a neighbour
Different kinds of link





Point-to-point
Point-to-multipoint
Broadcast
Each node calculates the best route to all other
nodes
Checkpoints

Voting av entire database for link state at a sequence
number
45
LS routing protocol architecture
change
route
lookup
change
Link state
database
Routing
algorithm
Routing table
Protocol for handling
of changes
46
Flooding of link state

Statistically reliable

Each node forwards on all interfaces

All incoming link state packets



If sequence number of large than earlier sequence
numbers
Will most probably reach all node in the network
Content

Sequence number


Node ID of the source



Avoid broadcast storms
Topology
Identify bi-directional links
List of all direct neighbour nodes with a cost
function
47
Link state, LSA
Routin database i D
A B C D E F
5
A
2
1
3
B
2 9
D
A
C 5
1
E
F
2
2 5 1
B 2
2
C 5
D 1 2 9
E
F
48
Link state problems/strengths

Problems


Selection of a node that reports for a
shared medium
Flooding does not scale for large networks


Division into hierarchical networks to limit
flooding
Strengths


All nodes have full topology knowledge
Error have only local relevance
49
Link state problem
a
area 1
area 2
b
We have two problems with the link state method
1.
Static cost factor
•
Can be the source of congestion, all traffic is routing through a single link
2.
Oscillation effects in forwarding traffic
•
At one point in time a is the preferred router between areas
•
Then routing information is exchange
•
New tables are computed and b becomes the preferred router
50
Router and Routecenter
Routers do not have
to participate in a
routing protocol
Route center
Routing center
receives status
reports from routers
Transfers
forwarding table to
routers
Network with router center
51
Connection-oriented forwarding

Establish a channel = a path through the
network






Examples ATM, MPLS, X.25
Explicit signalling
Data-driven signalling
Signalling protocol
Routing protocol to choose the nodes that
should form the path
In each node establishing a forwarding table

Incoming interface, channel – to outgoing
interface, channel
52
53
54
Virtual lines
C
B
1
b
1
b
2
1
1
A
a
c
2
2
1
3
4
D
a
node x
V.C. tables
a
b
1;c;3 1;c;2
2;c;4 2;c;1
c
1;b;2
2;b;1
3;a;1
4;a;2
2
2 c
node y
V.C. tables
a
b
1 ; b ; 1 1 ; a; 1
2 ; b ; 2 2 ; a; 2
3;c;1
4;c;2
c
1 ; a; 3
2 ; a; 4
55
Dynamic cost in route
computation

Adaptation of routes to load


Move traffic to lines with lower load
Main problem



Delay between measurement and computation
Delay between route computation and traffic
arrival
Fast variation in load



Bad predictability
Route flapping (oscillations)
Overhead of exchanging the routing
information
56
Performance of the network


Performance of the networks means
capacity, delay, delay variation (jitter),
and reliable
Has several elements

Transmission delay



Sending delay
Signal propagiation time
Node delay


Processing time
Queueing time
57
Measuring the link state
Node N
packet
T1
Node (N+1)
packet
T0
ACK
58
Topology example
B
C
(7)
Link 1
Link 1
(2)
(2)
A
Link 2
Link 3
G
D
(2)
(1)
Link 2
(3)
F
E
(6)
(3)
(2)
(2)
(4)
H
59
Shortest path tree for nodes
B
E
A
B
A
F
C
C
G
E
H
H
D
F
D
G
60
Routing tables for shortest path
trees
Node A
Node E
Node Link no
Node Link no
B
C
D
E
F
G
H
1
1
1
1
1
1
1
A
B
C
D
F
G
H
1
1
2
2
2
3
2
61
Packet size and delay
62
Modified load variation
Kost
Køteoretisk
forsinkelse
5
4
3
2
Satellitt
1
Jordbunden
0,5
1,0
Utnyttelsesgrad
63
Timing in line and packet
switching
64