- Switch
- A switch
is an interconnection device
that connects
multiple networks
that uses
the same network protocol
I.e.:
- A switch is a "Multiple Port Bridge"
- A switch
is an interconnection device
that connects
multiple networks
that uses
the same network protocol
- Networking Babel:
- In this course:
- A switch connect (multiple) networks of the same type
- In this course:
a device that connect
networks of the different type
is called:
- A router (See: click here )
- However, some poeple
will call
device that connect
networks of the different type:
- A switch.....
- In this course:
- Just like a bridge, for each port
of the switch, there is a
forwarding table :
Learning:
- When the switch
receives a
frame from
a node:
the switch will enter/update the source network address into the forwarding table of the corresponding port
- When the switch
receives a
frame from
a node:
- Forwarding Operation of a
switch:
- Suppose an Ethernet switch
receives an Ethernet frame
on some port:
- it will use the Destination Ethernet Address to perform forwarding
-
If the Destination Ethernet Address
is found in the
forwarding table of
same
Ethernet segment, then
the
switch will
ignore the frame:
- If the
Destination Ethernet Address
is found in
one of the
forwarding tables,
the switch will
forward the frame to
that port:
- If the
Destination Ethernet Address
is not found in
any of the
forwarding tables,
the switch will
broadcast (= send to all)
the frame to
all other ports:
- Suppose an Ethernet switch
receives an Ethernet frame
on some port:
- Example network 1:
Notice that:
- The node H is
not on a network
that is directly attached
to switch 1
- But H can still be stored in the forwarding table of switch 1 !!!
- The node H is
not on a network
that is directly attached
to switch 1
- How the frame
from A ⇒ H is
forwarded:
- Switch 1 finds H
in its
forwarding table
and forwards the
frame:
- Switch 2 finds H
in its
forwarding table
and forwards the
frame:
- Switch 1 finds H
in its
forwarding table
and forwards the
frame:
- Example newtork 2:
Changes made to the topology:
- The switches are
connected directly to
each other
- There are no networks in between the switches
- The switches are
connected directly to
each other
- How the frame
from A ⇒ H is
forwarded:
Result:
- Only nodes on the source network and the destination network will hear the frame transmission !!!
- Network fault:
- Links can
fail (cut) (but this event is
rare)
- Nodes (switches) can fail (short-circuit) (this kind of failure is very likely)
- Links can
fail (cut) (but this event is
rare)
- Fault tolerant computer
networks:
- Fault tolerant computer networks = a network topology that can keep on operating when some component(s) have failed
- How to design
fault tolerant computer networks:
- Fault tolerance is
achieved by
providing
alternate routes
between nodes:
- A fault will
cause disrupt
some route
- When one of the route is disrupted, the frames can still reach the destination through another route
- Fault tolerance is
achieved by
providing
alternate routes
between nodes:
- Nasty consequence
of fault tolerant networking:
- A fault tolerant network
will (always) contain
loops:
(Loops in a network is a real pain in your-know-where....)
- A fault tolerant network
will (always) contain
loops:
- Effect of
loops in a
computer network
- Frames (messages
may be repeated
indefinitely
if a network contains
loop !!!
- When messages are
repeated
indefinitely,
the network will
become
extremely congested
- Network performance will drop dramatically and the network becomes useless !!!
- When messages are
repeated
indefinitely,
the network will
become
extremely congested
- Frames (messages
may be repeated
indefinitely
if a network contains
loop !!!
- Example: a network with
a loop
- The loop problem illustrated:
- We begin with
empty
forwarding tables:
- When node X transmits
a frame to
node Y:
Notice the entries of X in the forwarding tables
- After one more round of
forwarding:
Notice that:
- The entry for X in the two forwarding tables of switch Sw1 !
- After one more round of
forwarding:
- Now the network
is in trouble:
because more copies of the frame will keep going ariund in the loop:
Notice there are now 2 copies of the frame !!!
- We begin with
empty
forwarding tables:
- Conclusion:
- A loop will cause
frames (messages) to
cycle indefinitely in
the network
- Furthermore:
- The number of copies of a message may multiply if there is a loop in the network !!!
- A loop will cause
frames (messages) to
cycle indefinitely in
the network
- A loop-free network is
called:
- a tree
Example:
- Logical network vs.
Physical network:
- Physical network =
the network consisting of the
actual nodes and links
Example:
- Logical network =
the network consisting of the
nodes and links that
are in use
Example:
- Physical network =
the network consisting of the
actual nodes and links
- Providing fault-tolerance:
- The physical network must
contain loops
- Because alternative routes will necessarily form a loop !!!
- We will create a
loop-free logical network
that is a
sub-network (graph)
of the
physical network by:
- Disabling one or more ports on a switch that is part of a loop
The switch will forward frame using the logical network !!!
- The physical network must
contain loops
- Next:
- We will study a
distributed algorithm
that the Ethernet switches
execute to:
- Form a tree (= the logical network)
- The algorithm is called the IEEE 8021D Spanning Tree Algorithm
- We will study a
distributed algorithm
that the Ethernet switches
execute to:
-
Note:
- The IEEE Standard call an
Ethernet switch:
- a (multiway) bridge
- The IEEE Standard call an
Ethernet switch: