- Two heterogeneous networks
cannot be inter-connected
physically
without using a "translation" device...
-
(Because the hardware is
incompatible)
Two heterogeneous networks can be inter-connected logically with the help of a translation device...
- Devices that
interconnect
tow heterogeneous (physical) networks
are called
routers
- Physical networks:
-
So far, we have studied
physical (real) networks
Examples:
- Ethernet
- Token ring
- Token bus
- Ethernet segments interconnected by Ethernet switches
- Characteristic of existing physical networks:
- A physical network
existing today typically
forms
one single broadcast domain.
I.e., broadcasting (sending messages to all nodes) is a trivial (built-in) operation
- Hosts connected to a
physical network
uses
addresses (identifiers)
that are hardwired
into the Network Interface Card (NIC)
E.g., Ethernet hosts has NICs with an Ethernet Address hard wired into the NIC
- A physical network
existing today typically
forms
one single broadcast domain.
- Logical network:
- A logical network
is a network formed by
software using
virtual (assigned) identifications
- Two logical networks
are inter-connected by
a router
- Two inter-connected logical network can interchange logical network messages with each other
- Example: logical
network message
- IP packet (message) !!!!
(We need to explain what a router does...)
- A logical network
is a network formed by
software using
virtual (assigned) identifications
- NOTE:
- I will explain the operation of the
router by using
2 Ethernet networks
Although the networks are homogeneous, the operation will also illustrate the difference between an Ethernet bridge/switch and a Router
- Furthermore, I will use the IP messages to explain the logical inter-connection
- I will explain the operation of the
router by using
2 Ethernet networks
- The Internet is a
virtual network
Fact:
- You cannot buy a network card that uses IP addressess
- So the Internet
is a figment of our imagination....
There are many virtual stuff that are pretty real...
You can even sell virtual islands for real money:
- Recall the Ethernet Frame
format:
7 bytes start 48 bit 48 bit 16 b 46 B <= n <= 1500 B 32b +-----------+----------+-------+------+--------+--------------------+-----+ | Preamble | 10101011 | Src | Dest | Length | Network Layer Data | CRC | +-----------+----------+-------+------+--------+--------------------+-----+
-
We call this a Layer 2 frame
because it is processed by the
Datalink layer (layer 2)
Notes:
- Src = Source Ethernet Address
- Dest = Destination Ethernet Address
- (Abreviated) Structure
of an IP packet:
4 bytes 4 bytes 32 <= 64K +-------+--------+-----+----------------+ | SrcIP | DestIP | CRC | Data | +-------+--------+-----+----------------+
-
We call this a Layer 3 frame
because it is processed by the
Network layer (layer 3)
Notes:
- SrcIP = Source IP (Internet) Address
- DestIP = Destination IP (Internet) Address
- The complete picture:
- A router operates as follows:
- A router is connected to some
physical networks
.... (e.g., Ethernet)
- Routers can receive
layer 2 frames
(e.g., Ethernet frames)
- When a router receives a
level 2 frame (Ethernet frame),
it extracts the
data
(= an IP packet !!!) from the
level 2 frame
- The router then
examines
the
Destination IP address
in the IP packet
- The router determines where to forward the IP packet based on the value of the destination IP address
- And forward the IP packet (unchanged) as data in a new level 2 frame (Ethernet frame)
Graphically:
- A router is connected to some
physical networks
.... (e.g., Ethernet)
- Routing:
- The decision on
where to forward
an IP packet is
called
routing
- We will discuss this topic in more details later...
- The decision on
where to forward
an IP packet is
called
routing
- Forwarding:
- Forwarding decisions
are made based on
physical addresses
(e.g., Ethernet address)
- Physical address are
layer 2 addresses
- Each Ethernet node must have a unique Ethernet address to make forwarding work
- Forwarding decisions
are made based on
physical addresses
(e.g., Ethernet address)
- Routing:
- Routing decisions
are made based on
logical addresses
(e.g., IP address)
- IP address are
layer 3 addresses
(An IP address is inside the data of a layer 2 frame !!!
It's buried one level deeper !!!)
- Each IP host must have a unique IP address to make routing work
- Routing decisions
are made based on
logical addresses
(e.g., IP address)
- Note:
restrictions on
IP address assignment
- Because IP address are
logical entities,
you can potentially
assign an Internet host with a
arbitrary (unique)
IP address
and it "would" work....
- But...
due to some
routing restrictions
(subnet schemes)
in the Internet, the
IP address assignment
is
not completely
arbitrary, but
need to obey some rules.
We will discuss that later....
- Because IP address are
logical entities,
you can potentially
assign an Internet host with a
arbitrary (unique)
IP address
- Routers must
maintain
information
to help them make
routing decisions
- The routing information are store in a data structure
called:
- routing table
The table contains entries of the form (Destination IP address, Output Link)
- routing table
- The following figure summarizes the
forwarding procedure
performed in
bridging:
Notes:
- A bridge
uses promiscuous ports
That means a bridge will receive and process every frame transmitted on the LAN segment.
- A bridge
do not alter (change)
the
Ethernet header information
of a received frame
I.e., the Ethernet frame is repeated verbatim onto the destination LAN segment
- A bridge
uses the destination address in
the Ethernet frame header
to
perform forwarding
- A bridge maintain a forwarding table containing Ethernet addresses
- A bridge does not examine the content in the data portion of the Ethernet frame
- A bridge
uses promiscuous ports
- The following figure summarizes
the routing operation
performed by a router
when an IP-packet is transmited ("routed")
over 2 different LANs:
Notes:
- A router uses
non-promiscuous ports !!!!!
That means:
- A router will only receive an Ethernet frame (containing an IP packet) if the Ethernet frame header contains the Ethernet Address of the router !!!!
- A router will
alter the Ethernet frame header
- In fact, a router will use its own Ethernet address as the source Ethernet address in the transmitted frame !!!
- A router
will look inside the
data portion
of an Ethernet frame
- In fact, the data portion
must contain
information in the IP packet format
(I.e., information fields must be given in a specific order)
- A router needs to look inside the data portion to find the destination IP address !!!
- In fact, the data portion
must contain
information in the IP packet format
- A router uses the
destination IP address
(stored inside the data portion
of the Ethernet frame
to make
the routing decision.
- A router maintain a routing table containing IP addresses
- A router uses information stored in a routing table to make routing decisions
- A router uses
non-promiscuous ports !!!!!
- Achieving scalability:
- One of the important factors
that affect the
scalability
of a networking technique is the
use of broadcasting
- Broadcasting =
sending a message to
all nodes
in the network
- Fact:
- Broadcasting is very intrusive in nature
- One of the important factors
that affect the
scalability
of a networking technique is the
use of broadcasting
- The Internet is a
huge network with
millions of computers
A broadcast message on the Internet will be extremely disruptive
Fact:
- The Internet
does not
support a
broadcast operation
(Before 1992, the Internet only support point-to-point communication.
Since 1992, the Internet supports point-to-multi-point communication - called multicast communication)
- The Internet
does not
support a
broadcast operation
- Effect of an
Ethernet broadcast message in routers:
Notes:
- The Ethernet
broadcast message
will be received
by the router
- The router will perform
its usual operation:
- Extract the
IP packet
- Examine the
Destination IP address
- Forward the IP packet to the destination (Ethernet) network
- Extract the
IP packet
Result:
- Ethernet broadcast messages are not broadcasted on all other LAN segments by routers !!!!
- The Ethernet
broadcast message
will be received
by the router
- In contrast, a bridge
(Ethernet switch)
will forward
Ethernet broadcast messages
blindly onto all
connected LAN segments:
- Network fact:
- Routers are the
preferred way
to inter-connect
multiple networks
- But routers are also more expensive than bridges...
- Routers are the
preferred way
to inter-connect
multiple networks
- Pre-requisite:
- A router that
interconnects 2 different types of networks
will
obviously need use
2 different types of network interface adapters
to connect to each of the networks...
- The networking drivers (software that controls the network cards for each type of network interface will be different (obviously)
- A router that
interconnects 2 different types of networks
will
obviously need use
2 different types of network interface adapters
to connect to each of the networks...
- Other than these
obvious changes,
there is nothing different
in the router's operation
to support interconnecting
two heterogeneuos networks
- Example:
Notes:
- Each type
of network interface adapter
recognizes (and transmits)
its particular frame format.
- Because the router extracts the IP packet from the source frame and puts the (same) IP packet in a new frame (of a different frame type) , the problem of heterogeneous networks is solved
- Each type
of network interface adapter
recognizes (and transmits)
its particular frame format.
- The Internet consists
of many heterogeneous networks,
- At the heart of the Internet
is a group of core routers
that make up of the
Internet Backbone:
- Each company construct
their own backbone
Example: ATT's backbone network
Example: MCI's backbone network
- The different backbones are
interconnected together
to for a global network
- IP messages are
exchange in this
network carried inside the data portion
of the frames:
- Encapsulation:
- The technique where an
entire packet of
a protocol A is
carried inside a
packet of protocol B is
called
encapsulation
Graphically:
- The technique where an
entire packet of
a protocol A is
carried inside a
packet of protocol B is
called
encapsulation
- Real life analogy:
- No, not Russian dolls:
- Rather: FedEx service
Here is the encapsulation technique explained using an analogy:
- No, not Russian dolls:
- Before the advent of the Internet,
there are
many different network protocols
Examples:
- DECNet from
Digital Equipmeent Corp -
see:
click here
- IPX from Novell Networks (used to be very popular for connection PCs) - see: click here
- DECNet from
Digital Equipmeent Corp -
see:
click here
- Problem: (heterogeneity at the network level)
- Packet formats are
different is each of these
network protocols
- Result:
- A computer connected on a type A network cannot communicate with another computer connected on a type B network...
- Packet formats are
different is each of these
network protocols
- With the advent of the
world wide
Internet
protocol, many of these Network Level protocols became
"obsolete"
(they were actually abandoned for economical reasons... - e.g., Novell lossed so much market share that they had to convert to IP or go Chapter 7)...