2 Network Basics

„ types of communication service

„ how communication services are implemented

z switching

z multiplexing

„ network performance measures

EE210 Data Communications and Computer Networks

1
2.1 Types of service in a layered network architecture
• connection-oriented:
¾ establish a connection
¾ use the connection (for data transfer)
¾ release the connection
• modelled on the telephone system
• essential feature: sender pushes objects (e.g. bits, packets) in at
one end of the connection, and receiver takes them out in the
same order at the other end
• connectionless: each message is sent independently of any other
messages going from the same sender to the same receiver
• modelled on the postal service
• essential features: each message must include the receiver’s
address, and messages can be received in a different order to
the order in which they were transmitted
EE210 Data Communications and Computer Networks
2
Types of service in a layered network architecture (cont.)
• each type of service can be characterised by its reliability: whether
or not the service guarantees to correctly deliver the data
• a reliable service is typically implemented by having the
receiver confirm to the sender that it correctly received each
message (which introduces extra overhead and delays)
• reliable connection-oriented service has 2 variations:
• message stream: preserves message boundaries
(e.g. 2 1-kB messages are received as 2 1-kB messages, not
1 2-kB message, 4 512-byte messages, or anything else)
• byte stream: doesn’t preserve message boundaries
(e.g. 2 1-kB messages are received as a 2048-byte stream)
• unreliable connection-oriented service
• e.g. real-time audio or video: tolerates some errors or losses in
transmission (quality decreases as errors/losses increase)
EE210 Data Communications and Computer Networks
3
Types of service in a layered network architecture (cont.)
• connectionless service can be unreliable (no 100% delivery
guarantee), acknowledged (receipt confirmed), or request-reply (a
single short message contains a request, another the reply)

EE210 Data Communications and Computer Networks

4
Types of service in a layered network architecture (cont.)
• Note: communication services are defined based on end-to-end
properties, not on how bits are transported in the network
• a single network could offer more than one type of service
• e.g. the Internet supports all of them, more or less
• may have different implementations of the same service
• e.g. connection-oriented delivery of a voice bit stream can
be implemented by packet voice or a dedicated circuit

EE210 Data Communications and Computer Networks

5
2.2 How are communication services implemented ?
• it is economically infeasible to directly connect every pair of
sender-receiver pairs (e.g. mesh or star) in a large network
• technology limitations mean that broadcast solutions don’t scale
to large numbers of hosts or large geographical distances
• therefore network resources must be shared between the users,
while still allowing senders to transmit data to their receivers
• the two basic techniques that permit connectivity while sharing
resources are switching and multiplexing
¾ switching: sharing network resources among multiple
transmissions
¾ multiplexing: sharing a single link among multiple
transmissions
EE210 Data Communications and Computer Networks
6
2.2.1.1 Circuit switching
• a path is set up in the network between the sender and the
receiver (by making the appropriate connections in the switches)
• the necessary network resources are reserved for the
connection prior to any data transfer; if this is not possible,
the connection request is blocked
• these reserved resources are then held for the duration of
the connection, regardless of actual usage

Switch = a device
that can create a
temporary
connection
between an input
link and an output
link

EE210 Data Communications and Computer Networks

7
Circuit switching (cont.)
• network links are not shared at the same time
• the links in a path are monopolised for the duration of the
connection, then released so that they are available for other
connections
• the connection set-up delay can be significant (>1 second)
• circuit switching is ideal for “smooth” network traffic
• e.g. telephone network
• what if the traffic from sender to receiver is “bursty” (which
means it varies widely around its average value) ?
• computer-to-computer traffic can be very bursty
¾ could set up a new circuit for each burst
¾ could hold original circuit for duration of data transfer
• both of these solutions are wasteful of network resources

EE210 Data Communications and Computer Networks

8
2.2.1.2 Packet switching
• packet = string of bits (up to a few thousand, typically)
• uses store-and-forward operation:

In packet switching, an
upper limit is placed on
packet size, and packet
transmissions are
pipelined (reducing
the overall delay)

circuit message packet

switching switching switching

9
Packet switching (cont.)
• 2 basic types of packet switching: datagram vs. virtual circuit
• Datagram packet switching:
• each packet is treated individually within the network, so
successive packets may follow different routes through it
• each packet contains the receiver’s address and a sequence
number (so that receiver can put them into correct order)

Network nodes are routers, which have routing tables telling

them which output link to use for each possible destination 10
Packet switching (cont.):

• Datagram packet switching

• no connection set-up needed

• flexible routing possible (e.g. if a router crashes)

• network resources are not shared at the same time

• each packet monopolises a link during its transmission, after
which the link is available for other packet transmissions

• ideal for short-lived bursty traffic

• less suitable for long-lived &/or interactive bursty traffic

EE210 Data Communications and Computer Networks

11
Packet switching (cont.)
• Virtual circuit packet switching:
• a route is set up in the network between sender and
receiver (by making appropriate entries in the routing tables)
• resources may or may not be reserved for this route. If
resources need to be reserved and are not available, the
connection request is blocked
• each packet contains its virtual circuit identifier
Routers have routing tables telling
them which output link to use
for each established virtual circuit

12
Packet switching (cont.):

• Virtual circuit packet switching

• connection set-up required, which may involve significant delay

• network resources are not shared at the same time

• each packet monopolises a link during its transmission, after
which the link is available for other packet transmissions

• less work required at intermediate routers than for datagram p.s.

• given a packet’s input link and virtual circuit identifier, the router
can look up its routing table to find the output link

• virtual circuits not as robust to network problems as datagram p.s.

EE210 Data Communications and Computer Networks

13
Packet switching (cont.):

• Comparison
• virtual circuit p.s. represents a “compromise” between circuit
switching and datagram p.s.

• circuit switching creates a path in the network; virtual circuit p.s.

creates a route which exists only in software; datagram p.s. doesn’t
have routes

• in circuit switching, the links in the path cannot be shared during

the connection; in virtual circuit and datagram p.s. they can

EE210 Data Communications and Computer Networks

14
2.2.2 Multiplexing
• sharing a single link among multiple transmissions
• 3 basic possibilities:
¾ Frequency division multiplexing (FDM)
¾ Time division multiplexing (TDM)
¾ Statistical multiplexing

EE210 Data Communications and Computer Networks

15
2.2.2.1 Multiplexing: Frequency Division Multiplexing (FDM)

EE210 Data Communications and Computer Networks

16
2.2.2.2 Multiplexing: Time Division Multiplexing (TDM)

EE210 Data Communications and Computer Networks

17
 2.2.2.3 Multiplexing: Statistical Multiplexing
• both FDM and TDM divide the link into independent channels
• inefficient if traffic is bursty, since no sharing allowed
• in statistical multiplexing, the idea is that the link should never
be idle when there is data to be transferred

Frames have an additional overhead (compared to FDM or TDM) to indicate

which input stream they belong to; the packet queueing delay is now variable;
and a (possibly complex) method is needed to decide which packets to multiplex
EE210 Data Communications and Computer Networks
18
Transmission Modes

Simplex

Half-duplex

Full-duplex

EE210 Data Communications and Computer Networks

19
2.3 Network Performance Measures
l = length of signal path in communication medium (metres)
v = signal propagation speed in the medium (metres/second)
L = average length of frame or packet (bits)
C = transmission rate (bits/second)
• Propagation delay = l / v , in seconds
• shows how long a bit takes to propagate along the path
• Transmission time = L / C , in seconds
• shows how long it takes to get packet onto the medium
• Throughput: how fast data can pass a certain point
• can be measured in bits/second, packets/second, …
• Efficiency is related to throughput, e.g.
efficiency = throughput (in packets/sec) * packet transmission time
EE210 Data Communications and Computer Networks
20
Network performance measures: Example 1
Consider an optical fibre 3000 km long with a transmitter
transmitting at 1.5 Gbps (1 Gbps = 1 000 000 000 bps).
The signal propagation speed in optical fibre is
approximately 200 000 km/sec. Suppose packet switching
is being used with a packet length of 2000 bits.

What is the bit propagation delay along the fibre ?

prop_delay = (3 000 000)/(200 000 000) = 15 millisec

What is the packet transmission time here ?

pkt_trans = (2 000)/(1 500 000 000) = 1.3333 microsec

How many packets have been transmitted and are propagating

over the fibre when the first bit reaches the
destination ?

num_pkts = (15×10-3)/(1.3333×10-6) = 11 250 packets

(Note that this is 22 500 000 bits)
EE210 Data Communications and Computer Networks
21
Network performance measures: Example 2
Consider a route in a store-and-forward network going
through 8 intermediate nodes. The packets contain 1000
bits and are transmitted at 64 kbps. Assume propagation
delays over the links are negligible. As a packet travels
along the route, it encounters an average of 5 packets
when it arrives at each node. How long does it take for
the packet to get to the receiver if the nodes transmit
on a “first come first served” basis?
At each intermediate node, 6 packets must be transmitted
in order for “our” packet to be transmitted: our packet
finds 5 packets ahead of it, which will be transmitted
first due to the “first come first served” policy.
The packet transmission time at every node is
pkt_trans = (1 000)/(64 000) = 15.625 millisec
The total travel time for our packet through the network is
travel_time = (transmission delay at sender) +
8 × (delay at each intermediate node)
= pkt_trans + 8 × 6 × pkt_trans = 766 millisec
EE210 Data Communications and Computer Networks
22
Network performance measures: Example 2 (cont.)
Note that the “pure” transmission delays only account for
route_trans = pkt_trans + 8 × pkt_trans = 141 millisec
Our packet endures a queueing delay in each intermediate
node of
node_q_delay = 5 × pkt_trans = 78 millisec
and since there are 8 intermediate nodes, the total queueing
delay along this route is
route_q_delay = 8 × node_q_delay = 625 millisec
which represents just over 80% of the total travel time.
• Queueing delay can account for a substantial “extra” delay experienced by
packets in the network
• One way to reduce this queueing delay FOR SOME PACKETS would be to use
a different policy in the intermediate nodes, rather than first-come-first-served.
• This could result in some “higher-priority” packets getting to their receivers
much quicker, while “lower-priority” packets would experience longer delays.
EE210 Data Communications and Computer Networks
23
Network performance measures: Example 3
150 nodes are connected to a 1000 metre length of coaxial
cable. Using some (unspecified) protocol, each node can
transmit 70 frames/second, where each frame is 1000 bits
long. The transmission rate at each node is 100 Mbps.

What is the per-node throughput ?

thru_node = 70 × 1000 = 70 000 bps
What is the total throughput (of the 150 nodes) ?
thru_total = 150 × thru_node = 10 500 000 bps = 10.5 Mbps
What is the efficiency of this protocol ?
efficiency = (total throughput, in bps)
× (bit transmission time)
= (10 500 000) × (1/100 000 000) = 0.105, or 10.5%
OR efficiency = (total throughput, in frames per second)
× (frame transmission time)
= (70 × 150) × (1 000/100 000 000) = 0.105, or 10.5%
EE210 Data Communications and Computer Networks
24
Network performance measures: Example 3 (cont.)
What would give us 100% efficiency, and why is the
efficiency so far below 100% here?

If some node in the network was transmitting at every

instant of time, the total throughput would be 100 Mbps
and the efficiency would be 100%. However, the protocol
being used to regulate access to the medium (the coaxial
cable) introduces some delays between transmissions
– either because it permits collisions to occur, which
must be recovered from; or because some “permission to
transmit” token must be passed from the node currently
in possession to the next node allowed by the protocol
to transmit. In either case, these inter-transmission
gaps result in a drop in efficiency.

• Note that, although this efficiency seems low, it may be perfectly acceptable.

• For example, if the objective was to run the network with a minimum
throughput of 10 Mbps, this protocol works OK.

EE210 Data Communications and Computer Networks

25