Lecture #16: Network Layer and

Internetworking
Contents

 Network Layer: functions and services 2

 Network Layer: technologies 6

 Internetworking 7
 Concatenated Virtual Circuits 10

 Connectionless internetworking 12
 Fragmentation 15
 Firewall technology 19 1
 OSI Network Layer
Application layer User application 1 ...
Encryption/ compression/ Choice of
Presentation layer decryption expansion syntax

Session layer Session Session Session to transport Session

control synch. mapping management

Transport layer Layer and flow Error Multiplexing

1/18 control recovery

Network layer Connection Routing Addressing

control
Data link Error Flow Synch Framing
Link layer
establishment control control
Access to Physical and Activation/
Physical layer transm. media electrical interface deactivation of con.

Connection control: establishment, maintaining and terminating network

connections between source and destination open systems
Routing: considerations associated with hop-by-hop services transparent to the
underlying resources such as data link connections .
Addressing: globally unique identification of a service access point of an end
system (transparent to subnet technology (routers/LANs…) and topology (# of
hops) including naming 2
 NL Services to the Transport Layer
 The basic service of the network layer is to provide the
transparent transfer of data between transport entities.
This service allows the structure and detailed content of
submitted data to be determined exclusively by layers above
the network layer.
 The network layer contains functions necessary to provide the
transport layer with a firm network/transport layer boundary
which is independent of the underlying communications media
in all things other than quality of service.
 Thus the network layer contains functions necessary to mask
the differences in the characteristics of different transmission
and subnetwork technologies into a consistent network service.
3
 Services provided to the
transport layer
 Transparent transfer of data between transport entities.
This service allows the structure and detailed content of
submitted data to be determined exclusively by layers
above the network layer.
 Firm network/transport layer boundary which is
independent of the underlying communications media in
all things other than quality of service.
 Mask the differences in the characteristics of different
transmission and subnetwork technologies into a
consistent network service.

4
 Network Layer Service Types
 Connection oriented - virtual circuit (VC) - supported
by the lower network layers (DLL):
– setup and release of the connection
16/1 – connection parameters negotiation
– sequenced delivery of packets
– receiver’s overflow prevented by flow control
– options:
• priority of delivery
• confirmation of delivery
– reliable
– unreliable (rare usage)
– Examples: most popular X.25
16/2
 Connectionless oriented - datagrams exchange -
reliability issues (if present) supported by the transport
layer
– send/receive directives (confirmed/nonconfirmed services)
– independent packets’ (“datagrams”) delivery with full
destination address 5
– Examples: most popular IP (required when using TCP/IP)
 Network Layer Technologies
 Datagram Exchange
– Addressing: full source and destination address in each
datagram
– State information: not needed nor hold 5/2
– Routing: independent routing of the subsequent packets
– Node Failure effects: packets loss
– Congestion control: not typical, rarely applied
– Complexity: in transport layer (above the subnet!)
– Application: connectionless services but also connection
oriented
 Virtual Circuit
– Addressing: short VC number in each packet
– State information: kept in the subnet table for each VC
– Routing: only during the VC setup
– Node Failure effects: VCs termination
– Congestion control: consists of and depends on buffering
– Complexity: in the network layer (in the subnet!) 6
– Application: connection oriented services
 Internetworking - Terms
 Internetworking - multinet structure including different types of
networks and protocols

 Internetworking glossary:
– Communication network: a facility providing data transfer service among
stations attached to the network
5/33

– Internet: a collection of communication networks connected by bridges and/or

routers
– Subnetwork: a constituent network of an internet
– Intermediate system (IS): a connection device between any two
subnetworks
– Repeater: IS that connect two identical subnetworks on the physical level,
repeats the bit sequence without storing of any data.
– Bridge: IS that connects two LANs with identical protocols. Bridges are address
filters that use store-and-forward mechanism without modifying the packets’
contents. It operates on DLL level
– Router: IS that connects two networks with potentially different protocols
(“multiprotocol router”); store-and-forward address filter operating on the
Network Layer
– Gateway: internetworking protocol converters acting on the Transport and
Application layers. Modifications: full and half gateways 7
5/34
5/35 Networks Characteristics
 Protocol stack: OSI/IP/Novel/DECnet/AppleTalk/...
 Addressing scheme: flat files (802.X) vs. hierarchical (IP),
implementation of directory services 9
 Service types incl. QoS: connectivity, confirmed/
/nonconfirmed services, special features support (e.g.real time)
 Parameters: system of timeouts, buffer sizes etc.
 Flow/error control: level of ordering and error protection
 Security: levels of privacy, encryption, identification etc.
 Routing and congestion control: different
mechanisms
 Broadcasting and multicasting: yes/no
 Packet size: maximum size varies substantially
 Accounting rules: yes/no; by traffic/time 8
 Addressing
Uniqueness: Addressing allows the DTE to be uniquely
identified so that data may be routed globally to the correct
destination.
Levels of addressing
Network Level (and above)
SAP: Uniquely identifies the DTE within the internet
16/3 DTE may have more than one SAP, each of them is unique to that particular
DTE
Global Internet Address (GNA) = (network, host or station) parameters
Form: (network identifier, end system identifier)
Subnet Level
A unique address for each DTE attached to the subnet
Referred to as the Subnetwork Attachment Point Address (SAPA)
Host parameter of GNA and SAPA may be the same but are often not
Different networks use different addressing formats and lengths (ARP, RARP)
Some host have more than one attachment point to the subnet 9
Host parameter (GNA) has global significance, SAPA has local significance
 Concatenated Virtual Circuits
CVC is End-to-End connection that consists of several
consecutive Point-to-Point links between:
source host and subnet
5/36 subnet and multiprotocol router (“full gateway”)
[subnet and subnet, connected by shared “half-gateways”])
subnet and destination host
Features:
the data routes are identified by VC numbers
during the session data packets traverse the same sequence of GWs
and arrive in order
the routes are supported by VC tables containing
the ID number of the actual VCs
the next destination for each VC
the number of the next concatenated VC
Application: internetworking in set of subnets of similar type of
services (e.g. either reliable or unreliable). Usually implemented on
Transport layer (e.g. TCP - End-to-End transport protocol) 10
 Concatenated Virtual Circuits
Pro’s Contra’s
• reservation of buffers and • waste of buffer space (table
communication capacity in space) for each open
advance connection
• guaranteed sequencing,
• static routing during the
delivery and stable delays
session i.e. bad congestion
• possible implementation of
control
any type services
• short addressing (small • vulnerability to router failures

communication overload due • complicated implementation in

to the headers) unreliable datagram
• small communication overload subnetworks
due to packets retransmission
and losses 11
 Connectionless Internetworking
Applies Datagram model

Features:

• independent routing for each packet thus optimizing the

5/37
the congestion

• not-in-order delivery

• datagram packets can be routed around network failure

points in d.g. subnetworks

• requires universal addressing system - Internet, IPX,

OSI, SNA, AppleTalk address standards

12
 Connectionless Internetworking
Pro’s Contra’s
• communication overhead due
• adaptive dynamic routing and
to longer address fields,
adaptive congestion control
repeated in each datagram
• low buffer space needed at • communication overhead due

routers to unreliable unordered

services
• robustness to router failures
• dispersed delay duration
• applicable for any type of
• requires universal addressing
subnets incl. unreliable ones system

13
 Tunneling
• Tunneling is a technique for connection of two similar
5/38 networks through the arbitrary type[s] of intermediate
network[s]
• Data entities (datagrams, packets) of two ends are
packed together with their control information
(addressing, ordering, error control fields, etc.) into the
payload field of the intermediate network’ NL packets
• The original control information is not being interpret
anywhere in the intermediate network but in both ends
• Therefore, tunneling needs multiprotocol routers only on
the both ends of the “tunnel” where the original data
entities are constructed/restored
14
 Fragmentation
• Fragmentation is the process of splitting of the data
structures into the entities that are suitable to transmit over
the various networks and the reverse process of restoring the
original structures out of the fragments.
• Fragmentation factors:
• Transmission method (bit error rate, multiplexing method, etc.)
• Operating system (read/write blocks of 0.5 kB)
• Protocols (packet length field limitation)
• Standardization
• Service discipline and resource sharing in the end stations and intermediate
systems (IS): routers, gateways (e.g. SJF “shortest job first”, RR “Round
Robin” etc.)
• Examples of payload size:
• ATM cell carries 48B
• IP packet carries 64kB
• Data packets are broken into fragments and each
fragment is sent in separate internet packet. 15
 Fragmentation Methods
• Each network in the internet is bounded by gateways which
are the entry point and the output point of the packets
5/41a

traversing that network

• 1st approach: transparent fragmentation. Large packets
are fragmented (if needed!) into fragments at the small-
packet-network entry point (gateways G1, G3) and
resembled back at the network output point (G2, G4). Note
that all the fragments should reach the same network output
point!
• Example: ATM networks hardware fragmentation/defragmentation
of the packets into ATM cells at each entry/output point
• Requirements/features:
• additional counting of the number of fragments in connectionless
networks or End-of-the-packet flag in the last fragment in the
connection-oriented networks
• congestion control and performance are affected by the requirement
for similar routing of all the fragments
• multiple fragmentation/defragmentation cycles may occur during an
16
internet route of a long packet
 Fragmentation Methods (2)
•2nd approach: nontransparent fragmentation. Large
5/41b
packets are fragmented (if needed!) at the small-packet-
network entry point (gateway G1), then traverse the internet
as independent packets and are resembled back only at the
destination host.
•Requirements/features:
• defragmentation capabilities of each host
• communication overhead for each fragment during the whole route
• better possibility for congestion control and dynamic routing (in the
datagram model)
• only one fragmentation/defragmentation cycle (if any!) may occur
during an internet route of a long packet
• possibility for hierarchical fragmentation: fragmentation of already
fragmented packets in case the route passes network of even
smaller packets: tree-numbering of the fragments that can be
extended hierarchically (e.g.
18 [0.]  [0.0, 0.1, 0.2 ...]  [0.0.0, 0.0.1, 0.0.2 … 0.1.0, 0.1.1 ...] ... 17
 Fragmentation Methods (3)
Requirements/features (cont.):
• fragmentation to some elementary frame size.
Fragments are short enough to be carried by any
intermediate network. An internet packet contains one
or more elementary frames. Additional flagging:
• packet ID number
5/42 • ordering number of the first elementary fragment in the packet
• end-of-the-packet flag (1 bit: end/no_end)

18
19
20
21
22
23
24
25
26
27
28
29
30
31