Packet Switched Data Networks

1
Packet Switched Data Networks
2
What is a PSDN?

• The telephone network is widely referred to as a
  Public Switched Telephone Network (PSTN).
• This means that there is public access (if you
  buy a telephone and pay an installation fee, line
  rental and line usage charges).
• There are similar services available for computer
  communications. These are called Packet Switched
  Data Networks (PSDN).
• If you are prepared to buy a computer, pay an
  installation fee, line rental and line usage
  charges, you could connect to such a network.
• Usually, however, only large businesses,
  universities and government organisations can
  afford to do so.

3
Why use PSDNs?

• Most PSDNs are built and run by telephone
  companies.
• PSDN offer high data rates at a relatively low
  price (when compared with the cost of using a
  modem over a telephone line) over large
  distances.
• PSDN are particularly cost effective if data
  traffic is high.
• If the amount of data is low but constant, then a
  leased line (permanently connected telephone
  line) may be better because installation cost is
  lower.
• For low amounts of data traffic that needs to be
  sent only occasionally, an ordinary modem and
  telephone line or an ISDN line is adequate (e.g.
  a home user connecting to the Internet).

4
System Network Architecture (SNA)

• In the early 1970s, IBM developed a reliable
  system that enabled mainframes to communicate
  with each other.
• Each subarea network is controlled by a
  mainframe, which manages dedicated switching
  minicomputers.
• SNA uses a 7-layer architecture similar to the
  OSI 7-layer reference model (in fact the OSI
  model was originally based on the SNA model).
• At the Data Link Layer, SNA uses a system called
  SDLC (Synchronous Data Link Control).

5
Synchronous Data Link Control (SDLC)

• SDLC provides a set of interface primitives to
  the network layer (similar to LLC). It also
  defines the format of the frames sent over the
  data links.
• Every frame starts and ends with the bit pattern
  01111110.
• Bit stuffing is used within the frame body so
  that any group of five 1s is replaced by 111110.
• Each host on the subarea network is assigned a
  unique address, which is used in the address
  field of the frame.
• The control field contains bits representing
  sequence numbers, acknowledgements and type
  information.
• The data field contains an arbitrary number of
  bits.
• The checksum is a slight variant of a 16-bit CRC.

6
SDLC Control Field

• There are three types of frame information
  frames, supervisory frames and unnumbered frames.
  The contents of the control field is different
  for each type.
• Information frames use a 3-bit sliding window
  protocol. The sequence number is placed in the
  seq bits.
• P/F stands for Poll/Final. It is set to P to
  invite a response from a terminal. The frames
  sent back in response have the P bit set except
  for the last one which is set to F.

7
SDLC Control Field

• Rather than sending separate acknowledgement
  frames, most acknowledgements are piggybacked on
  response frames.
• It makes sense because the overhead of sending a
  whole frame for every acknowledgement is too
  much.
• The next field contains this acknowledgement but
  it is slightly unusual. It does not contain the
  sequence number of the frame being acknowledged
  but rather the sequence number of the first frame
  not received.
• This allows several frames to be acknowledged at
  once.
• Supervisory frames are used to carry
  non-piggybacked acknowledgements.

8
X.25 Packet Switching

• The main recommendation regarding PSDNs is X.25.
  This recommendation describes a low cost,
  reliable network protocol.
• This protocol is designed to work over a variety
  of existing connections of varying quality.
• When a connection is initiated, an X.25 network
  will first establish a virtual circuit from
  source to destination.
• Messages are broken into data packets containing
  up to 128 bytes. These packets are transmitted
  across the network and are reassembled by the
  destination device.

9
X.25 Sending Packets

• Individual hosts are not normally connected
  directly to the X.25 network.
• Hosts are usually connected to a LAN.
• A special device called a PAD (Packet Assembler
  and Disassembler) is also connected to the LAN.
• Hosts send data remotely by first sending the
  data to the PAD which converts the data into
  packets suitable for transmission across the X.25
  network.
• The PAD takes the data and places it into the
  User Data field of X.25 packets.
• These packets are then passed on to the Data Link
  Control which encapsulates them into frames.
• The frames are then transmitted to the Data
  Circuit Terminating Equipment (DCE) in the X.25
  network.

10
X.25 Encapsulating Packets

• X.25 packets are encapsulated in the payload of
  data link frames.
• Typically Linked Access Protocol (LAP),
  LinkedAccess Protocol Balanced(LAPB) or High
  LevelData Link Control (HDLC)is used to
  transmit framesover individual links.
• The LAP, LAPB and HDLC protocols arevery
  similar to SDLCwith the exception ofa few extra
  commands that make them more suited to
  peer-to-peer communication.

11
X.25 Reference Model

• X.25 predates the OSI 7-layer model. This means
  that the layers in the X.25 reference model do
  not correspond exactly to the OSI model.
• In X.25, thetransport layerand above
  arelabelled as theuser definedprocess.
• The OSI networklayer, data linklayer
  correspondto the packetlayer, frame layerin
  X.25.

12
X.25 Performance

• X.25 networks were designed to work across a wide
  range of relatively noisy (high error rate)
  links.
• This is because X.25 was designed at a time when
  there were few standards in digital communication
  and the connections were of relatively low
  quality.
• X.25 pays the price in terms of header overheads
  and computational overheads. Each packet must be
  unpacked and repacked into frames at every node
  it passes through. Flow control and error
  checking are also performed at each node (at the
  frame level).
• Packets are routed at the network level. This
  enables the network to utilise different data
  link protocols working over different media.
• X.25 networks typically provide data rates of
  between 150bps and 56kbps.

13
Switched Multimegabit Data Service (SMDS)

• SMDS is designed to connect remote LANs.
• In situations where remote LANs might be
  connected using leased lines, SMDS offers a cost
  effective and secure alternative.
• SMDS provides a connectionless broadband data
  service (CBDS) which means that it uses packet
  switching and that the frames are modulated
  within the SMDS network.

14
SMDS Sending Packets

• Like the X.25 system, hosts are not usually
  connected directly to the SMDS network.
• Instead, hosts transmit their packets to an SMDS
  gateway, which converts them to SMDS packets and
  transmits them to the SMDS network via a MAN.
• SIP3 embeds the MAC frame between a standard
  header and trailer.
• SIP2 is the link access protocol for sending data
  over the MAN to an SMDS router.
• SIP1 sends the bit patterns over the MAN (e.g.
  DQDB).

15
SMDS Frame Format

• SMDS effectively acts as a packet delivery
  service.
• It takes a packet from one network, embeds it in
  an SMDS packet, transmits it across the SMDS
  network, extracts the packet from the SMDS packet
  and places it on the destination LAN.
• The source and destination addresses consist of
  up to 15 decimal digits, each stored in a 4-bit
  field (i.e. binary coded decimal). They are like
  telephone numbers.
• When an SMDS packet arrives at the first SMDS
  router, it checks the source address to make sure
  it corresponds to the line on which it came (this
  avoids billing fraud).

16
SMDS Features

• SMDS typically provides data rates of about
  45Mbps.
• The average data rate that can be transmitted
  depends on the level of service paid for. If
  this average data rate is exceeded by the
  customer, the excess packets are discarded by the
  SMDS network.
• SMDS customers can arrange to restrict the
  addresses to and from which packets can be sent
  and received.
• This prevents unauthorised access to other
  networks and also prevents unauthorised access
  from other networks.
• This security feature is important for companies
  and organisations dealing with confidential
  information.
• Special broadcast addresses can also be arranged.
  Packets sent to one of these addresses are sent
  to a list of predefined addresses.

17
Frame Relay

• Frame relay is a simplified version of X.25.
• Frame relay was designed to use modern (lower
  error rate) links. It was also designed to use
  the same data link protocol throughout the
  network (whereas X.25 could use many different
  data link protocols).
• Rather than routing data at the packet level,
  Frame relay routes data at the frame level.
  Frame relay also does not include error checking
  information.
• This means that there are fewer overheads
  involved in routing and error checking.
• Packets do not need to be extracted from frames.
  Instead the frames are routed without the need to
  look inside them.
• Error checking is performed by the communicating
  hosts.

18
Frame Relay Features

• Like X.25, frame relay uses a virtual circuit to
  transmit data. Unlike X.25, all virtual circuits
  are leased (permanently set up by pre-arrangement
  with the service provider).
• This means that there is no connection
  establishment delay.
• It also means that networks are more secure since
  a network can only be accessed from a small
  number of other networks.
• Frame relay provides an unreliable
  connection-oriented service that can carry data
  at rates of about 1.5Mbps.
• This is not as high as SMDS but frame relay is
  considerably cheaper.

19
Frame Relay Frame Format

• The frame relay frame format starts and ends with
  special flag bytes just as with SDLC.
• The header field consists of
• A 10-bit virtual circuit ID called DLCI (Data
  Link Connection Identifier).
• A 1-bit field called FECN (Forward Explicit
  Congestion Notification) that is set by the frame
  relay network to tell the destination host that
  the frame experienced congestion.
• A 1-bit field called BECN (Backward Explicit
  Congestion Notification) that is set by the frame
  relay network in frames travelling in the
  opposite direction. This tells the source host
  its frames are experiencing congestion.

20
Frame Relay Frame Format

• A 1-bit field called DE (Discard Eligibility)
  that is set by the source host to tell the
  network that a frame can be discarded if the
  network is low on resources.
• A 1-bit field called CR (Command/Response) that
  is not currently used but is reserved for future
  use.
• The data field carries the users data.
• The checksum is a 16-bit CRC.
• The notion behind FECN and BECN is that by
  informing the source and destination hosts of
  network congestion, high-level protocols will be
  able to take appropriate flow control actions.