The Destination (and optionally the source) of a packet
switched call are identified by an address consisting of 12 decimal digits,
digits 1 to 4 are allocated by CCITT to identify the country and the network
within that country, for example:
234 identifies the UK
2342 identifies Public Network
2341 identifies International Public Network
2348 identifies UK telex
These 4 digits are known as the DNIC (Data Network Identification Code).
The remaining 8 digits identify the customers port, on Public Network there are 3 digits
to identify the Packet Switching Exchange (PSE) and 5 digits to identify the
port on that exchange.
This X.25 address, which is the packet switching equivalent of a telephone
number, is sometimes known as an NUA (Network User Address).
In addition there is an optional 3rd and 14th digit that is passed through
the network transparently and can be used by the customers equipment to identify
the system or task required, for example the digits could be used to specify
whether the user wishes to connect to the companies mailbox or the companies
database.
The 84 CCITT standard also specifies an address extension facility field which allows an OSI Network Service Access Point (NSAP) address which may be up to 40 decimal digits.
6.2 Network-Layer Addressing
By 1984 agreements were reached between international standards bodies, ISO
and CCITT, about an addressing scheme. This scheme relies on the concept of
the network service access point (NSAP), which is the interface address between
the network layer and the Transport layer of the OSI model. In addition, some
progress was made towards establishing an international registration authority
to control and administer the allocation of NSAP addresses, in a similar manner
to the way in which CCITT administers data network identification codes (DNICs)
to public networks.
Existing numbering plans such as X.121,F.69,E163, etc., as defined by CCITT for public networks, are considered inadequate for OSI addressing. They exclude private networks and LANs that do not attach to public networks and terminals on a private network, which can have more than one address assigned to them if that private network is attached to one or more public networks.
6.1 Slide Net 11 ................ Multiplexing and routing
10 - Routing algorithms
Deterministic routing
Shortest path algorithms
This approach is simple to implement and provides the minimum distance. However,
it does not account for network traffic or overloads on a link or a node.
Subject of research in graph theory. Dijkstras algorithm provides the shortest
path from a network node to every other node.
10.1 Floyd's Algorithm
This algorithm generates a series of matrices by successively including a
new node to provide the shortest path for each node pair. In a given matrix
Dk, the path between any two nodes uses only nodes 1 to k. Hence, true shortest
paths are obtained when k=N, where N is the total number of nodes.
10.3 Minimum delay Algorithms
10.4 - Reliability in deterministic routing
A reliable network should provide alternate routes for message transmission
in order to handle such conditions as link or node failure
10.5 flooding technique
Here a message arriving in a node is transmitted to all neighbouring nodes
or to a group of selected neighbouring nodes, the network is flooding with several
copies of messages. Duplicate copies are discarded at the destination node.
This technique is simple to implement, and the message almost always reaches
the destination node, however it is extravagant in use of network resources.
10.6 Disjoint Deterministic Routes
For reliability and to maximise backup capability alternative routes between two nodes should be arranged so that no link or node is shared by any two routes.
10.7 Adaptive routing strategies
10.8 Centralised Adaptive Routing
Adaptive routing, because of its requirement for adjustment to network conditions,
is difficult to manage from a central point. Parameters such as message queue
lengths and expected delays at each node have to be collected in the central
node. The inherent delay in receiving and disseminating this information
can result in addressing traffic conditions. However a centralised scheme is
feasible and desirable in some situations.
10.9 Random Routing
Random routing procedures determine routing to the next node based on a probability
distribution over the set of neighbouring nodes. For example, one random routing
policy sends the message on any one of the links with equal probability.
10.10 Adaptive Routing in ARPANET
Each node periodically estimates the minimum expected delay route to every
destination node and enters the corresponding outgoing link address in a table.
The destination node and its most desirable link address are stored in a routing
table. When a packet arrives in a node, it is sent on the link corresponding
to its destination node entry in the routing table. Every node transmits to
each of its neighbouring nodes an estimate of minimum delay for every destination
node. The delay of each node to itself is set to zero. The delay vector is stored
as entries in the delay table for the link on which it arrived.
10.11 Hierarchical Adaptive Routing
For a large number of nodes, network routing table sizes grow and require
large amounts of storage.
Network nodes can be divided into groups or clusters of nodes. Then, the routing table size in a node can be reduced by providing less information about nodes in other clusters.
10.12 slide net5 - making a call
* Level 3 shown in white
bar over top in trace
10.13 - Call packet - Addresses
byte 4 - Address
lengths
This byte contains 4 bits each containing the lengths of the
called and calling addresses as follows,
bit 8 7 6 5 4 3 2 1
msb lsb msb lsb
|<-- calling -->|<--- called --->|
address length address length
this means that although the calling address is first in the address field,
when the address lengths are displayed on a data analyser the called address
will be displayed first.
Byte 5 on
addresses, calling (to) address first, 12-14 BCD digits may be less for international calls.
called (from) address - optional, 0,12,13 or 14 BCD digits
When necessary, the address field shall be rounded up to an integral number
of bytes by inserting zeros in bits 4,3,2, and 1 of the last octet of the field.