1. William Stallings Data and Computer Communications
Chapter 10
Packet Switching

2. Why Packet Switching? Using circuit switching for data Inefficient
Resources dedicated to a particular connection
Much of the time a data connection is idle
Data rate is fixed
Both ends must operate at the same rate
Limits the utility of high-speed workstations

3. Use of Packets

4. Advantages Line efficiency Data rate conversion
Single node to node link can be shared by many packets over time
Packets queued and transmitted as fast as possible
Data rate conversion
Each station connects to the local node at its own speed
Nodes buffer data if required to equalize rates
Packets are accepted even when network is busy
Delivery may slow down

5. Switching Technique Station breaks long message into packets
Packets are received, stored briefly (buffered) and past on to the next node on the network
Store and forward
Packets handled in two ways
Datagram
Virtual circuit

6. Datagram Each packet treated independently
Packets can take any practical route
Packets may arrive out of order
Packets may go missing
Up to receiver to re-order packets and recover from missing packets

7. Virtual Circuit Preplanned route established before any packets sent
Call request and call accept packets establish connection (handshake)
Each packet contains a virtual circuit identifier instead of destination address
No routing decisions required for each packet
Clear request to drop circuit
Not a dedicated path

8. Virtual Circuits vs. Datagram
Network can provide sequencing and error control
Packets are forwarded more quickly
No routing decisions to make
Less reliable
Loss of a node looses all circuits through that node
Datagram
No call setup phase
Better if few packets
More flexible
Routing can be used to avoid congested parts of the network

9. Packet Size

10. Circuit vs. Packet Switching

11. Routing
Finding a path through network nodes from source to destination
Characteristics of routing strategies
Correctness
Robustness
Stability
Fairness
Optimality
Effieciency (overhead for routing processing)

12. Performance Criteria Criteria used for selection of route Minimum hop
number of links
simplest
minimize the use of network resources used for routing decision
Least cost
minimize the total link cost
minimize delay
maximize throughput
minimize the monetary cost
more complex algorithms

13. Costing of Routes

14. Decision Time and Place
Routing decision time
Datagram (no virtual circuit)
for each packet
Virtual circuit
when the virtual circuit is established
route may change dynamically depending on current conditions

15. Decision Time and Place
Routing decision place
Which nodes are responsible for routing decision?
Distributed routing
each node decides the next one
most common
Centralized routing
designated central node (e.g. network control center)
not so robust since loss of this node may block the whole network
Source routing
routing decision is made by the originating station

16. Network Information Source and Update Timing
Routing decisions usually (not always) are based on knowledge of network (e.g. traffic load and link costs)
like flooding
Distributed routing
Nodes use local knowledge
cost of each outgoing link
May collect info from adjacent nodes
May collect info from all nodes on a potential route
Central routing
Collect info from all nodes

17. Network Information Source and Update Timing
When is network info updated?
No information used (fixed, flooding)
no updates
Local information used only
update is continuous
Other nodes’ info used
regular updates to adapt the route changing conditions
Tradeoff between good adaptive routing and efficiency

18. Key Routing Strategies
Fixed
Flooding
Random
Adaptive

19. Fixed Routing
Single, permanent route for each source to destination pair
Determine routes using a least cost algorithm (Bellman-Ford, Dijkstra - appendix 10A)
Route fixed, unless there is a change in network topology
link costs should be static (e.g. cannot be current traffic or load)
Simple, works well with reliable and stable load networks
Disadvantage: lack of flexibility

20. Fixed Routing Tables

21. Flooding No network info required
Packet sent by node to every neighbor
Incoming packets retransmitted on every link except incoming link
Eventually a number of copies will arrive at destination
Each packet is uniquely numbered so duplicates can be discarded at destination

22. Flooding Example

23. Flooding Precautions against unlimited grow in circulation
Nodes can remember packets already forwarded to keep network load in bounds
Include a hop count in packets.
Set to a maximum value (e.g. diameter of the network)
Decrease one at each hop
Discard when 0

24. Properties of Flooding
All possible routes are tried
Very robust
can be used for emergency messaging
At least one packet will use minimum hop count route
Can be used to set up virtual circuit
All nodes are visited
Useful to distribute information (e.g. routing info)

25. Random Routing
Node selects one outgoing path for retransmission of incoming packet
Selection is at random
equally likely
all outgoing links are utilized in long-run
can select outgoing path based on a probability
e.g. probability based on data rate
good traffic distribution
No network info needed
Route is typically not least cost nor minimum hop

26. Adaptive Routing Used by almost all packet switching networks
Routing decisions change as conditions on the network change
Failure
Congestion
Requires network info exchanged among nodes

27. Adaptive Routing - Drawbacks
Decisions more complex
Processing burden
Tradeoff between quality of network info and overhead
Overreaction may cause oscillations

28. Adaptive Routing - Advantages
Robustness
Improved performance
Aid congestion control
adaptive routing balances the loads

29. Classification Based on information sources
Local
rarely used
Adjacent nodes
All nodes
Route to outgoing link with shortest queue length
Good for load balancing, but may not be good for optimal routing
Can include bias for each destination (a value for negative preferrence)
outgoing link is determined by minimizing queue length plus bias

30. Isolated Adaptive Routing

31. ARPANET Routing Strategies
First Generation
1969
Distributed, adaptive
Bellman-Ford algorithm
Node exchanges delay vector with neighbors (every 128 ms)
Update routing tables based on incoming info
Doesn't consider line speed, just queue length (delay)
Queueing delay is not the whole delay

32. ARPANET Routing Strategies
Second Generation
1979
distributed, adaptive
uses “measured” delay as performance criterion
computes delay every 10 secs
informs all others (using flooding) if there is a significant change
After all nodes change their routing tables, the routing tables may become obsolete
causes oscillations (under heavy loads)
result: not all routes can get the best path under heavy loads

33. ARPANET Routing Strategies
Third Generation
1987
Link cost calculations changed
Measure average delay over last 10 seconds,
Transformed into link utilization
Normalized based on current value and previous results
Resulting estimated link utilization is then transformed into delay

34. Bellman-Ford Algorithm
Minimum cost paths
Iterative
find the shortest paths from a source to all possible destinations using only link
then using max. two links by adding appropriate links to the paths of step 1
then using max. 3 links on top of paths with two links
so on .. until no improvement is gained by adding more links
See example on board