Computer Networks Circuit Switching Packet Switching Switching Networks

Computer Networks Circuit Switching - Packet Switching

Switching Networks Ñ Long distance transmission is typically done over a network of switched nodes Ñ Nodes not concerned with content of data Ñ End devices are stations Ñ Computer, terminal, phone, etc. Ñ A collection of nodes and connections is a communications network Ñ Data routed by being switched from node to node

Nodes Ñ Nodes may connect to other nodes only, or to stations and other nodes Ñ Node to node links usually multiplexed Ñ Network is usually partially connected Ñ Ñ Some redundant connections are desirable for reliability Two different switching technologies Ñ Circuit switching Ñ Packet switching

Simple Switched Network

Circuit Switching Ñ Dedicated communication path between two stations Ñ Three phases Ñ Establish Ñ Transfer Ñ Disconnect Ñ Must have switching capacity and channel capacity to establish connection Ñ Must have intelligence to work out routing

Circuit Switching - Applications Ñ Inefficient Ñ Channel capacity dedicated for duration of connection Ñ If no data, capacity wasted Ñ Set up (connection) takes time Ñ Once connected, transfer is transparent Ñ Developed for voice traffic (phone)

Public Circuit Switched Network

Telecomms Components Ñ Subscriber Ñ Devices attached to network Ñ Local Loop Ñ Subscriber loop Ñ Connection to network Ñ Exchange Ñ Switching centers Ñ End office - supports subscribers Ñ Trunks Ñ Branches between exchanges Ñ Multiplexed

Circuit Switch Elements

Circuit Switching Concepts Ñ Digital Switch Ñ Provide transparent signal path between devices Ñ Network Interface Ñ Control Unit Ñ Establish connections Ñ Generally on demand Ñ Handle and acknowledge requests Ñ Determine if destination is free Ñ construct path Ñ Maintain connection Ñ Disconnect

Blocking or Non-blocking Ñ Blocking Ñ A network is unable to connect stations because all paths are in use Ñ A blocking network allows this Ñ Used on voice systems Ñ Short duration calls Ñ Non-blocking Ñ Permits all stations to connect (in pairs) at once Ñ Used for some data connections

Space Division Switching Ñ Developed for analog environment Ñ Separate physical paths Ñ Crossbar switch Ñ Number of crosspoints grows as square of number of stations Ñ Loss of crosspoint prevents connection Ñ Inefficient use of crosspoints Ñ All stations connected, only a few crosspoints in use Ñ Non-blocking

Crossbar Matrix

Multistage Switch Ñ Reduced number of crosspoints Ñ More than one path through network Ñ Increased reliability Ñ More complex control Ñ May be blocking

Three Stage Switch

Time Division Switching Ñ Partition low speed bit stream into pieces that share higher speed stream Ñ e. g. TDM bus switching Ñ based on synchronous time division multiplexing Ñ Each station connects through controlled gates to high speed bus Ñ Time slot allows small amount of data onto bus Ñ Another line� s gate is enabled for output at the same time

Routing Ñ Many connections will need paths through more than one switch Ñ Need to find a route Ñ Efficiency Ñ Resilience Ñ Public telephone switches are a tree structure Ñ Static routing uses the same approach all the time Ñ Dynamic routing allows for changes in routing depending on traffic Ñ Uses a peer structure for nodes

Alternate Routing Ñ Possible routes between end offices predefined Ñ Originating switch selects appropriate route Ñ Routes listed in preference order Ñ Different sets of routes may be used at different times

Alternate Routing Diagram

Principles of Packet switching Ñ Circuit switching designed for voice Ñ Resources dedicated to a particular call Ñ Much of the time a data connection is idle Ñ Data rate is fixed Ñ Both ends must operate at the same rate

Basic Operation Ñ Data transmitted in small packets Ñ Typically 1000 octets Ñ Longer messages split into series of packets Ñ Each packet contains a portion of user data plus some control info Ñ Control info Ñ Routing (addressing) info Ñ Packets are received, stored briefly (buffered) and past on to the next node Ñ Store and forward

Use of Packets

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

Switching Technique Ñ Station breaks long message into packets Ñ Packets sent one at a time to the network Ñ Packets handled in two ways Ñ Datagram Ñ Virtual circuit

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

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

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

Packet Size

Circuit v Packet Switching Ñ Performance Ñ Propagation delay Ñ Transmission time Ñ Node delay

Event Timing

External and Internal Operation Ñ Packet switching - datagrams or virtual circuits Ñ Interface between station and network node Ñ Connection oriented Ñ Station requests logical connection (virtual circuit) Ñ All packets identified as belonging to that connection & sequentially numbered Ñ Network delivers packets in sequence Ñ External virtual circuit service Ñ e. g. X. 25 Ñ Different from internal virtual circuit operation Ñ Connectionless Ñ Packets handled independently Ñ External datagram service Ñ Different from internal datagram operation

Combinations (1) Ñ External virtual circuit, internal virtual circuit Ñ Dedicated route through network Ñ External virtual circuit, internal datagram Ñ Network handles each packet separately Ñ Different packets for the same external virtual circuit may take different internal routes Ñ Network buffers at destination node for re-ordering

Combinations (2) Ñ External datagram, internal datagram Ñ Packets treated independently by both network and user Ñ External datagram, internal virtual circuit Ñ External user does not see any connections Ñ External user sends one packet at a time Ñ Network sets up logical connections

External Virtual Circuit and Datagram Operation

Internal Virtual Circuit and Datagram Operation

Routing Ñ Complex, crucial aspect of packet switched networks Ñ Characteristics required Ñ Correctness Ñ Simplicity Ñ Robustness Ñ Stability Ñ Fairness Ñ Optimality Ñ Efficiency

Performance Criteria Ñ Used for selection of route Ñ Minimum hop Ñ Least cost Ñ See Stallings appendix 10 A for routing algorithms

Costing of Routes

Decision Time and Place Ñ Time Ñ Packet or virtual circuit basis Ñ Place Ñ Distributed Ñ Made by each node Ñ Centralized Ñ Source

Network Information Source and Update Timing Ñ Routing decisions usually based on knowledge of network (not always) Ñ Distributed routing Ñ Nodes use local knowledge Ñ May collect info from adjacent nodes Ñ May collect info from all nodes on a potential route Ñ Central routing Ñ Collect info from all nodes Ñ Update timing Ñ When is network info held by nodes updated Ñ Fixed - never updated Ñ Adaptive - regular updates

Routing Strategies Ñ Fixed Ñ Flooding Ñ Random Ñ Adaptive

Fixed Routing Ñ Single permanent route for each source to destination pair Ñ Determine routes using a least cost algorithm (appendix 10 A) Ñ Route fixed, at least until a change in network topology

Fixed Routing Tables

Flooding Ñ No network info required Ñ Packet sent by node to every neighbor Ñ Incoming packets retransmitted on every link except incoming link Ñ Eventually a number of copies will arrive at destination Ñ Each packet is uniquely numbered so duplicates can be discarded Ñ Nodes can remember packets already forwarded to keep network load in bounds Ñ Can include a hop count in packets

Flooding Example

Properties of Flooding Ñ All possible routes are tried Ñ Very robust Ñ At least one packet will have taken minimum hop count route Ñ Can be used to set up virtual circuit Ñ All nodes are visited Ñ Useful to distribute information (e. g. routing)

Random Routing Ñ Node selects one outgoing path for retransmission of incoming packet Ñ Selection can be random or round robin Ñ Can select outgoing path based on probability calculation Ñ No network info needed Ñ Route is typically not least cost nor minimum hop

Adaptive Routing Ñ Used by almost all packet switching networks Ñ Routing decisions change as conditions on the network change Ñ Failure Ñ Congestion Ñ Requires info about network Ñ Decisions more complex Ñ Tradeoff between quality of network info and overhead Ñ Reacting too quickly can cause oscillation Ñ Too slowly to be relevant

Adaptive Routing - Advantages Ñ Improved performance Ñ Aid congestion control (See chapter 12) Ñ Complex system Ñ May not realize theoretical benefits

Classification Ñ Based on information sources Ñ Local (isolated) Ñ Route to outgoing link with shortest queue Ñ Can include bias for each destination Ñ Rarely used - do not make use of easily available info Ñ Adjacent nodes Ñ All nodes

Isolated Adaptive Routing

ARPANET Routing Strategies(1) Ñ First Generation Ñ 1969 Ñ Distributed adaptive Ñ Estimated delay as performance criterion Ñ Bellman-Ford algorithm (appendix 10 a) Ñ Node exchanges delay vector with neighbors Ñ Update routing table based on incoming info Ñ Doesn't consider line speed, just queue length Ñ Queue length not a good measurement of delay Ñ Responds slowly to congestion

ARPANET Routing Strategies(2) Ñ Second Generation Ñ 1979 Ñ Uses delay as performance criterion Ñ Delay measured directly Ñ Uses Dijkstra� s algorithm (appendix 10 a) Ñ Good under light and medium loads Ñ Under heavy loads, little correlation between reported delays and those experienced

ARPANET Routing Strategies(3) Ñ Third Generation Ñ 1987 Ñ Link cost calculations changed Ñ Measure average delay over last 10 seconds Ñ Normalize based on current value and previous results

Required Reading Ñ Stallings chapters 9 and 10 Ñ ITU-T web site Ñ Telephone company web sites (not much technical info - mostly marketing)