Packet Switching

Since its introduction in the early 1970s, packet switching has received widespread acceptance. Public networks have been constructed in most developed countries and many developing countries. The internetwork ITU-T X.75 protocol provides for interlinking of national networks at an international level. The ITU-T X.25 Recommendation is the original standard for packet-switching architecture.

Packet switching has several advantages over conventional circuit-switched networks. The circuit-switched network maintains a fixed bandwidth between the transmitter and receiver for the duration of a call. Also, the circuit-switched network is bit stream transparent, meaning it is not concerned with the data content or error-checking process. This is not the case for packet switching, where bandwidth is allocated dynamically on an "as required" basis. Data is transmitted in packets, each containing a header that contains the destination of the packet and a tail, or footer, for error-checking information. Packets from different sources can coexist on the same customer-to-network physical link without interference. The simultaneous call and variable bandwidth facilities improve the efficiency of the overall network. The buffering in the system which allows terminals operating at different bit rates to interwork with each other is a significant advantage of packet switching. The obvious disadvantage is the extra dimension of complexity with respect to the switches and network-to-customer protocol.

Furthermore, in certain circumstances, packet switching has several advantages over other methods of data communication:

1. Packet switching might be more economical than using private lines if the amount of traffic between terminals does not warrant a dedicated circuit.

2. Packet switching might be more economical than dialed data when the data communication sessions are shorter than a telephone call minimum chargeable time unit.

3. Destination information is contained in each packet, so numerous messages can be sent very quickly to many different destinations. The rate depends on how fast the data terminal equipment (DTE) can transmit the packets,

4. Computers at each node allow dynamic data routing. This inherent intelligence in the network picks the best possible route for a packet to take through the network at any particular time. Throughput and efficiency are therefore maximized.

5. The packet network inherent intelligence also allows graceful degradation of the network in the event of a node or path (link) failure. Automatic rerouting of the packets around the failed area causes more congestion in those areas, but the overall system is still operable.

6. Other features of this intelligence are error detection and correction, fault diagnosis, verification of message delivery, message sequence checking, reverse billing (charging), etc.

Packet switching (B) compared
to Packet switching (A).

Packet networks

The intelligent switching nodes within packet-switching networks, in general have the following characteristics:

1. All data messages are divided into short blocks of data, each having a specific maximum length. Within each block there is a header for addressing and sequencing. Each packet usually contains error control information.

2. The packets move between nodes very quickly and arrive at their destinations within a fraction of a second.

3. The node computers do not store data. As soon as a receiving node acknowledges to the transmitting node that it has correctly received the transmitted data, by doing an error check, the transmitting node deletes the data.

For packet switching, a data switching exchange (DSE) is required, which is a network node interlinking three or more paths. Data packets move from one DSE to another in a manner that allows packets from many sources and to many destinations to pass through the same internode path in consecutive time sequence. This, of course, is in contrast to the circuit-switched network that seizes a specific path (link) for the duration of the message transfer.  The boundary of the network is usually defined as the point where the serial interface cable is connected to the DCE. This point is also often referred to as the network gateway

The X.25 protocol

ITU-T Recommendation X.25 is recognized by the data communications fraternity as one of the most significant milestones in the evolution of networking architecture. Within the X series of recommendations, X.25 specifies the physical, link, and network protocols for the interface between the packet-switching network and the DTE/DCE at the gateway.

The DTE is connected to the DCE, which is equivalent to the subscriber telephone connection to the central office. X.25 does not describe details of the DTE or the packet data network equipment construction, but it does specify, to a large extent, the functions the DTE and packet network must be able to perform. The X.25 Recommendation specifies two types of service that should be offered by a carrier: (1) virtual call (VC) service and (2) permanent virtual circuit (PVC) service.

Example of a Packet switching network

Virtual call service. Prior to the transmission of any packets of data in a VC service, a virtual connection must first be set up between a calling DTE logical channel and the destination DTE logical channel. This is analogous to placing a telephone call before beginning the conversation. Special packets having specific bit sequences are used to establish and disconnect the virtual connection. Having made the virtual connection, the two DTEs can proceed with a two-way data transmission until a disconnect packet is transmitted. The term virtual is used because no fixed physical path exists in the network. The end points are identified as logical channels in the DTEs at each end of the connection, but the route for each packet varies depending on other packet traffic activity within the network.

Where the TDM world had its limitations of the fixed bandwidth allocation, a newer service was created that allowed the bandwidth to be allocated on the fly. Instead of simply putting the data into a fixed time slot, the user data is broken down into smaller pieces called packets, each containing both the source and the destination addressing information, as well as other control functional information. When a user sends data in a burst, multiple packets will be generated and routed across the network based on the addresses contained in the packets. The network creates a virtual circuit from each source to each destination to keep track of the packets on each connection. This is a different form of multiplexing, more of a statistical time-division multiplexing scheme. STDM uses the analyses of the past users to allocate more interleaved packet slots to the heavier users and less interleaved slots to the lighter users. The major drawback to this scheme is the penalty paid in "speed" of delivery. Where guaranteed data delivery and integrity was a prerequisite for the development of the X.25 networks, the delay in processing the data across the network was the price we paid. Clearly, something had to be done to overcome these limitations.

What the standards bodies attempted to accomplish was to use the best of all worlds. Taking the features of the switched network and the packet arrangements, the network arrived at a frame relay service that should meet the needs of the user. This comparison took into account the following service connections:

" Speed The speeds available to the user are important. As the capabilities of the dial-up digital network are now supporting the high-speed connections of up to 2.048 Mbps, the X.25 network only provides speeds of up to 64 Kbps. Thus, the decision is to use a digital connection of up to 2.048 Mbps.

" Network set-up delay The call set-up time on a digital switched network is relatively low. This is also true for the X.25 network, for general terms. However, as network congestion builds up the switched network circuits of the X.25, the network can introduce extensive delays. Thus, the network suppliers were looking to implement a low-delay call set-up. The PVC allows for this because the call set-up time is eliminated.

Routing The routing of the switched network is fairly static. When one uses a switched call set-up, the routing is used once to establish the link. From the perspective of a network delay or a failed link, the user will have to reestablish the call through a new dial sequence. This is not as robust as expected. The X.25 network is robust, in that if a problem occurs on the link, the packet nodes in the network will immediately reestablish the connection with the next packet that runs across the network. Therefore, the suppliers and standards bodies opted for the robust and dynamic call set-up. The delay in this is very low.

• Signaling In the switched network design, the signaling is also static, in that it is used only in the initial set-up or tear-down of the call. Therefore, if the call is interrupted, the signaling disappears. In the packet-switched networks, the signaling is dynamic. The signaling is contained in every packet, thus it is easily reestablished. The standards bodies opted for the dynamic signaling arrangements of the packet-switched networks.

• Bandwidth Once again, the bandwidth consideration is important. In the circuit-switched networks the bandwidth is fixed. If you dial up a 56-Kbps circuit, you get a full 56 Kbps whether you need it or not. However, the variable length of a packet allows for the maximum of 56 or 64 Kbps, but you can use much less. The disadvantage here is that the speed cannot exceed the range of 56 or 64 Kbps. The robustness rests in the packet-switched service as opposed to the circuit-switched world. Therefore the standards bodies opted to arrive at a dynamically allocated bandwidth on demand concept.

• Costs Circuit-switched services are relatively low priced. The costs per minute for 56, 64, 128, 384, or up to 2.048 Mbps are very reasonably priced. Pricing approximates $0.50 per minute for a switched service, and a Tl switched service is dropping well below the costs that you might expect, putting pressure on the costs of a dial-up analog service. In the packet-switched services, the costs are very low, priced at a cost per kilo-packet. There are numbers that approximate $0.23 per kilopacket. This is extremely low in terms of the dial-up network. The only issue here is what the best method of call processing will be if frames are routed across a PVC.

Unfortunately, this one issue was the stumbling block to wide acceptance and use of the frame relay networks. Pricing was introduced in various trends that went from low to very high. The standards bodies do not get involved with pricing issues. The network suppliers have changed their pricing plans, and are getting very aggressive. As a matter of fact, because frame relay falls under the auspices of a value-added service, it is not governed as a tariffed service. The network suppliers are now offering frame relay access and use on an individual case basis. They might require a nondisclosure agreement, before they will submit a proposal for pricing a frame relay network. This was initially a turn-off for customers who were interested in pilot programs introducing frame relay for the WAN connections. The problem was solved through the creative pricing arrangements, and now frame relay is becoming more readily accepted.