Taking these in turn:
These are the traditional data services provided by the PTOs, providing transparent circuits for voice and data multiplexer links, video codecs, mainframe to terminal connections etc. Typically, they are time critical data streams with data clocks being maintained across the network.
These services may be of a bursty nature, but the data must arrive at the other end of the link in a well - ordered fashion, and must not experience large differences in the delay across the link. Typical services would be digital voice and digital video, even though they are usually transmitted synchronously.
Such service is typically that required by client workstations communicating with a file server. Typically, the data will be transmitted in bursts of high data volumes. Unlike other services, however, it does not matter particularly (to the carrier!) when the data actually arrives - long gaps in reception of up to several hundred milliseconds can usually be tolerated.
To handle all these services, the PTOs required a new form of time division multiplexing, that could handle both time critical and non - time critical data in a way which would maximise use of the available bandwidth between locations. The technical solution was ATM. ATM is based on the concept of virtual circuit switching, in contrast to data networks such as Ethernet, Token Ring and FDDI which use packet switching, and telephone networks which use physical circuit switching.
ATM transmits data in "cells". A cell is 53 bytes long, and consists of two parts - a header and a payload. This form of packet structure is used in conventional packet switching, but there are major differences between the two technologies: an ATM cell does not have a source and destination address, and an ATM cell header has practically no protocol overhead. The header simply contains a virtual channel identifier (VCI), and, if required, a virtual path identifier (VPI). The method of operation is as follows:
An ATM connection is set up by a pre - communication session between the ATM terminal equipment and the ATM communications equipment (known as an ATM switch, for reasons which will become clear). Sessions between ATM devices start by the ATM network itself finding out where the required destination address is - that is, the routing of an ATM session is found once only, at call set up time, just as the routing of a telephone call is only made once at the start of the call. ATM addresses are based on a combination of the E.164 ISDN international "telephone" number and a 6 byte End System Identifier (ESI) which corresponds to the MAC address of a local area network device.
The routing information is then given a virtual channel identifier (a VCI/VPI combination) by the ATM network. Each ATM switch simply looks at the VCI/VPI, and does a wire - speed lookup to see which port that cell should be switched to. There is no packet processing, no attempt to decode protocol layers or provide filters - the routing from switch to switch is simply done on the VCI. (The virtual path identifier or VPI acts as a simplification, when several VCIs share common routes) The actual switching is done in hardware, not software, which means that the achievable speeds are very high, - in equipment available today, a 16 port switch can bidirectionally switch data on any port to any other port at 155 Mbps, giving a total switch capacity of 2.5 Gbps (eight full duplex links at 155 Mbps in both directions)
The other advantage of using hardware switching is that the delay, or latency of the network can be made very low indeed, which improves the data network performance and enables use of ATM on time critical services.
Following telephony practice, a network of ATM switches can be built. Interconnections between ATM switches are simple - any port of any ATM switch can be connected to any other ATM switch. If more bandwidth is required, then several ports are connected, as the routing algorithms will automatically find the least used route between switches.
The connection between ATM switches is thus a "mesh" network - it can be built up as required, and does not demand any special topology. This is one of the most attractive features of ATM - it is a solution which can grow with the demand, and has no fixed limits on the data transmission capacity.
The use of cells to transmit data does not mean that the protocols of today are not used: these would be embedded in a stream of ATM cells. ATM is totally protocol transparent - the cell payloads are passed by ATM switches without even being "read" by the switches at a binary level.
ATM uses the concept of end - to - end flow and error control, in contrast to a conventional packet switched network, which uses internal flow and error control. That is, the network itself does not check the payload data for errors, but leaves this up to the end terminal device. (In fact, the only error checking on the cells is on the header, so that the integrity of the VCI/VPI is ensured). The reason for this is that the quality of data transmission has dramatically increased in the past few years, to a point where the error control mechanisms of the "traditional" line protocols such as HDLC are really no longer necessary. Similarly, because the switches do not store and forward the cells, but switch them "on the fly" at wire speed, there is no flow control in the network.
If you have found this summary useful, please feel free to contact K-Net Ltd (atm-info@k-net.co.uk) about any ATM related issue. We manufacture a unique range of ATM Video Codecs which are compatible with all standards compliant ATM switches. In addition, we resell a complete line of ATM equipment from leading vendors such as Fore Systems, NetEdge Systems and Fibercom, allowing us to build total ATM solutions
Return to home page