ATM (Asynchronous Transfer Mode)

A follow-up to previous packet-based technologies (X.25, Frame Relay, and IP), ATM is a similar network technology based on transferring data in cells or packets - but is uses a fixed size for every cell.  In addition, the cell used with ATM is relatively small compared to units used with older technologies. The small, constant cell size allows ATM equipment to transmit video, audio, and computer data over the same network, and assure that no single type of data hogs the line. 

ATM vs Frame Relay

Like Frame Relay, ATM is a Layer 2 protocol, that uses PVC's to guide the packets of data from one end to the other, across a series of switches.  In addition, both protocols use fixed-size data units across the backbone.  ATM uses 53 byte cells from end-to-end.  

Frame Relay backbone SAR (Segmentation And Reassembly) Problem - Frame Relay uses standard Q.922 Annex D frames from the endpoints which are typically 2048 or 4096 bytes  -  but uses fixed-size sub-frames across the backbone which vary, depending on the manufacturer of the switches (Alcatel uses 132-byte sub-frames, for example).  Therefore the ingress edge switches on a Frame Relay network must fragment (segment) large frames to small frames, while the egress edge switches must defragment (reassemble) small frames into large frames.  With ATM, the ingress edge switch receives cells from the customer, and simply forwards those cells to the next switch.  The cell size is never changed, so there is no SAR problem within the backbone.  Therefore, ATM  is capable of much higher speeds than Frame Relay, since the edge switches do not have to fragment incoming ATM data .  .  .  it is all 53 byte cells !!

ATM customer end SAR Problem - at the customer end, the ATM cells are created, sent, and received.  Like Frame Relay frames, ATM cells are used to encapsulate data from higher-level applications, such as IP.  Therefore both protocols must fragment packets if the packets exceed the basic transport data unit size.  But ATM uses small cells  (53 bytes), so it must encapsulating virtually any Layer 3 protocol.  FR, on the other hand, can use large frames of up to 4096 bytes, and the frame size is not fixed.  Therefore Frame Relay can encapsulate most protocols without fragmenting.

For example, the most common application for ATM is transport betweeb two Ethernet LAN's via IP over ATM.  Since layer 2 Ethernet packets max out at 1518 bytes, even when they are encapsulated within layer 3 IP packets, the size 


ATM is a Layer 2 protocol, and comparing it to IP (which is Layer 3) is a bit like comparing Apples and Oranges.  Nevertheless, a continuing debate has ensued over which protocol is "better".  The IP contingent complain that ATM is inflexible, and is not inherently routable .  ATM proponents counter that IP has never successfully offered COS (Class Of Service), and point out that RSVP has been a failure.  

Many providers  run IP across their ATM backbone, and have implemented MPLS to monitor and reroute congested paths.  

Some providers offer a "pure" IP network and it is built "ahead" of traffic demands.  Since additional bandwidth is available for traffic spikes, MPLS is not needed.  This also frees up the pure ATM network for dedicated ATM customer traffic, since it does not have to support IP traffic.  

Common implementations of ATM support data transfer rates of from 25 to 622 Mbps (megabits per second). This compares to a maximum of 100 Mbps for Ethernet, the current technology used for most LANs.