Asynchronous Transfer Mode (ATM)
UMTS Radio Access Network uses ATM backbone (for carrying both signaling and user data). So let us have quick brush-up on ATM.
In typical (switched) digital network, multiple nodes are connected not directly but rather through a mesh of switching nodes. This backbone network is called switching network and it fills up the gap between network layer and physical layer. The data would be in the form of chunks (packets) and the switching network may provide connection oriented and/or connection less services depending on switching techniques used. The connection is more likely to be virtual in nature i.e. may not be mapped directly to "dedicated channel" in physical layer. In fact, one the main intentions of having switching network is efficiency. Consider internet traffic - it is bursty in nature whereas video and voice call traffic is likely to be low delay/constant bit rate in nature. Switching network should be able to support all types of traffic while efficiently using physical resources. Switching network achieves this by providing generic interface in terms of (virtual/logical) connections which higher layers can use as per their requirements. Multiplexing is the key here. Switching techniques should be able to multiplex (and de-multiplex) various types of traffic keeping their individual requirements in mind.
Frame relay is one such switching technique wherein data is transmitted in terms of variable size frames.
ATM is also a packet switching technique, but uses fixed size packets called cell; ATM is also known as cell switching technique. The cells are of size 53 bytes - 5 bytes for header and 48 for data. It uses statistical TDM operation i.e. instead of allocating time slots (cells) to a connection dedicatedly, cells are sent only when there is data to be sent. This way bandwidth can be allocated dynamically and with fine granularity.
As ATM network divides higher layer data in terms of cells with each cell having header, overhead is more compared to variable packet switching techniques. This looks interesting: ATM is still recommended for higher speeds. ATM is preferable when we want to high speed networks (like Broadband ISDN) based on high speed physical medias like optical fiber. The reason is: most of ATM cell switching is possible to be done in hardware due to fixed size of cells that is to be switched. This is not possible in other packet switching techniques where due to variable size nature of packets, software intervention is needed to and/or costlier hardwares are needed for switching.
Connections are managed with the help of virtual paths and virtual connections. Paths are configured, established, and released by network. A path consists of multiple connections. Connections can be made by user or network. Header field of cell contain identifier for virtual path (VPI) and virtual connection (VCI). VPI + VCI identify virtual channel. Switching nodes makes sure that cells are switched appropriately (i.e. transmitted to next switching node or user) with the help of mapping between incoming and outgoing paths and connections. Connections can be permanent, semi-permanent, or switched. Permanent connections are not established, but rather they are configured during subscription time. Switched connections require establishment procedure. Preassigned VCI values are used for basic/mandatory functions related to signaling and control.
Note that ATM do not distinguish between higher layer signaling traffic and user traffic. Both types of data are carried in cells including ATM's own signaling and control traffic.
ATM has three layers:
|ATM Adaptation layer (AAL)|
AAL has different types of interfaces from Type 1 thru 5, distinguished based on type of service parameters (like connection oriented/less, constant or variable bit rate etc.). AAL is responsible for converting higher layer data to cell format to transmit over ATM and other way round (segmentation and reassembly).
|AAL1||Used for real time data requiring constant bit rate. Provides connection oriented mode.|
|AAL2||Variable bit rate, but real time (like VoIP). Provides connection oriented mode. UTRA Iu-CS, Iur, and Iub uses it.|
|AAL3/4||Can be used for non real-time and variable bit rate data (bursty data). Cell loss should be least. Provide both connection oriented and connection less modes.|
|AAL5||Non real-time data and variable bit rate in connection less mode. Good for high speed IP traffic, less overhead and better error detection. UTRA Iu-CS, Iu-PS, Iur, Iub, and Iu-BC uses it.|
ATM layer provides header for cell. It takes care of multitude of functions like multiplexing, taking care of user requirements of QoS (when and how to send cells), mapping between virtual paths and connections, flow control etc. ATM header differ slightly for user-to-network interface (UNI) and network-to-network interface (NNI) by GFC (Global Flow Control). Other than GFC (for UNI), VPI, VCI, header contain fields: Payload type (3 bits, values - user data, OAM etc.), cell-loss priority (1 bit, 1 indicate that cell can be discarded in case of congestion, 0 indicate cell can be discarded when no alternative present), and error control field for header data (Header Control Field, HEC). HEC can correct single bit errors.
Physical layer does the actual transmission. ATM does not specify particular media for transmission, but high speed and high bandwidth medias (higher than T3/E3 speeds of 44.74 Mbps) are best suited. Normally optical fibers are preferred. Optical fibers have high bandwidth, no magnetic interference, no crosstalk between fibers, low attenuation (unlike wired medias). In North America, it is popularly known as SONET (Synchronous Optical NETwork) and internationally SDH (Synchronous Digital Hierarchy). Physical transmission is divided in 51.84 Mbps blocks, called OC-1 to 2488.32 Mbps (OC-48). OC-n is n times OC-1, this makes multiplexing and de-multiplexing easier.
Further reading would be: ATM switching matrix, Flow/Congestion control, LAN/IP over ATM etc.
|Copyright © Samir Amberkar 2010||§|