Professional Documents
Culture Documents
Introduction
Managing a wide variety of traffic over a network is a constant challenge to public and private network operators alike. There are different traffic characteristics and network performance requirements that must be addressed, as well as the need for networks to efficiently support multiple applications using multi-vendor network equipment. ATM networks are, by their very nature, designed to carry traffic generated by a wide variety of applications. To efficiently support multiple services in a potentially multi-vendor network environment, the ATM Forum has specified traffic management functions and procedures - such as traffic contract, conformance, policing, shaping, connection admission control, flow control and congestion control - by way of its TM4.0 specification, which was approved in April 1996. The main aims of the TM4.0 ATM traffic management specification are: Protect the network and users to achieve netowrk performance ogbjectives; Optimise the use of network resources; Protect applications against network congestion.
Diagram 1
CBR service
The CBR service is designed to support real-time applications (such as circuit emulation and voice), and therefore assumes a real-time delay constraint and delay jitter. The allocated rate is defined and the application is assumed to offer traffic constantly at this rate. The acceptable loss rate, maximum delay and jitter are also specified. The CBR traffic is deterministic and can be characterised by a constant cell inter-arrival which corresponds to a known peak emission rate.
VBR service
The VBR service is mainly used for video applications, frame relay, SMDS traffic, or any other application which has known or predictable bursty traffic characteristics. VBR traffic can be characterised by an ON/OFF source. During the ON period, the cells' inter-arrival is constantly based on a peak emission rate. No cells are generated during the OFF period. The duration of each ON and OFF period is also well characterised. VBR service is further divided into two subcategories, based on the delay requirements of the applications. The sub-categories are referred to as real-time VBR (RT-VBR) and non-real-time VBR (NRT-VBR).
ABR service
The ABR service is designed to implement a feedback mechanism which allows end systems to adapt their sending rate to the congestion status of the network, thereby minimising cell loss. The feedback is conveyed via a control loop which originates from an end system. In the case of ATM, the end system's response to the feedback is specified as part of the protocol. The traffic characterisation of an ABR connection is dynamic. An ABR connection can be allocated/reserved bandwidth corresponding to minimum cell rate (MCR) and extra bandwidth is made available via the flow control protocol.
UBR service
UBR service does not include bandwidth allocation and does not mandate the enforcement of a traffic parameter. All UBR connections share the remaining bandwidth with no specific control mechanism at the ATM layer. With a UBR connection, there is no guaranteed QoS, and the user must tolerate whatever capacity and cell loss the network can provide at the instant the cell goes through the network. End-to-end protocols used in the Internet, such as TCP/IP, regulate the flow of traffic based on packet loss. This type of traffic can run efficiently using the UBR service enhanced with a frame discard congestion control technique.
Traffic descriptors
The ATM Forum's TM4.0 specification defines source traffic descriptors for each connection. These represent the connection's traffic characteristics, and include: Peak cell rate (PCR); Sustainable cell rate (SCR) and maximum burst size (MBS); Minimum cell rate (MCR); Cell delay variation tolerance (CDVT). Peak cell rate represents the peak emission rate of the source. The inverse of the PCR represents the theoretical minimum inter-arrival time of cells at a queuing point. PCR can be limited by the physical speed of the source or via ingress traffic shaping. To allow for statistical gains in the bandwidth allocation, connections may specify a sustainable cell rate. The maximum burst size parameter represents the connection's burstiness factor. It specifies the maximum number of cells that can be transmitted at PCR while still being compliant to the negotiated SCR. Theoretically, the MBS represents the
averaging period over which to calculate SCR. For connections with dynamic bandwidth requirement (ABR), the minimum cell rate represents the minimum required bandwidth. The MCR may be specified as zero. The bandwidth available from the network may vary, but shall not become less than MCR. The cell delay variation tolerance parameter is used to characterise the delay jitter/cell clumping (CDV) that can be encountered by cells in the network. This parameter is used in conjunction with PCR and SCR monitoring to ensure that cells which were generated at the appropriate interval, but suffered a positive CDV, are still seen as conforming to the traffic descriptors. Essentially, CDVT allows some number of cells to be generated at link rate while still being compliant with the PCR descriptor.
1. The value of CLR is not signalled. CLR is low for sources that adjust cell flow in response to control information. Whether a quantitative value for CLR is specified is network specific.
Table 1
Diagram 2
Flow control
Flow control protocols are used to avoid congestion, react to impeding congestion and prevent congestion collapse. The TM4.0 specifies a flow control protocol for the ABR service. The specification is flexible and allows for various implementation options. In ABR, the source creates a control loop by generating resource management (RM) cells at a fixed interval of data cells (the default is every 32 data cells). The destination turns around the RM cells, inserts the congestion information, then sends the cells in the backward direction of the connection. The source adjusts the shaping rate of the virtual connection (VC) - up or down - based on the congestion information contained in the backward RM cells. Switches in the path indicate congestion status either via a single bit (binary mode) or by inserting an explicit rate value, which represents the share of the bandwidth allocated to the VC (explicit rate mode). The source adjusts the sending rate, taking into account both the binary and explicit feedback, and accounting for the potential loss of RM cells. The binary mode is backward-compatible with existing switching elements in the field since it is based on the EFCI bit, which is specified as part of the ATM Forum UNI 3.1[1] specification. One key parameter which impacts on the performance of the protocol is responsiveness. This is mainly driven by the feedback loop delay (transmission + propagation + queuing). The propagation delay becomes a significant factor when the protocol is used over a WAN. To alleviate potential performance degradation due to a long feedback loop, ABR defines the concept of virtual sources and virtual destinations VS/VD) which segment the feedback loop into sub-loops by assigning the functions of source and destination to intermediate nodes in the network. The coupling between adjacent loops is flexible and allows for product differentiation. VS/VD functions can be at the ingress and/or the egress of a node, and be fully coupled, or totally independent.
Diagram 3
Congestion control
The ATM Forum TM4.0 specification defines congestion control in terms of discard priority management and frame discard. Discard priority management allows network elements to discard CLP= 1 cells with higher probability than cells with
CLP=0. This only applies when the CLP significant mode is used. In the CLP transparent case, cells with CLP=1 have the same priority as cells with CLP=0. The frame discard mechanism consists of discarding complete AAL5-PDUs when congestion is impending. Frame discard improves the performance of the applications, and optimises network efficiency by preventing partial packets from using up network resources. These are potential methods of implementing the mechanisms to provide a range of efficiency and fairness. These methods of implementation are not subject to interoperability agreements (eg, ATM Forum specification).
Conclusion
The ATM Forum TM4.0 specification defines a set of interoperable functions and procedures that allow efficient management of multi-service networks. Using these functions, a network operator can define flexible competitive service offerings. References [1] ATM Forum. UNI 3.1 specification. March 1993. [2] ATM Forum. TM4.0 specification. April 1996. Click here to go back to my Home Page Enter your email address to receive email when this page is updated. Your Internet email address:
times.