ATM BASICS - Chapter 4 docx

59 Chapter 4 ATM Services The chapter explains and summaries a range of concepts and mechanisms, which allow ATM for supporting of large variety of services. It presents the concept of Quality of Service in ATM from the services perspective. It describes QoS categories and traffic classes, QoS parameters as well as traf- fic parameters that are used in ATM devices. The chapter also explains basic traffic management mechanisms, which ensure that requested QoS is delivered to the customers. The reading of the chapter is advised prior to discussing ATM applications. Before presenting different ATM services and applications, one has to real- ize the great capabilities of ATM technology in terms of provisioning guar- anteed bandwidth and strict QoS conditioning. In fact whenever ATM ser- vices are discussed, the QoS objectives and mechanisms are concerned. ATM was designed to deliver the QoS in a way incomparable to any former tech- nology. The state-of-the-art transmission and switching constraints impli- cated a very detailed and coarse definition of the ATM QoS. The unique com- bination of capabilities and mechanisms allow for ensuring that all types of applications can be given a suitable set of parameters. Therefore, different services can be supported simultaneously with adequate QoS treatment. 4.1 Quality of Service Categories Since ATM supports transport of several different connection and traffic types, as well as service types, a service classification was developed. Initially, ITU-T took into account three factors that are used to differentiate between services. Service classes introduced by ITU-T are dependent on the timing relationship between source and destination, bit rate, and the con- nection mode. Next ATM Forum presented its own classification of so called QoS service categories. The relations between two classification models are given in the Table 4-1. ATM Basics 60 Table 4-1, QoS Categories and Service Classes ATM Forum defined also an updated version of UBR QoS category called Differentiated UBR. QoS mainly relates to the needs of the user for the particular application. Therefore, a number of categories have been defined. Voice services, for example, require a minimal end-to-end delay to minimize the need for echo cancellers. They also require a minimum level of delay variation. On the other hand, data services are more tolerant of delays and delay variation but can hardly tolerate any data loss. The ATM Forum QoS categories are list- ed here: •CBR (Constant Bit Rate) is intended to support constant bit rate connection-oriented traffic where end-to-end synchronization is required. This category is equivalent to ITU-T Class A performance requirements. No error checking, flow control, or other processing is done. This service should meet the current requirements for perfor- mance comparable to digital private line services such as E1. All real- time streaming applications can be considered as the CBR traffic. Hence, the CBR services are used to emulate synchronous or plesiochronous transmission. This capability is called Circuit Emulation Services (CES). •VBR-RT (Variable Bit Rate – Real-Time) has been defined to sup- port variable bit rate connection-oriented traffic where end-to-end syn- chronization is required. This is otherwise known as ITU-T Class B per- formance requirements. This service is intended for packetized video and voice applications, such as video conferencing systems. Compressed voice with silence detection/suppression can be given as the example of traffic that is well fitted into this category. •VBR-NRT (Variable Bit Rate – Non Real-Time) is for types of traf- fic which behavior is predictable, yet does not require as strict timing relationship to be maintained on end-to-end basis. This service can be used, for instance, for interconnecting LANs where multimedia email are exchanged and any variation in cell delivery times will be eliminat- ed before the email is viewed. •ABR (Available Bit Rate) service is designed for economical support of applications with vague requirements for throughputs and delays. It has been introduced and covered under the UNI V4.0. ABR has an embedded congestion control mechanism that is based on the feedback control. In result the temporary volume of traffic can be adapted accord- Chapter 4 61 ingly to the current network conditions. A typical example might be for use in a company, which requests for the guaranteed minimum load dur- ing a day and which wishes to benefit from additional capacity during idle periods of the day. •UBR (Unspecified Bit Rate) it’s the ATM version of the ‘best effort’ service, with no reservation of bandwidth and no QoS parameters guar- anteed. All UBR cells are accepted and if there is capacity left, they are delivered. Under congestion conditions, UBR cells are discarded, without any notification sent to the sender. Signaling used to set up and clear down calls is normally transmitted as UBR, as is Local Area Network Emulation (LANE) traffic. Recently the definition of UBR has been updated with Differentiated UBR, which is the ATM Forum response to the growing popularity of DiffServ model deployed in IP networks. •GFR (Guaranteed Frame Rate) is a service category that was defined in the late 90’s. It is intended to provide a mechanism that will make certain guarantee at the frame level. What is guaranteed is a frame rate rather than a cell rate. This approach was thought to offer an efficient tool for transmission of network layer packets (e.g. IP packets). Note, that AAL discussed in the previous chapter are correlated with service classes and QoS categories. In fact different types of AAL were designed to be capable of serving traffic described by a specific service class. For instance Class A traffic (CBR QoS category) should be transported with the AAL 1 (as the requirement) or AAL 5 (as an option). ATM Basics 62 4.2 Quality of Service Parameters QoS in an ATM network is defined by the set of six parameters that char- acterize the performance of a given virtual connection. The parameters quantify the performance of the connection all the way across the ATM net- work, but excluding the end stations. Three of the six parameters are nego- tiable between the end-stations and the network as part of the traffic con- tract. The remaining three of the parameters is non-negotiable. 4.2.1 Negotiable QoS Parameters The values of the negotiable parameters are signaled during the setup process between user equipment and a switching device and further within the ATM network. The user can request for QoS specifying values for these parameters prior to the connection setup. The network may accept a new call and establish the connection. Alternatively, the network may reject the call due to the negative result of connection admission control process. One of the negotiable QoS parameters relates to the potential loss of trans- mitted cells. Cell loss occurs due to the buffer overflow and component or link failure. ATM switch buffers may overflow when several bursts of cells arrive simultaneously from different virtual connections. The Cell Loss Ratio (CLR) is defined as shown in the equation shown in the Fig. 4-1. Lost cells include cells that did not reach the destination, cells received with an invalid header, cells corrupted by errors. In brief, lost cells include only those which are sent but do not arrive within expected time. The total transmitted cells figure in the equation is the total number of con- forming cells, that is, cells that conform to the traffic descriptor. Cells that are lost due to violation of their traffic descriptors are not counted when cal- culating the CLR. Hence, only cells, which were admitted to the network, are counted. The CLR, which in UNI 4.0 became a negotiable parameter, has a value in the range 10 -15 to 10 -1 in increments of 10 -x . Chapter 4 63 For many services it is extremely important that the delay on end-to-end basis is minimized and has a stable value. Although ATM is an asynchro- nous transmission method, it can monitor and guarantee the delay for the transmitted traffic. The second negotiable QoS parameter, Cell Transfer Delay (CTD) is influenced by a number of factors. The factors include prop- agation delay, switch design, buffer capacity and traffic load. An application may require a service that can set an upper limit on the value of CTD. This limit is referred to as Maximum Cell Transfer Delay (maxCTD), which can be requested in the range of 10 ms to 16,777,215 ms with increments of 10 ms. Please note that CTD is a variable that changes its value randomly with every cell received. ATM Basics 64 Fig. 4-1, Cell Loss Ratio The control of the maxCTD parameter does not ensure that every real-time application can operate properly. Some services and applications require that the delay the delay should not change its value during transmission of sensitive data. For instance, the transmission of voice services over packet switched networks is possible as long as the end-to-end delay fluctuates within defined limits. The major source of variation of the CTD is related to one of the fundamental ATM mechanisms - multiplexing of cells from dif- ferent users. The processes of buffering and switching inevitably introduce certain level of uncertainty in terms of delivery time for a single cell. Chapter 4 65 Fig. 4-2, Cell Delay Variation Consequently, there was a need for the definition of another QoS parame- ter: the Peak-to-Peak Cell Delay Variation (peak-to-peak CDV). It is the dif- ference between the maximum CTD and the minimum CTD. Note that the minimum CTD is made up of the fixed-delay components, that is, the unavoidable delays such as a propagation, transmission and switching delay. Peak-to-peak Cell Delay Variation, has been a negotiable parameter since UNI 4.0 was introduced, is mainly affected by the ATM switch design, buffer capacity, traffic load and the number of nodes in a system. CDV is measured in the range 10 ms to 655350 ms, with an increment of 10 ms. The relation between maxCTD and peak-to-peak CDV is explained in the Fig. 4-3, where the cell arrival distribution as a function of arrival time is presented. As can be easily noticed, for a given source, destination and route, some minimum (fixed) delay always exists due to the physical char- acteristics of the transmission path. However, only some cells make it in the minimum time. Most of the cells, according to the probability density func- tion are delivered within ranges set by the value of the maxCTD. But the probability density function says that it is also possible that some cells can reach the destination later than it is desired. Such cells will not be counted as delivered properly and they are treated as lost cells. By choosing a value of CTD, the customer and the carrier make an agreement on how late a cell can be delivered and still count as a correctly delivered cell. Normally, CDV is chosen so that, the fraction of cells that rejected for being too late be on the order of 10 -10 or less. ATM Basics 66 4.2.2 Non-negotiable QoS Parameters The remaining three QoS parameters refer to the characteristics of the vir- tual connection in terms of errors that can be caused at the time of trans- mission and switching processes. The parameters’ definitions are given in the Fig. 4-4. Chapter 4 67 Fig. 4-3, The relation between CTD and CDV CER (Cell Error Ratio) is defined as the number of ‘errored’ cells over the number of successfully delivered cells plus the number of errored cells. An errored cell is one that has been modified incorrectly, in such a way that it cannot be efficiently recovered by error correction mechanisms. The CER ratio is affected by the characteristics of the physical media being used. Hence, CER has typically lower value if the physical medium is fiber as opposed to a copper wire. SECBR (Severely Errored Cell Block Ratio) is the fraction of N-cell blocks of which M or more cells contain an error. Hence, it is defined as the num- ber of severely errored cell blocks over the total transmitted cell blocks. A cell block is the number of user cells transmitted between successive O&M cells. The knowledge of this parameter can be helpful when there is a need to investigate the nature of errors. ATM Basics 68 Fig. 4-4, Non-negotiable QoS parameters [...]... should decrease or disappear This approach 76 Chapter 4 has a number of constraints such dependence on the round trip delay and customer behavior However, when combined with other traffic management mechanisms it may come out as the sufficient solution 4. 4.7 Rate-based Congestion Control in ABR Service Both CBR and VBR traffic sources are of real-time or semi-real time nature For these traffic types it... ATM Basics the connection is established, the network permanently checks if the traffic contract is not broken by the user equipment and ensures that the connection does not interfere with other services As can be easily noticed, traffic management in ATM is not restricted to only one function In fact several different mechanisms are needed to provide QoS 4. 4.1 Connection Admission Control In high-speed... cells, but it is treated in a special way along the route When it reaches the destination, it is examined, updated, and sent back to the source The path for RM cells is presented in the Fig 4- 7 77 ATM Basics Figure 4- 7 , The path taken by RM cells in ABR operation The congestion control is based on the idea that each sender has a current rate ACR (Actual Cell Rate) that is greater than MCR but lower than... a specific algorithm 73 ATM Basics called the Generic Cell Rate Algorithm (GCRA) The objective for the GCRA is to check every cell to see if conforms to the parameters for its virtual connection The ATM Forum documents propose the two possible implementations of the GCRA: virtual scheduling algorithm and a leaky bucket algorithm The latter algorithm is presented in the Fig 4- 6 GCRA has two parameters:... incrementing factor, which is the maximum allowed arrival rate (PCR in case of CBR) and a limiting factor, which is the amount of variation herein that is tolerable (CDVT in case of CBR) Fig 4- 6 , The leaky bucket algorithm 74 Chapter 4 In normal conditions, all the cells transmitted at PCR enter the leaky bucket algorithm As long as the minimum cell arrival time (the reciprocal of PCR, T = 1/PCR) is equal to the... conformance checking of the SCR Each type of AAL uses its own set of traffic parameters The relation between QoS Categories and traffic parameters is given in the Fig 4. 5 Fig 4- 5 , Relations between QoS categories and traffic parameters 70 Chapter 4 There is also a traffic parameter that describes to the capabilities of network devices The CDVT (Cell Delay Variation Tolerance) informs how much variation will... connection The non-negotiable parameters mentions are not signaled between user and ATM network Their meaning is descriptive, which means they can be measured, collected and analyzed, but they cannot be requested from the ATM network as QoS objectives 4. 3 Traffic Descriptors When a virtual circuit is established, both the user (typically the process in the user device) and the network operator (the ATM network... applications There is no standardized mechanism for traffic shaping In practice shaping needs only to take place once, and that is at the customer premises equipment (CPE) 4. 4.5 Cell Loss Priority Control As it was presented in Chapter 2, every ATM cell contains the CLP field in the header This field is used to implement a very rudimentary control mechanism If the CLP bit is set to1 (CLP = 1), the cell can... requested and try to establish the connection again The implementation of the CAC mechanism is not subject to standardization as it is internal to a switch and dependent on the carrier policies 72 Chapter 4 4 .4. 2 Resource Reservation and Management Resource reservation as well as resource management is the techniques of reserving resources in advance, usually at setup time Once the CAC process has accepted... substantial period In fact for VBR-RT service category there is a request to deploy two instance of CGRA One of them will test the traffic with incrementing factor of 1/PCR and limiting and CDVT, and the second will use accordingly 1/SCR and BT parameters In addition to providing a tool for conformance testing the GCRA can also shape the traffic and remove some of the burstiness 4. 4 .4 Traffic Shaping Workstations . single cell. Chapter 4 65 Fig. 4- 2 , Cell Delay Variation Consequently, there was a need for the definition of another QoS parame- ter: the Peak-to-Peak Cell Delay Variation (peak-to-peak CDV) UNI 4. 0 became a negotiable parameter, has a value in the range 10 -1 5 to 10 -1 in increments of 10 -x . Chapter 4 63 For many services it is extremely important that the delay on end-to-end basis. •VBR-RT (Variable Bit Rate – Real-Time) has been defined to sup- port variable bit rate connection-oriented traffic where end-to-end syn- chronization is required. This is otherwise known as ITU-T