2001 ALCATEL BELL N.V. ALL RIGHTS RESERVED VHBE Alcatel University. Page 1

Transcription

1 1 Page 1

2 ADSL coaching team Frank Vercamme Marc Debacker Werner Van Loock Brian Bogaerts Customer Documentation 2 Page 2

3 Intro to ATM: Overview 1. Why ATM? 2. ATM & ADSL 3. The ATM cell header 4. The B-ISDN Protocol Reference Model 5. QOS (Quality( Of Service) 3 Page 3

4 What is the similarity between a duck and ATM? ATM 4 Page 4

5 They are both built to go around with more (MULTI) media Air Ground Water Voice Video Data ATM 5 A duck can fly, walk and swim. We want to create a network capable of transporting voice, video, data, We also want to we want to introduce quality of service (QOS). Page 5

6 Intro to ATM: Some History Voice Network Voice Network Voice Network Analogue transmission Analogue transmission Digital transmission 6 For more than a century, the primary international telecommunication infrastructure has been the public circuit switched telephone system. This system was designed for analogue voice transmission and is inadequate for modern communication needs. In 1984 one agreed to build a new, fully digital, circuit switched telephone system. This new system, called (N-)ISDN has as its primary goal the integration of voice and non voice services and uses TDM (time division multiplexing) to achieve this.(tdm is not applicable for analogue transmission). It knew a slow growth. Main disadvantages: circuit switched and 64 kbps based. Page 6

7 The N-ISDN: Half - integrated PSTN N-ISDN Narrow (low bitrates) Half integrated 64 kbps based Internet Data Cable TV 7 -Today people want to use multiple (telecommunication) services at the same time. -The N-ISDN does not handle the needs of all telecommunication users (wishing to communicate voice, data, video,...) in a uniform way. -On one hand the user perceives the network as providing services in a more or less uniform way (i.e. an integrated) manner. -On the other hand once inside the network, dedicated networks handle different kinds of data. Examples: - a packet switching network - a frame relay network - a video conference network -... Page 7

8 The B-ISDN: Service independent - High (data) bitrates - Flexible PSTN B-ISDN High bitrates Multi service ATM based Cable TV Internet DATA 8 What do we want? To avoid the disadvantages of the N-ISDN, the idea merged to create a single, high capacity network, which is able to flexibly support all types of services (voice, data, video). Main characteristics: - Service independent network: The B-ISDN is multi service. It does not consider the contents of the ATM cells, in other words, the information of ATM cells is not examined inside the network. - High bitrate, broadband network: The network has a large capacity, it offers its services to many users and many applications. Typical bitrates of the network are 155 Mbps, 622 Mbps, or even more. - The network is capable of offering bandwidth usage on demand. It is suitable to carry information for fixed bit rate services (of any bitrate) and for variable bit rate services. - Future safe network. The network is able to adapt itself to changes in technology and user requirements. In the future new services can arise, which are unknown today.these services will all have to be treated by the same network. To realise such a B-ISDN network a new technology is necessary, which describes how the information has to be treated in that network. This technology is ATM (Asynchronous Transfer Mode) and it requires a new type of switches, called ATM switches. Usually the interconnections of the exchanges are realised by a digital transmission system, called SDH (Synchronous Digital Hierarchy). Page 8

9 Some History: : Circuit Switching - MRCS - FCS 125 µs Voice Network Signal Strength 1 timeslot per voice call occupation of a fixed bandwith (64 kbps) granularity & variable bit problem Hz not suited for (high speed) data traffic Voice frequencies are 300 to 3400 Hz Double sampling (Nyquist): 8000 Hz Each sample is 1 byte long = 8 bits 8 bits samples/s = 64 kbps Not complex but not flexible!! 9 When you or your computer places a telephone call, the switching equipment within the telephone system seeks out a physical `copper` (including fibre & radio) all the way from your telephone to the receiver`s telephone. This technique is called circuit switching. Once a call has been set up, a dedicated path between both ends exists and will continue to exist until the call is finished. The voice information of multiple sources is combined on 1 line using the technique of TDM. The transmission frame, which lasts for 125us, is divided into a number of slots. Each user is assigned to one timeslot for the duration of the communication, so there is a one to one relationship between a user and a timeslot. An important property of circuit switching is the need to set up an end to end path before any data can be sent (connection oriented). The elapsed time between the end of dialling and the start of ringing can easily be 10s. This results in no danger of congestion (once the call has been set up). Required bandwidth is statically reserved prior to releasing information into the network. Even when no bandwidth is needed the timeslot remains occupied. This leads to not being efficient. Since data traffic has a bursty nature CS is not flexible enough. In case we need e.g. 70 kbps one timeslot is not enough. Granularity problem: not all intermediate values of bitrates can be reached. Variable bit problem (bursty data traffic). What must you allocate for a variable bandwith source? The highest bitrate? Then you over-allocate. The average bitrate? Then you risk quality reduction. MRCS = multi rate circuit switching: allocation of n basic channels. FCS = fast circuit switching: for sources with fluctuating & bursty nature. No allocation of resource when not needed (fast associating signalling). Page 9

10 Some history: Packet Switching D D C Bandwith D C D C D Bursty nature of data transport time Packet Switching Network A Variable length packets per data call no occupation of a fixed bandwith no granularity & variable bit problem well suited for (high speed) data traffic B B A a lot of overhead (error control, identification, ) B A B A Flexible but complex!! 1 0 In packet switching the user information is transmitted inside variable length data packets. These data packets are transmitted whenever there is free capacity available in the network, which can be on any moment of time. Therefore different information cannot be identified anymore with respect to a particular timeslot. To handle this, all packets have to carry a header field, identifying them, and whose main purpose is to provide information for the routing of the packet. The average bitrate of a connection can be modified by sending more packets or packets with larger contents. When data packets arrive in a switch, error control is performed. To correct an erroneous data packet, retransmission of this information is necessary. Meanwhile the right sequence of the data packets that belong to the same packet has to be maintained (connectionless!). This means that a high amount of processing is required for every packet, and this in every switch within the network. This complexity results in delays and thus in a low throughput capability of the packet switching network. Page 10

11 Evolution of Transfer Modes Unflexible & Easy Make Flexible Multirate Circuit Switching ATM Circuit Switching Fast Circuit Switching Flexible & Complex Simplify Frame Switching Frame Relaying Packet Switching 1 1 ATM (= cell switching) combines the best of both CS & PS, i.e. it combines the simplicity of circuit switching & the flexibility of packet switching. This way ATM becomes `the` transfer mode suitable for transporting data at high speed. Page 11

12 Circuit & Packet Switching: properties RS 232 Circuit Switching Packet Switching ATM bit level no yes yes yes Basic Characteristics of ATM? cell level? no yes no yes connection level? N.A. connection level (Asynchronous Transfer Mode) No link by link error control ATM is connection oriented Cells with small, fixed size fields Transparent delivery of information yes no no Connection oriented? N.A. yes possible yes Connectionless? N.A. no possible no 1 2 ATM basic characteristics: Connection oriented: sequence preservation, service guarantee,faster routing, reduced header. Fixed cells: easy for buffers and memory Small cells: small delay and higher granularity The above table should be interpreted with care: Synchronicity at cell level might sound strange when discussing packet switching; we just want to indicate that the boundaries of a packet are not known in advance since the length of a packet might differ. Even so it is strange to talk about cells when considering serial communication (RS232). RS232 is not synchronous at bit level, that is why start and stop bits are present (to synchronise after periods of idle time). Examples of connection oriented packet switching: X25, Frame Relay, ATM is a special case of connection oriented packet switching. Page 12

13 Intro to ATM: Overview 1. Why ATM? 2. ATM & ADSL 3. The ATM cell header 4. The B-ISDN Protocol Reference Model 5. QOS (Quality( Of Service) 1 3 Page 13

14 13 Internet Access (ADSL) ISP1 POP Service Management Centre ATM NT NT I* Bus ADSL LT ADSL LT 155 Mbit/s ADSL mux Broadband POP PSTN PS ADSL modem Internet 1 4 Offering high speed Internet access, creating a high speed B-ISDN is not enough. We also need ADSL to ensure that the UTP (Unshielded Twisted Pair) over which the data (and voice) are sent does not become the bottleneck. Combining the B-ISDN and ADSL data can be sent at high speeds. Page 14

15 ANT : ADSL Network Termination Line 10 Base T ATM Forum 25,6 Mb/s Ethernet upstream buffer Round Robin Scheduling ATM-F upstream buffer Ethernet ATM-F 1 5 The ADSL modem, A(D)NT (ADSL Network Termination),ATU-R (ADSL Transceiver Unit at Remote location) or SpeedTouch modem has 2 interfaces. 10 Base T: Ethernet interface ATM Forum interface: ATM interface The ATM interface supports SVC (Switched Virtual Circuits), and keeps up the transparency of sending information, even in the ANT. Both interfaces can work simultaneously. If both interfaces are active at the same time, the upstream bandwith is 50% for each interface (Round Robin based multiplexing). Demultiplexing (downstream) is based on the VP/VC value. Page 15

16 Intro to ATM: Overview 1. Why ATM? 2. ATM & ADSL 3. The ATM cell header 4. The B-ISDN Protocol Reference Model 5. QOS (Quality( Of Service) 1 6 Page 16

17 The ATM cell header Header Payload 5 Bytes Bytes 2 formats : UNI or NNI 1 7 An ATM cell consists of 2 parts: header & payload. The header exists in 2 formats: UNI: User to Network Interface NNI: Network to Network Interface Page 17

18 The difference between UNI and NNI Public ATM network Private ATM network NNI UNI UNI UNI UNI UNI NNI UNI 1 8 Page 18

19 UNI format of the ATM - cell GFC:Generic Flow Control VPI:Virtual Path Identifier VCI:Virtual Channel Identifier 16 bit PT: Payload Type 3 bit CLP : Cell Loss Priority 1 bit 4 bit 8 bit GFC VPI VPI VCI PT HEC Payload (48) C L P Header HEC : Header Error Control 8 bit 1 9 The GFC mechanism assists in the control of the flow of traffic from ATM connections. More specifically, the GFC mechanism is used to control traffic flow in order to alleviate short-term overload conditions that may occur. (Stop & Continue messages with respect to receiving & retransmission buffers) GFC is only available at the user to network interface. Any end user- to - end user flow control requires additional network facilities. GFC is not supported in ADSL Page 19

20 NNI format of the ATM - cell VPI VPI:Virtual Path Identifier 12 bit VCI VCI:Virtual Channel Identifier 16 bit PT: Payload Type 3 bit CLP : Cell Loss Priority 1 bit PT HEC Payload (48) C L P Header HEC : Header Error Control 8 bit 2 0 Page 20

21 What means VP ( Virtual Path ), VC ( Virtual Channel )? User physical interface 1 VC for video 1 VC for french audio 1 VC for english audio 1 VC for spanish audio ATM network Virtual Path Connection Server 2 1 VP/VC forms a two-level identification; just like First/Sir Name or Street Name/House Number. To group a number of houses, family members or virtual connections one only uses the Street Number, Sir Name or VP value. Imagine one wants to look at a VOD (Video On Demand). Both the video and the audio signals are transported within the same VP pipe. This enables faster routing as the smaller pipes (VCs) do not have to be translated (considered). Typically there is a STB (Set Top Box) in between the ANT and the TV; a box where the audio signal (language) is selected and where the ATM cells are being translated into VHS signals. Page 21

22 What means VP ( Virtual Path ), VC ( Virtual Channel )? Users Concentrator ATMnetwork VP switch ISP 1 (Internet Service Provider) Internet Physical interface Virtual Path Connection 2 2 Page 22

23 Routing of the cells Each header of the cell contains the cell identification that routes the information towards the right destination Virtual path identification VPI 2 /9 6 /4 Virtual channel identification VCI 2 3 Page 23

24 ATM : Virtual Path AND Virtual Channel Virtual Channel identified by VCI Virtual Path identified by VPI Optical Fibre or Copper wire ATM cell VPI VCI Data Header Payload 2 4 The above picture is to visualize the meaning of VP/VC. It has got nothing to do with a physical representation. There are huge differences between the STM & the ATM mode: STM: - Reservation of physical channels - Fixed bitrate of channels - Low granularity of bitrates - Physical resources are permanently reserved, even if nothing is transmitted ATM: - Reservation of virtual channels - Variable number of channels - Variable bitrates of channels - High granularity of bitrates - Physical resources are not used, if nothing is transmitted Using PVCs (Permanent Virtual Connections) there is an `Always On` ATM connectivity. Since the VP/VC values are translated in every ATM switch using a fixed routing table, there is, at all times, a unique VP/VC path for every user. Page 24

25 The ATM switch Core-function : Switching the ATM cells as fast as possible 3/5 1/2 4/1 6/3 8/2 7/3 2 5 Inside an ATM switch there is a routing table. e.g.: 3/5 <-> 7/3 1/2 <-> 8/2 4/1 <-> 6/3 Since the ATM payload is transported transparently, only the header of an ATM cell is considered. Page 25

26 Local Significance of the VP/VC value 4/5 Port VP/VC Port VP/VC 1 2/9 2 4/5 1 6/4 3 2/9 Each ATM switch may change the VP/VC value. 2/9 6/ /9 The VP/VC value is unique on each physical interface 2 6 The VP/VC value must be unique on each physical interface in order to uniquely identify traffic on the respective virtual channels. Two adjacent, physically separated interfaces might have the same VP/VC. The fact that the VP/VC value on a particular interface has no global meaning is what we call local significance. Page 26

27 CLP : Cell Loss Priority Congestion CLP = 1 (low priority) CLP = 0 (high priority) 2 7 Filming a soccer game the camera might decide to mark the grass as low priority, since the player and the ball are far more important for the spectator to look at (easy to understand, more difficult to implement). In case of congestion some cells have to be dropped. At that point it is nice to know which cells are less important. Off course these cells are then dropped. Tagging (= increasing the CLP value) is also done by switches. When talking about UBR connections (see below) the switch will mark non guaranteed traffic as low priority traffic. GFR (Guaranteed Frame Rate) is another concept. Here there is intelligence at IP (TCP) level; it is preferable to drop two ATM cells belonging to the same IP packet rather than dropping two ATM belonging to different IP packets since an entire IP packet has to be transmitted when only 1 ATM (belonging to this packet) is discarded. Suppose that every 100 ATM cells 1 ATM cells is dropped; if IP packets contain (at average) more than 100 ATM cells no IP packet will ever get through. TCP will order a retransmission for every IP packet. Page 27

28 PTI : Type of payload ATM layer Maintenance cell PTI = 001 PTI = 000 Physical layer Empty cell TC PTI = Since we have to keep up the synchronicity at bit and cell level we cannot allow idle time in between 2 consecutive ATM cells. In case of idle time (no release of information) empty cells have to be inserted in order to get a continuous bit (cell) stream. In case no empty cells were inserted delineation (synchronization procedure) would have to take place more frequently. Maintenance cells (OAM) are inserted, e.g. to reconfigure an ATM switch without having to go into the field. The PTI consists out of 3 bits: A B C A: User Cell = 0; OAM cell = 1 B: No congestion = 0; Congestion = 1 C: Normal Cell = 0; Last Cell =1 The last cell indicator indicates the last ATM cell of an IP packet that was cut down into ATM cells. The congestion indicator is used to identify a congestion going on in the ATM network. However, indicating which ATM switch caused the the congestion is not known end to end. What about the VP/VC of a maintenance/oam cell? -maintenance cell header: VP/VC: 0/0 PTI: 100 CLP: 1 -empty cell header: VP/VC: 0/0 PTI: 000 CLP: 1 Page 28

29 Intro to ATM: Overview 1. Why ATM? 2. ATM & ADSL 3. The ATM cell header 4. The B-ISDN Protocol Reference Model 5. QOS (Quality( Of Service) 2 9 Page 29

30 The B-ISDN Protocol Reference - model Signalling Data Plane Management Layer Management Management plane Control Plane Higher layers User Plane Higher layers ATM Adaptation Layer ATM Layer Physical Layer Data, video, voice, signalling Conversion of information between the upper layers and the ATM cells Multiplexing and switching of cells Translation of the cells to a physical level 3 0 User plane: transfer of user data Control plane: transfer of signaling messages. Signaling is not applicable when talking about PVCs, since it will be the operators responsibility to adapt the routing tables within switches whenever a new connection has to be setup. You might compare it with the old telephone system where an operator had to make the connection for you. In the current telephone system signaling takes care of setting up the connection; within short notice ATM signaling will take care of setting up ATM connections and SVCs will become dominant. The only reason why SVCs are currently not much in use is because of complexity. Management plane: used to maintain the network and to perform operational functions. It will e.g. take care of OAM cells. Plane Management: responsible for the management of the different planes. Page 30

31 How are the connections set up in an ASAM? User plane AAL ATM PHY Connections realised by the operators ( PVC ) = Permanent Virtual Connection Connections realised via signalling ( SVC ) = Switched Virtual Connection ATM ATM PHY PHY ATM ATM PHY PHY ATM ATM PHY PHY User plane AAL ATM PHY ATM network ADSL releases 3.1 and lower = PVC ADSL releases 3.2 and higher = SVC 3 1 Only 2 layers in the protocol stack of an ATM switch. We assume the lines to be of good quality; so that we can skip link to link control. This means that the payload inside an ATM cell is not checked in an ATM cell. This is meant when talking about sending information in a transparent way. Today SVCs are not yet supported by the 10 Base T (Ethernet) interface of the ANT. Page 31

32 The ATM layers In the ADSL system following layers are most important : The physical ATM layer : Transmission convergence Physical medium The ATM layer : Setup of virtual connections Cell routing The ATM adaptation layer: Guarantees the Quality of Service Translates the userdata into ATM - cell stream 3 2 Page 32

33 Physical layer Transmission Convergence (TC) Insertion, extraction of empty cells HEC (Header Error Control) : cell delineation, detection of cell header errors Adaptation towards the de transmission system (SDH, PDH ) Physical Medium (PM) Transmission and reception of bits (NRZ) in function of the medium (electrical, optical) 3 3 Since we want to avoid error multiplication (routing mistakes because of VP/VC mistakes) we do perform an error control on the ATM cell header in every ATM switch. Page 33

34 Adaptation towards the transmission system (SDH, ADSL,...) Transmission Convergence (TC) POH Path OverHead (SDH overhead) ATM cells Trame SDH Physical Medium (PM) SDH example 3 4 Usually the interconnections of the exchanges are realized by a digital transmission system, called SDH (Synchronous Digital Hierarchy). ATM is often chosen to run on top of SDH because in case of a link failure redundancy might be present. This corresponds to the departure and arrival of trucks at fixed times. A truck leaves no matter if the truck is fully loaded or not. The trucks themselves are loaded asynchronously. Page 34

35 TC : HEC code calculation Header 5 Bytes 5 Bytes Header Error Correction (Payload) 48 Bytes 48 Bytes Correction of single bit errors Detection of multiple bit errors ATM cell 3 5 The HEC field (1byte) consists contains a CRC value. To avoid error multiplication the header is protected with a Cyclic Redundancy Check (CRC), which is transmitted in the HEC field (1 byte). This HEC field is checked in every ATM switch. The sending side calculates the HEC value: the polynomial generated by the first four bytes of the header is multiplied by 8 and then divided by another polynom. The remainder of this division is transmitted in the HEC parameter field. At the receiving side, the HEC value is recalculated to check for errors. The HEC code is capable to correct a single bit error and to detect multiple bit errors. This implies that there is only enough redundant information present inside the HEC field to correct single bit errors. As soon as multiple consecutive cells show header errors, correction is given up since this often indicates a burst error. In case of a burst error multiple bit errors are very likely and the detection of multiple bit errors is not 100% safe. Multiple bit errors might slip through. That is why cells are discarded in the latter case. Page 35

36 TC : Cell delineation Receive Payload HEC Header Payload HEC HEC Header Transmission Convergence (TC) bytes byte cell Payload HEC Header Payload HEC HEC Header HEC? 3 6 Cell delineation enables the receiver to recover the cell boundaries. We need to recover the cell boundaries to keep up the synchronicity at cell level. Since an ATM cell has a fixed length we do not have to repeat the cell delineation procedure once we are working synchronous at cell level. To obtain this cell delineation, the correlation between the first four bytes of the cell header and the HEC value is examined. If the calculated value differs from the one that was transmitted we shift bit by bit and perform a recalculation until the boundaries of the header are recovered. Once this is done we can shift cell by cell (working in a synchronous way at cell level). Page 36

37 TC : Cell Rate Decoupling ATM layer Maintenance cell Physical layer Empty cell TC 3 7 Maintenance & empty cells are removed as soon as possible. Page 37

38 ATM layer ATM layer Cell multiplexing / demultiplexing VPI/VCI translation Cell header generation /extraction Generic flow control at the UNI Physical layer 3 8 Page 38

39 AAL-5 characteristics Byte stuffing Error recovery Length of the information 5 48 bytes bytes Message bytes ATM - cells PAD 0-47 bytes. Length 2 2 bytes. bytes CRC 4 bytes Header cell with Payload Type = 0 Header cell with Payload Type = 1 (last cell of a message) 3 9 An example: an IP packet is cut into pieces (ATM cells). The last ATM cell will have a last cell indicator set to 1. Page 39

40 Intro to ATM: Overview 1. Why ATM? 2. ATM & ADSL 3. The ATM cell header 4. The B-ISDN Protocol Reference Model 5. QOS (Quality( Of Service) 4 0 Page 40

41 Quality of service Quality of Service (QoS) : a fundamental question for ATM networks... The quality of service guarantees how each circulating cell of a connection will be handled by the network An agreement is made between the supplier and the customer. ATM network User Each connection inside the ATM network has his own QOS 4 1 Page 41

42 parameters Contrat Descripteur de trafic PCR: MCR: SCR: CDV : 10 Traffic Descriptors A set of service categories The ATM Forum has defined a set of QOS 4 2 QOS is not present when talking about IP networks. ATM introduces QOS! Consider an ADSL connection (max downstream capacity of 8 Mbps). If father is watching a VOD mother would not want to disturb him by making a phone call (VoDSL), listening to MOD (Music On Demand) or by surfing the web. It is fundamental to allow only extra connections if the existing ones remain intact (will not suffer from others `logging on`). By defining service categories, described by traffic descriptors, the QOS of all active connections is guaranteed. Page 42

43 Main traffic descriptors PCR : (Peak Cell Rate) MCR : (Minimum Cell Rate) CDV : (Cell Delay Variation) The maximum cell rate at which the sender is planning to send cells The minimum number of cells/sec that the customer considers acceptable Measures how uniformely the cells are delivered 4 3 Page 43

44 ATM traffic categories Sub-adaptation layers Constant Bit Rate Variable Bit Rate Available Bit Rate Unspecified Bit Rate Connection ATM traffic categories CBR Class A VBR Class B Class C Virtual connection Bitrate Constant Variable Time Real time No real time Applications AAL 1 AAL 2 AAL 3/4 AAL 5 (SEAL) (Simple Efficient Adaptation Layer) ABR / UBR voice, video voice /compressed video Trafic FR, X25 Class D No connection Trafic LAN The AAL maps the user / control / management PDUs (Protocol Data Units) into the information field of one or more consecutive ATM cells of a virtual connection, and vice versa. To obtain different traffic categories 3 basic parameters are considered. (i) time relation between source and destination In a 64 kbps voice call there is a clear time relation between source and destination. Information transfer between computers has no time relation. Services with a time relation are called real time services. (ii) bit rate Some services have a constant bit rate, others have a variable bit rate. (iii) connection mode Services can be either connectionless or connection-oriented. Class A: voice (circuit emulation), fixed bit rate video Class B: variable bit rate video & audio Class C: connection oriented data transfer & signaling Class D: connectionless data transport, e.g. SMDS (Switched Multimegabit Data Services). Page 44

45 Used for ADSL : CBR ( Constant Bit Rate ) service category PCR Bitrate Guaranteed + CDV ( Cell Delay Variation ) Time 4 5 In CBR all traffic up to the peak cell rate is guaranteed. All traffic that exceeds the PCR is discarded. Page 45

46 Used for ADSL : UBR ( Unspecified Bit Rate ) service category Bitrate PCR MCR Not guaranteed Guaranteed Time 4 6 In UBR all traffic up to the minimum cell rate is guaranteed. All traffic in between the MCR & the PCR is not guaranteed (best effort). The amount of allocated bandwith depends on the network traffic. All traffic that exceeds the PCR is discarded. Page 46

47 ABR ( Available Bit Rate ) service category PCR MCR Bitrate Guaranteed + Rate feedback ( feedback about congestion Cell loss will be low) Time 4 7 Currently not supported in ADSL. Page 47

48 CDV : Cell delay Variation Videoconference ATM network t1 t2 t 1 t 2 Theoretical arrival time Real arrival time CDV??? CDV 4 8 Because of cell delay variation (introduced by cell multiplexing, ) it is possible for ATM cells to arrive early or late; however since ATM is connection oriented the sequence is preserved. You might imagine the confusion during a video conference. It is up to a higher layer protocol, like RTP for audio (Real Time Protocol, RTP runs on top of UDP/IP/AAL2/ATM ), to include a time indication (time in between RTP packets) to correct these delays. Traffic shaping partially compensates for the effects of CDV on the peak cell rate of the ATM connection. Examples of traffic shaping are PCR reduction and reduction of CDV by suitably spacing cells in time. Traffic shaping may be done in the CPE (however it is not done in ADSL so far) to be in compliance with the traffic descriptor and associated parameter values that were negotiated within the network. Page 48

49 Traffic control functions The user asks a connection and specifies some of the traffic descriptors The network checks if it is possible to set up such connection CAC : Connection Admission Control Contrat Descripteur de trafic PCR: MCR: SCR: CDV : 10 Once a connection is set up, the network checks via UPC (Usage Parameter Control) / NPC ( Network Parameter Control) if the agreement has been respected ( policing ) 4 9 Prior to setting up a connection the network checks if there is spare capacity left for the connection. Therefore CAC (Connection Admission Control) checks are performed. E.g. the SANT card of an ASAM performs up- and downstream CAC checks. If the CAC checks fail the operator will receive a `download rejected` message; telling him that there is no spare capacity left. This way QOS is guaranteed. Once you received a connection the policing function is watching you. Even if traffic shaping is performed in the ANT the policing function is still required because the user cannot be trusted. The CDV should always be smaller then the CDVT (Cell Delay Variation Tolerance). In case of CBR it is the GCRA (Generic Cell Rate Algorithm) that calculates the PCR by measuring the inter arrival times of ATM cells. If a cell arrives too fast the cell is discarded. Page 49

50 CAC : Can the network accept a new connection? 6m 3 /h 9m 3 /h 5m 7m 3 /h Pollution Purification plant 20m 3 /h 5 0 If the influx of dirty water in a purification plant exceeds the maximum capacity of the plant then the river will become filthy. Page 50

51 Low level protocols of the Alcatel-1000 ADSL system ATM forum 25,6Mb/s ATM/ADSL ASAM ATM/SDH ANT ATM/PDH ATM network Ethernet 10BaseT ATM/PDH R-ASAM ANT AWS 5 1 Page 51