Chapter 1 ATM reference model
The B-ISDN protocol reference model is defined in ITU-T I.321, as shown below. It consists of three faces: user plane, control surface and management, and in each face, it is divided into physical layer, ATM layer, AAL layer, and high layer.
The three faces in the protocol reference model complete different functions:
User plane: Using a hierarchical structure, providing the transmission of user information flow, and also has certain control functions, such as flow control, error control, etc.
Control plane: Using a hierarchical structure, complete call control and connection control functions, and use signaling to create, monitor, and release;
Management planes: including layer management and surface management. The layer management uses a hierarchical structure to complete management functions related to the resources and parameters of each protocol layer entity, such as metallic signaling. At the same time management also handles the OAM traffic associated with the layers; face management does not hierarchical, which completes the management function related to the entire system and coordinates all planes.
The following chapters do not directly describe the various layers of ATM, but correspond to the OSI seven-layer model.
Chapter 2 Physical Layer
First, the physical layer of ATM, physical interface, ATM switch, ATM physical layer
ATM is an Asynchronous Transfer Mode. This model can be used as a comparison with the synchronous T1 line. Each 125US is generated every 125US in the T1 line, which is controlled by the main clock, and 1 byte data from the same source is controlled in the kth time slot of each frame. T1 is synchronized. The ATM is not strictly requiring the cell to alternately arrive from different sources, and each column has no special mode, the cell can come from any different source, and it is not required to come from a computer. The cell stream is continuous, and the data cell can have intervals, which are filled with special idle cells (Idle Cell).
ATM does not standardize the transmissive cell format. In fact, it indicates that only single cells are available and indicate that the cell can be loaded onto T1, T3, SONET, or FDDI (fiber LAN) line. For these examples, there is a standard specified that the cell is encapsulated into the frames provided by these systems.
In the initial ATM standard, the main rate is 155.52Mb / s, and there is another 4-fold (622.08MB / s). Select this rate to be compatible with SONET, SONET is a divided standard standard for fiber optic lines in the telephone system. ATMs based on T3 (44.736MB / s) and FDDI (100MB / s) have also appeared.
The transmission medium of the ATM is often fiber, but the coaxial cable or the 5-class twisted pair of coaxial cables within 100m can be. Fiber can reach thousands of meters long. Each link is between the computer and an ATM switch or between the two ATM switches. In other words, the ATM link is point-to-point (not the LAN, which has many senders and recipients on a cable). By allowing the cell to enter the switch from a line and output from a plurality of lines, the broadcast effect can be obtained. Each point-to-point link is one-way. Two links are required for all duplex operations, and traffic in each direction takes up one.
The physical layer of the ATM includes two sub-layers, i.e., a physical media sublayer (PM), and a transmission condensation (Tc) sub-layer. Among them, the physical media sublayer provides bit transport capability, compared to specific time and line coding, etc., and defines its corresponding characteristics for the physical media used (such as fiber, coaxial cable, twisted pair, etc.); transmission convergence The main function of the sub-layer is to achieve a conversion between a bitstream and the cell stream.
For the output, the ATM layer provides a cell sequence, the PDM sublayer performs the necessary encoding, and it is transmitted in a bit stream. For input, the PDM sublayer obtains the input bits from the network, and a bit stream is submitted to the TC sublayer. The boundaries of the cells are not marked, and the TC sublayer is responsible for finding where the cell is started and end. But this is not only difficult, but it is not in theory, so the TC layer removes this feature. Because the TC layer manages the division, it belongs to the data link function, so we discuss it in Chapter 3. Second, physical interface
The ITU-T and ATM forums divide the physical interface into three categories, namely SDH, based on cell and PDH. The following is a different perspective:
Traditional digital signaling.
DS064Kbit / s DS1 (T1) 1.544Mbit / S DS2 (T2) 6.312Mbit / S (4 DS1, 96 DS0) DS3 (T3) 44.736Mbit / S (28 DS1, 672 DS0) DS4274.176Mbit / S (4032 DS0) Synchronous Transmission Lettering (STS)
STS-1 (OC-1) 51.84 Mbit / s STS-3 (OC-3) 155.52Mbit / S (3STS-1) STS-12 (OC-12) 622.08Mbit / S (12STS-1) STS-24 ( OC-24) 1244.16Mbit / S (24STS-1) STS-48 (OC-48) 2488.32Mbit / S (48STS-1)
ANSI standard
STS-151.84Mbit / SSTS-3C155.52Mbit / SSTS-12C 622.08Mbit / SDS344.736Mbit / S CCITT standard
DS11.544Mbit / SE12.048Mbit / SDS26.312Mbit / SE28.448Mbit / SE334.368Mbit / SDS344.736Mbit / SSTM-1155.52Mbit / S (same as STS-3) STM-4 622.08Mbit / S (with STS-12 is the same)
Third, ATM switch 1, ATM basic queuing principle
ATM exchange has two fundamental points: cell exchange and statistical multiplexing between each virtual connection. The cell exchange is about to exchange the ATM cell through various forms of exchange media, from one VP / VC to another VP / VC. Statistical reuse performance exchange resources such as the exchange of cells of the virtual linkage of the virtual connections, in order to solve the competition of these resources, to solve the competition of these resources, and must separate the cells, borrow circuits The idea of exchange, can be considered that statistical reuse is reflected in the exchange, and is achieved by queuing mechanism.
The queuing mechanism is an extremely important content in the ATM exchange. The overflow of the queue caused the band loss, the cell line queue is the main reason for the exchange of delay and delay jitter, so the queuing mechanism has a decisive impact on the performance of the ATM switch. There are three basic queuing mechanisms: input queue, output queuing and central queuing. Each of these three ways, such as the input queue has a letterhead blocking, the load is less than 60%; the output queuing memory utilization rate is low, the average captain is long, and the central queuing memory rate is high, and the memory management is complex. At the same time, there are various ways, the input queue is low on the memory rate, the central queue is high, and the output queue is between the two, so it is not directly utilized in practical applications, but integrated Adopted some improvement measures. Improved methods mainly:
Reduce the header blocking of the input queue.
Using the input and output of the back pressure control queuing method.
Way of queuing mechanism with loopback.
Shared output queuing method.
A plurality of output sub queues are set on a output line, which are logically operated as a single output queue.
2, ATM switch
In order to achieve a large-capacity exchange, in order to increase the scalability of the ATM switch, the basic exchange unit of the small capacity is often constructed, and then these switching units are configured to become a ATM switching mechanism (FABRIC), and for the ATM switching mechanism The main problem of research is to transmit media structure and circuitry between the exchange units, and how to reduce competition and reduce blocking. The ATM exchange mechanism classification method is different. There is a category of time division exchange and space separation, where time division exchange includes shared bus, shared ring, and shared memory structure, space division exchange includes full interconnection network and multi-level interconnect.
3, ATM switch
The general model of the ATM cell switch is shown below. It has some input lines and some output lines, typically equal to the number (because the lines are bidirectional). A cell (if any) is obtained from each input line every cycle. By the internal exchange fabric, it is transferred on the appropriate output line. From this perspective, the ATM switch is synchronized.
A universal ATM switch
The switch can be a pipeline, that is, the entered cell may appear on the output line after a few cycles. The cell is actually asynchronous to reach the input line, so there is a primary clock to indicate the beginning of the cycle. Any cell that is completely reached at the time of time can be exchanged within the cycle. The unfinely arrived cell must wait until the next cycle.
The cell usually arrives at the ATM rate, typically at around 150MB / s, that is, about 360,000 cell / s, which means that the switch has a cycle of approximately 2.7 um. A commercial switch may have 16 to 1,024 input lines, that is, it must receive and exchange 16 ~ 1,024 cells per 2.7um. At the rate of 622Mb / s, a group of cells entered the switching structure per 700ns. Since the cell is fixed and smaller (53 bytes), this may make such a switch. If a longer variable length group is used, the high-speed exchange will be more complicated, which is the reason why ATM is short and fixed.
4, classification of ATM switches
Various ATM swap devices are different due to the different applications, and the completion of the completion is slightly different, and the main differences include interface types, exchange capacity, and processing signaling.
In the public network, there is access switch, a node switch, and a cross-connect device. The location of the access switch is equivalent to the user switch in the telephone network, which is located on the edge of the ATM network, connects various service terminals into the ATM network. The status of the node switch is similar to the service switch in the existing telephone network. It completed the VP / VC exchange, requiring the exchange capacity to be large, but the interface type is simple than the access switch, only the standard ATM interface, mainly the NNI interface, also There are UNI interfaces or B-ICI interfaces, signaling aspects, as needed to process ATM signaling. The cross-connect device is similar to the cross-connect equipment in the existing telephone network, which completed VP exchange in the backbone, and does not need to perform signaling processing, thereby achieving extremely high speed exchange.
In the ATM private network, there is a private network switch, an ATM LAN switch. The private network switch acts equivalent to a node switch in the public network, a UNI and an NNI interface with a private network, complete the signaling processing of P-UNI and P-NNI, with strong management and maintenance. The ATM LAN switches complete the access of the LAN, and the ATM LAN switch should have a local area network interface and an ATM P-UNI interface, and the various layers of protocols of the DPRK network, and ATM signaling.
Chapter 3 Data Link Layer
First, the data link layer in the ATM, the cell transmission is three, the cell receives one, the data link layer in the ATM
The ATM physical layer generally includes an OSI physical layer and a data link layer, including a physical medium like an OSI physical layer determines a sublayer and a transmission assembly (TC) sub-layer as the data link function. For ATM, there is no special physical layer feature. Instead, the ATM cell is shipped by SONET, FDDI, and other transmission systems. Therefore, we will focus on the data link function of the TC sublayer. When an application generates a message to be sent, this message is to enter the transmission line, transfer to the ATM protocol stack, plus the head and tail, and put the segment into the ATM cell. Finally, these cells arrive at the TC sub-layer for transmission. Let us look at the door, what happened on the road.
Second, the cell transmission
The first step is to perform the header checksum. Each cell has a 5-byte head, including 4 bytes of virtual circuits and control information, and 1 bytes of checksum. The checksum only includes the first 4 head bytes without occupying the payload bytes. It is made of 32 headers in addition to a polynomial x ^ 8 x ^ 2 x 1, the remainder obtained. Check and add constant 01010101.
Make a decision to check the header, in order to reduce the possibility of incorrect delivery of the cell due to the header error, but also to avoid the valid load field of the payload field to avoid the verification of the payload field. If you need to verify the payload field, you will get this function on a higher layer. Since the checksum fields are located only on the head, the 8-bit checksum fields are called header errors control HEC (Header Error Control).
Once HEC is generated, insert the cell head, this cell is ready to send it. Transmission means are divided into two groups: asynchronous and synchronous. When using asynchronous methods, as long as you are ready to send it, you can send, there is no time limit.
Using synchronous mode, the cell must be sent in advance according to the predetermined time. If there is no data cells available when needed, the TC sublayer must invent one, this cell is called an idle cell.
Another type of not data is the operation and maintenance of OAM (Operation and Maitenance). The ATM mechanism also uses OAM cells to exchange control and other necessary information to ensure the operation of the system. The ATM output rate matches the rate of the transmission system is an important task of the TC sublayer.
At the receiver, the idle cell is processed in the TC sub-layer, but the OAM cell is given to the ATM layer.
Another important task of the TC sub-layer is: If there is, the system is generated in accordance with the system that is transmitted, generates frame information. For example, an ATM camera only produces a series of cells on the line, but it may also generate SONET frames with ATM cells, embed in the SONET payload. In the latter case, the TC sublayer will generate SONET or frames, and package the ATM cell, which is not completely unnecessary step because the SONET payload cannot support the integer multiple of 53 byte cells.
Although the telephone company explicitly uses SONET as an ATM transmission system, it can also be defined as payload fields corresponding to other systems, and this new frame is working. In particular, it is also possible to map T1, T3 or FDDI frames.
Third, the cell receive
At the output, the operation of the TC sublayer is a series of cells, increasing a HEC on each cell, converting this result into a bit stream, and multi-bit stream matches the bit stream to perform a physical transmission system s speed. At the input, the TC layer accurately reverse transform. It comes to the bit stream that arrives, sets the cell boundary, determines the cell header (discarding the cell with the unlawful header), and processes the OAM cell, and upload the data cell to the ATM layer.
The most difficult part is to set the cell boundary in the arrival bit stream. In some cases, the physical layer for transmitting provides help. However, sometimes the physical layer does not provide help. What should I do if I do this? Tips are HEC. As the bitstream reaches the TC sub-layer, a 40-bit shift register is retained, and the bit stream enters from the left, and the right is coming. The TC sub-layer then reviews these 40 points to see if there may be a legal cell head. If there is, the rightmost 8 bits will be legitimate HEC, and the leftmost 32 bid is not. If this is there, there is no legal cell in which the buffer does not exist. In this case, all the bits in the buffer move one bit to the right, so that the backend is empty, so a new input bit Add it to the leftmost end. This process is continuously repeated until a legal HEC is positioned. At this time, the cell boundary is clarified because the shift register includes a valid head.
Chapter IV Network Layer
First, the network layer in ATM
Second, the cell format
Third, the connection is established
Fourth, routing selection and exchange
V. Service type
Six, service quality
Seven, traffic integrity and control
Eight, congestion control
First, the network layer in ATM
The ATM layer handles the cell from the source to the destination, in the ATM switch, does include a routing algorithm and protocol, which also handles global addressing issues. Therefore, the ATM layer functions as the same function as the network layer. The ATM layer does not guarantee 100% reliability, but the protocol of a network layer does not need this.
Because the ATM layer has the function of the network layer without having the functionality of the data link layer, and the ATM layer is similar to the existing network layer, so we are still discussing the ATM layer protocol in this chapter.
The only problem is that the ATM layer does not have the characteristics of the data link layer protocol: a single-station protocol between the machines used for the wires, just like the protocols 1 to the protocol 6 in Chapter 3. The ATM layer has a function of a network layer protocol: end-to-end virtual circuit connection, exchange, routing.
For connection-oriented protocols, the ATM layer is unusual because it does not provide any confirmation. However, the ATM layer still provides a powerful guarantee: the cells sent along a virtual circuit will never lose order. If the block occurs, the ATM subnet is allowed to discard the cell, but in any case, it cannot reordbound the cells transmitted in a separate virtual circuit. However, the credit transmitted in different virtual circuits does not provide a sequential guarantee.
Second, the cell format
In the ATM layer, there are two interfaces that are very important, namely, user-network interface UNI (user-network interface), and network-network interface NNI (Network-network interface). The former defines the boundary between the host and the ATM network (in many cases, between the client and the carrier), the latter is applied to the two ATM switches (routers in the ATM). The ATM cell heads of the two formats are shown below. The cell transmission is the leftmost byte priority, and inside the one byte is the leftmost bit priority.
Figure (a) ATM head in UNI; (b) ATM head in NNI
Third, the connection is established
Technically, the connection establishment is not part of the ATM layer, but is processed by a highly complex q.2931 (Stiller, 1995) used by the control platform. However, logically processing the location of the network layer connection is the network layer, and similar network layer protocols are settled here, so we discuss it here.
Messages for connecting to establish and release
Message is sent by the host when sending the meaning of the meanings when sending the meaning setup. Please create a virtual circuit to enter the call call processing, I saw the entry call will try your call request Connect I accept the entry call to accept your call request Connect Ack Thank you, thank you, Call Release Please terminate the other end. Release completion of Release Complete's confirmation to Release confirmation ATM network allows you to create a multi-game channel. A multipoint broadcast channel has a sender and more than one recipient. They are established by the following method: establish a connection between the source and destination between the source and destinations, and then send the Add Party message to the second destination to the front call returned to the virtual circuit, next You can send the rest of add Party to increase the number of destinations.
ATM has three address formats. The first byte indicates which of the three address formats. The first 20-byte length is based on the OSI address format. The 2nd and 3rd bytes indicate the country, the 4th byte gives the format based on the address part. Others include 3-byte specified permissions, 2 bytes indicating domain (Domain), 1 byte indicated area, and 6 words The address of the section, and some other information items. In the second address format, the 2nd and 3rd bytes specify an international organization, not the country; the rest of the address and the format are the same as the first. The other is the old use of 15-bit decimal number of ISDN phone numbers (CCITT E.164) as the format of the address.
Fourth, routing selection and exchange
When a virtual circuit is established, the SETUP message walks along the network from the source end. The routing algorithm determines the path to the message to go, and thus determines the path to the virtual circuit. Any specific routing algorithm is specified in the ATM standard, so people can choose one from our routing algorithm discussed in the previous few sections of this chapter, or selects different different algorithms.
Most of the workload of the switch is the choice of how to get the output line from how to get the output line from the virtual circuit information in a cell. In addition to the last station in each direction, the routes are performed on the VPI field instead of the VCI field; in the last station, the cell is transmitted between the switch and the host. Affihood is used between the two switches.
In the local area network, things are much simpler, and a simple virtual path can be used for all virtual circuits.
V. Service type
Constant bit rate CBR (constant bit rat) is mainly used to mimic copper or optical fibers. There is no error check, there is no traffic control, and there is no rest. This category makes a relatively smooth transition in the current telephone system and future B-ISDN systems, as the voice-level PCM channel, the T1 circuit, and the remaining telephone systems use a constant rate synchronous data transmission.
The variable bit rate VBR (variable bit rate) is divided into two sub-groups, which are set up for real-time transmission and non-real-time transmission. RT-VBR is primarily used to describe services with variable data streams and require strict real-time services such as interactive compressed video (eg, television conference). NRT-VBR is used primarily to be a communication occasion, in which case a certain number of delays and changes thereof can be endured by the application, such as email.
The available bit rate ABR (Available Bit Rate) is designed for burst information transmission that has been generally known for bandwidth. ABR is the only network that provides speed feedback to the sender. When the congestion occurs during the network, the sender will require the sender to reduce the transmission rate. Assume that the sender complies with these requests, the cell loss using ABR communications will be low. The ABR running is a bit like a moving passenger waiting for the opportunity: Bandwidth is available). Unspecified bit rate UBR (unspecified bit rate) does not do any promise, no feedback to congestion, this type is well suited to send IP datagrams. If congestion occurs, the UBR cell will also be discarded, but it does not send feedback to the sender, nor does it give the sender to slow down the expectations.
Various ATM service types
Service features CBR RT-VBR NRT-VBR ABR UBR bandwidth guarantees are optional inappropriate to real-time communication is not applicable to burst communication is not about congestion feedback, not
Six, service quality
Service quality is an important topic in the ATM network, which is because the ATM network is used as real-time transmission, such as audio and video. When a virtual circuit is established, the transport layer (typically a process, "customer") and ATM network layers in the host (for example, a network operator, i.e., "carrying provider") must comply with a definition service. agreement.
The first part of the agreement is a traffic descriptor. It describes the load to be provided. The second part of the agreement specifies the quality of service that the customer's request and the communication provider agreed. Whether it is load or service, it is to be described in a measurable amount, so that the agreement can be objectively determined.
In order to make the specific traffic agreement possible, the ATM standard defines a series of service QoS (Quality of Service), customer, and communication providers to negotiate the values of these parameters. For each service quality parameter, the value in the worst case is specified, requiring the communication provider to reach or exceed this value. In some cases, the parameter is a minimum, and in other case it is a maximum. Also here, the quality of quality is specified separately in each direction. Some of these more important columns are in the table below, but they are not applicable to all service types.
Some service quality parameters
Parameter Necklement Meaning Peak Cell Rate PCR Cell Send Maximum Rate Connection Cell Rate SCR Long Time Average Cell Transmission Rate Minimum Cell Rate MCR Minimum Acceptable Cell Transmission Rate Cell Delay Variable Extreme CDVT Maximum Acceptable cell jitter cell loss ratio CLR cell loss or submit too late proportional cell transmissions delay the time (intermediate value and maximum value) CMM delay change CDV cell submission time Change amplitude cell error ratio CER Submit the ratio of error-free cell comparison Severe error message ratio SECBR error cell proportional cell error destination ratio CMR cell submit to the error destination
Seven, traffic integrity and control
The mechanism to use and enhance the quality of service parameters is based on (partially) a particular algorithm, i.e., universal cell rate algorithm GCRA (Generic Cell Rate Algorithm). Its working principle is to check each cell to see if it complies with the parameters of the virtual circuit.
GCRA has two parameters that specify the maximum allowable arrival rate (PCR) and the arrival time variation (CDVT). PCR's reciprocal, T = 1 / PCR is the minimum cell arrival interval value.
The GCRA algorithm is called a virtual schedule algorithm, but it is equivalent to a drain bar algorithm from another perspective. A fluid that is imagined to be imagined into a T unit that pours a drain bar. This bucket leaks the liquid at a speed of 1 unit / US, so it is empty after TUS. If the cell is just arrived at 1 cell / TUS, then each arrival cell will find that the bucket just is empty, the cell will reload the tank in the tank. Therefore, when a cell arrives, the liquid level is rising to T, and then linearly decreases to zero. When a cell arrives in advance, the bucket should overflow. For a given T, if we set the L, the capacity of the bucket will be difficult to exceed T, so all the cells must be sequentially sent in a very specific interval. However, if we now increase the value of L, it is much more than t, the bucket will accommodate a lot of cells because T L >> T. This means that the sender can send some sudden data in a peak rate, while they can still be properly received.
GCRA is normal for normal cases by a given parameter T and L. T is just the countdown of PCR; l is CDVT. GCRA is also used to ensure that the average cell transmission rate does not exceed the SCR over a longer period of time.
In addition to providing a rule to see which cell is symptomatic, which is not agreeing, GCRA is also used to communicate shaping to eliminate certain burst transmissions. The smaller the CDVT means the better smooth effect, but it also increases the chance of the cell because it is not agreeing. The GCRA drain bucket is combined with a token barrel in some implementation to provide further smooth.
Eight, congestion control
The ATM network must handle long-term congestion caused by greater than the traffic capacity greater than the system processing capacity, but also to process short-term congestion caused by sudden transmission in communication. Results There are several different strategies. The most important of them can be divided into 3 categories:
1, license control
Many ATM networks have real-time communication sources that generate data at a fixed rate. Tell this type of communication source slowing the transmission rate is unreasonable (imagine a new digital phone with a red light. When the notification congestion occurs, the red light will be bright, the speaker will be required to be reduced by 25% ).
Therefore, the ATM network prevents congestion from being placed in the first position. However, for CBR, VBR, UBR channel traffic, there is no dynamic congestion control at all, so it will prevent congestion here to recover much better than congestion. One of the main tools to prevent congestion is the license control. When a host requires a new virtual circuit, it must describe the communication and services that wish to be provided, and the network is checked to see if it is possible, and the connection is processed without a harmful impact on the existing connection. You may need to check a number of possible lines, thus finding which one will work.
2, resource reservation
The same information is closely related to the prior predetermined resource, which is usually done when the call is established. Because the traffic descriptor gives the cell transmission peak rate, the network is possible to reserve sufficient bandwidth along the path to process the peak rate.
3, speed-based congestion control
In CBR and VBR communication, because the information source is solid, the real-time and semi-real characteristics, even in the case of congestion, it is generally impossible to slow the sender to slow down the transmission rate. No one will worry in the VBR service. If there are too many cells, you will drop more.
In ABR communication, the network is to inform one or more senders and request them to temporarily slow down the transmission rate until network recovery, which is also possible.
How to detect, notify and control the congestion in ABR communication is a hot topic in the development of the ATM standard, and the problem is concentrated in two aspects: First, credit-based solutions, one is based on speed-based solutions.
Switch vendors oppose credit-based solutions. They don't want all calculations to remember these credits, and do not want to provide a lot of buffers in advance and think that the total amount of overhead required is too large. Therefore, a speed-based congestion control system is employed. Its basic model is that each transmitting end passes a special resource management RM (Resource Management) cell after K cell data. The transmission path of this cell is the same as the K cell, but it is specially processed by the switch. When the RM cell reaches the receiving end, it is detected, modified, and then transmits it back to the transmitting end. In addition, two other congestion control devices are also provided. The first is that the overload switch can generate RM cells since its own, and send them back to the transmitting end. The second is that the overload switch can set the value of the PTI bit to the cell data transmitted from the transmitting end to the receiving end. Of course, these two methods are not completely reliable. Chapter 5 Transport Layer
First, the structure of the transport layer in ATM, the structure of the ATM adaptation layer, the AAL1 four, AAL2 5, AAL3 / 4 three, AAL5 seven, AAL protocols Comparison Eight, SSCOP - Specific Services for Connection Agreement 1, ATM Transport layer
It is difficult to say if ATM has a transport layer. On the one hand, the ATM layer has a function of a network layer, and there is a layer (AAL), from the hierarchical angle, AAL is a transport layer. Some experts agree to this view. One of the protocols used here (AAL5) is functionally similar to UDP, and UDP is undoubtedly a transport layer protocol. On the other hand, there is no AAL protocol to provide a reliable end-to-end connection as TCP (although these protocols only need to do small changes). In addition, in most applications, another transport layer is also used on AAL. It is no longer necessary, and the AAL layer and its agreement are discussed in this chapter, regardless of whether it is a real transport layer.
The AAL layer of the ATM network is essentially different from TCP. The main reason is that the designer is more interested in transmitting audio and video data streams, and quickly transmits more important than accurately. The ATM layer continuously outputs 53 bytes of cells. There is no error control in the cell, there is no traffic control and other types of controls. So, it does not meet the requirements of most applications. In order to make up this lack, in recommendation I.363, the ITU defines one end-to-end layer above the ATM layer. This layer is called ATM adaptation layer AAL (ATM Adaptation Layer), which has experienced a paragraph history: full of errors, repeated revisions, and unfinished work.
AAL target is to provide a useful service to the application and divide them with the transmitting end (party) into a cell, separating the cell reorganized the cell recommptive to the data. It organizes service space in three coordinate axes:
1, real-time service and non-real-time service.
2, constant bit rate service and change bit rate service.
3. Connected service and non-connected services.
In principle, eight different services can be defined using two values on three coordinate axes and each coordinate axis, as shown below. ITU feels that only four of them have value, and named A, B, C, D, respectively. Several other are not supported. Starting from ATM 4.0, this picture has some time, so it is proposed here to help readers understand why the AAL protocol is designed as current. At present, the main difference is between transport classes (ABR, CBR, NRT-VBR, RT-VBR, and UBR), not between the service classes supported by these AAL.
Basic service class supported by AAL (now outdated)
To deal with these four types of services, ITU defines four protocols and later discovering that the technical requirements for class C and class D are very similar to the AAL3 and AAL4 in AAL3 / 4. The computer industry was eager to succinct, and later found that they were not satisfactory. Later, another protocol -AAL5 is defined to solve this problem.
Second, the structure of the ATM adaptation layer
The upper portion of the ATM adaptation layer is called a Cenvergence Sublayer. Its role is to provide an interface to the application. It consists of two sub-parts: one is a common part (relative to a given AAL protocol) that is common to all applications, the other is a sub-part associated with the application. Each of them is related to the protocol, but can include packet fragmentation and error detection. At the transmitting end, the gravist is responsible for receiving a packet from the application's bitstream (data) or random length, and divides them into a unit of 44 to 48 bytes for transmission. The exact size depends on the protocol used, because some protocols should take a portion of the 48-byte ATM load as their head. At the receiving end, the sublayer recombines the cell to the original message. Typically, the message boundary (if present) is to be retained.
The part of the AAL is called the segmentation and recombinant SAR (Segmentation and Reassembly) sub-layers. It puts the data unit that the contributor is given to its data unit and the end and the tail constitutes a product payload. Next, these loads are handed over to the ATM layer for transmission. At the receiving end, the SAR layer will restructuring the cell. The SAR layers are basically involved in cells, while the contributor is derived from the packet.
SAR layers have additional features for some (but not all) service classes. In particular, it can sometimes perform error detection and multiplexing. SAR layers are existing for all service classes, but the strength is strong, depending on its specific protocol.
Third, AAL1
AAL1 is a protocol for Class A transmission. Class A transmission refers to real-time, constant bit rates, facing-oriented transmission, such as non-compressed audio and video data. Enter a bit stream, there is no message boundary. For this transmission, there is no use of an error detection protocol that is stopped-, because the delay introduced by the timeout and the retransmission mechanism is unacceptable. However, the application will notify the application when the cell is lost, and it takes a measure (if possible) to make up.
AAL1 uses a gravity and SAR layers. The gravity is detected to detect the loss and mistroducing cells, and the data rate of the gently entered is set to transmit the cell at a constant speed. Finally, the gravidine is decomposed into a 46-byte or 47-byte unit, and then handed over to the SAR sub-processing. At the other end (Receiver), it takes out these data units, and recombines into the original input. AAL1's contributor layer does not have its own protocol header information.
On the contrary, AAL1's SAR layers have their own agreement. Its cell format is shown below. Both formats are starting at 1 byte of head: which includes 3 bytes of cell serial number SN (for detecting whether it is lost or false into the cell); the field is 3-bit sequence protection field SNP (ie Calibration), you can correct individual errors in the positive cell serial number field and detect two errors.
When the packet boundary must be retained, P cells are used. Pointer field is used to give the offset of the next packet start position.
Fourth, AAL2
AAL1 is designed for simple,-facing, real-time data flow, in addition to having a detection mechanism for loss and mistroduction, it has no error detection. For simple uncompressed audio or video data, or occasionally there is no problem with any other data streams, AAL1 is already enough.
For compressed audio or video data, the data transfer rate will change with time. For example, many compressed schemes are periodically transmitted, and then transmits only the difference between adjacent order frames, and then transmits a complete frame. When the lens is still moving and there is no such thing, the difference frame is small. Second, the message division must be retained so that it can distinguish the next full frame start position, even when there is a loss of cell or bad data. For these reasons, there is a need for a better agreement. AAL2 is designed for this purpose. Like in AAL1, AAL2's contributor layer does not have its own agreement and SAR sub-layers have its own protocol. The format of SAR cell is as follows:
AAL2's cell format
The serial number SN (SEQUENCE NUMBER) field is used to record the number of cells to detect cell loss or misuse. The Information Type IT (INFORMATION TYPE) field is used to indicate that the cell is the beginning of the message, the middle or end. Length Indicator Li (Length Indicator) field indicates how much payload is byte (the payload may be less than 45 bytes). Finally, the CRC field is the verification of the entire cell, and an error can be detected.
There is no size of each field in the standard. It is said that in the final government of the standardization process, members of the Committee think that AAL2 has many problems, so that they can't be put into use, but it is too late, there is no way to organize standardized processes. Finally, members have fallen all the fields of field size setting to enable formal standards to be promulgated on time, but no one can actually use it.
Five, AAL3 / 4
At the beginning, the ITU has developed different protocols for services class C and D (service class C and D are sensitive to data loss or error, but do not have real-time connection and non-connection data transmission service classes). Later ITU found no need to specify two sets of protocols, so they will be combined into one, form a separate protocol, namely AAL3 / 4.
AAL3 / 4 can operate in two modes, namely flow, and packets. The message boundary information is not retained in the current mode. The flow mode will be discussed below. Reliable transmission and unreliable (ie, non-guaranteed reliability) may appear in each mode.
AAL3 / 4 has no performance in other protocols - support multiplexing. This feature of AAL3 / 4 allows multiple sessions from one host (such as remote login) to transmit along the same virtual circuit and separated at the destination. All sessions using a virtual circuit get the same quality service because it is determined by the nature of the virtual circuit itself.
And AAL1, AAL2 is different, AAL3 / 4 has a contributor-based protocol and a SAR sub-protocol. The packet reaches the maximum number of 65535 bytes from the application to the congoric layer. First, fill it to 4 integer waith bytes. Then add head and tail information. Reconstructuring the insulation in the contextor layer, and after the head and tail information are added, the packet is transmitted to the SAR layer, and the packet is divided into a maximum 44-byte data sheet by the SAR sublayer.
AAL3 / 4 has two-layer protocol overhead: Each message needs to increase 8 bytes, and each cell adds 4 bytes. In summary, it is a mechanism for a large overhead, especially for short messages.
Sixth, AAL5
From AAL1 to AAL3 / 4 protocols, it is mainly designed by the telecommunications industry and is standardized by ITU, which does not consider much to consider the requirements of the computer industry. Due to the complexity and inefficiency of the two protocol layers, plus the checksum fields (only 10), so that some researchers have born a new adaptive layer agreement. The protocol is called a simple and effective adaptation layer Seal (Simple Efficient Adaptation Layer). After argumentation, the ATM Forum has accepted SEAL and called AAL5 for it.
AAL5 provides several services to its applications. One option is a reliable service (ie, the flow control mechanism is used to ensure the transmission, to prevent overload); another option is unreliable service (ie, does not provide data transmission guarantee measures), pass options to make the check error or lost Or transfer to the application (but being identified as a bad trunk). AAL5 supports point-to-point mode and multipoint broadcast mode transmission, but multi-playback modes do not provide a guarantee measures for data transmission. Like AAL3 / 4, AAL5 supports packet mode and flow mode. In message mode, the application can pass the length from 1 byte to 65535 bytes to the AAL layer. When reaching the contextic sub-layer, populate the packet to the payload field and add the tail information, select the fill data (0 bytes ~ 47 bytes) to make the entire message (including the data and tail information) of the entire message (including the data and tail information) The integer multiple of the word one section. AAL5 does not have a grade head, only one 8-byte tail.
The user to the user uu (user to user) is not used in the AAL layer itself, but for its own purpose for a higher layer (possibly a specific service sub-part of the context), for example, sort or multiplexing. The length of the length indicates how much the real payload is in bytes, which does not include the number of bytes of the filled. 0 value is used to terminate unloaded packets. The CRC field is based on the standard 32-bit checksum of the entire packet, including filling data and tail information (CRC field set to 0). A 8-bit field of the tail will be used in future use.
The packet is handed over to the SAR layer and then sent out. Do not add any heads and tail information in the SAR layer, but divide the packet into a 48-byte unit, and send each unit to the ATM layer for transmission. It also notifies the ATM layer to set the PTI field of the last cell to 1 to retain the message division. (A problem occurs at this time: This is an incorrect protocol layer mixture because the AAL layer should not use the head information of the ATM layer.)
AAL5 is more efficient than the main advantages of AAL3 / 4. Although AAL3 / 4 has only an increase of 4 bytes of header information for each message, it also adds 4-byte header information for each cell, so that the capacity of the payload is reduced to 44 bytes, for long Packets, invalid data accounted for 8%. Each message of AAL5 has a slightly large tail (8 bytes), but there is no additional overhead of each cell. There is no sequence number in the cell, you can make up for a long check, so that the lost, misuse or erroneous cell can be detected without using sequence numbers.
In the Internet, the general method of interface with the ATM network is to transmit IP packets using AAL5's payload fields. Various issues related to this method are discussed in RFC 1483 and RFC 1577.
Comparison of seven, AAL protocols
Various AAL protocols do not seem to be similar, and considering that it is very difficult, and there is also doubts about the discolorer layer and SAR sub-layers, especially because AAL5's SAR layers do not have any own characteristics. With a slightly enhanced ATM layer header information, it is sufficient to provide image sorting, multiplexing, and data division.
The overall impression of AAL is a lot of variants, and there are many subtle differences between variants, and have not been completed. The original four services A, B, C, D are actually abolished. AAL1 may not necessarily exist; AAL2 is incomplete; AAL3 and AAL4 never have a day; AAL3 / 4 is less efficient and there is too little verification and field bit.
Everything in the future relies on AAL5, but so far, AAL5 has many other improvements. AAL5 packets should have a sequence number and a logo for distinguishing between data or control packets, so that it can be a reliable transport protocol. The above functions can be implemented with unused spaces in the tail.
Eight, SSCOP - Connection Agreement for Specific Services
Although there are so many different AAL protocols, there is no transmission connection that supports simple and reliable point-to-point. Applications that require this service can use another protocol-to-specific service-oriented connection protocol SSCOP (Service Specific Connection Oriented Protocol). However, SSCOP is only used to control and cannot be used for data transmission.
SSCOP users send packets, each packet being given a 24-bit sequence number. The packet is up to 64KB and cannot be separated. They must be transmitted in sequence. Unlike some reliable transmission protocol, it always selectively retransmit all packets when it is selectively retransmitted while losing the packet rather than return to the serial number n.
SSCOP is fundamentally a dynamic sliding window protocol. For each connection, the recipient reserves the window that is ready to receive the packet number, and indicates whether the packet has existing bitmap (Bitmap). This window can change the size during the protocol operation.
The unusual of SSCOP is a confirmation method: it does not have a mechanism. Instead, the sender periodically queries the recipient, requiring it to send a bitmap that indicates the state of the window. According to this, the sender discards the message received by the other party and updates its window.
Chapter 6 ATM Network Structure and Interface
First, the ATM Network Structure 2, ATM Main Interface 1, ATM Network Structure
The ATM network can be divided into three parts: public ATM network, special ATM network and ATM access network.
The public ATM network is an ATM network operated and managed by the telecommunications management department. It connects each dedicated ATM and ATM terminals through a public user network interface. As the backbone network, the public ATM network should ensure interoperability with existing networks, which should be able to support a variety of existing services including ordinary phones, and there must be a set of maintenance, management and billing. There is currently no commercial ATM network, and the agreement on public ATM network is constantly being improved.
The dedicated ATM network refers to an ATM network within a unit or department. Since its network scale is small than the public network, it does not require management procedures such as bills, the dedicated ATM network is first entering the practical ATM network, new ATM Equipment and technology are often used in the ATM private network. The current private network is mainly used for local area network interconnect or directly constitutes ATM LAN to provide high quality multimedia business and high-speed data transfer on the local area.
Access ATM network mainly refers to using ATM technology in various access networks, transmitting ATM cells, such as ATM-based passive fiber network (APON), mixed fiber coaxial (HFC), asymmetric digital loop (ADSL) And use ATM wireless access technology, etc.
Second, ATM main interface 1, UNI (user-network interface)
UNI is a user network interface in the ATM network, which is an interface between the user equipment and the network, directly facing the user. The UNI interface defines the interface criteria of the physical transmission line, that is, how the user can connect with the ATM network through what physical line and the interface, also defined the ATM layer standard, UNI signaling, OAM function, and management functions. According to the location where the UNI interface is located, it can be divided into the Uni (PUNI) of the public network, and the definition of the two UNI interfaces is basically the same, but the PUNI is too much because of the incomprehensible interface. Consider strict consistency, so PUNI's interface is more, more flexible, and more developed.
2, NNI (Network to NetWork / Network Node Interface)
NNI can be understood as an interface between network node interfaces or networks / networks, which is generally interface between two switches. Like UNI, the NNI interface also defines a specification of the physical layer, ATM layer, etc., and signaling, etc. Function, but since the NNI interface is related to the routing problem connected to the network, the route selection method is specifically described. Similarly, the NNI interface is also divided into NNI (PNNi) in public NNI and private network, the difference between the public NNI and PNNi is still quite large, such as the Broadband ISDN of the ISDN of the No. 3,7 signaling system. The user part B-ISUP, and PNNI is completely based on UNI interfaces, still uses the Uni signaling structure. 3, B-ICI (BiSDN Inter-Carrier Interface)
B-ICI is defined as an interface between two public ATM networks, providing a connection to the UNI interface belonging to two operators, which is based on NNI interface, which is characterized by supporting multiple services between different networks, including Based on cell-based PVC mode business, PVC mode frame relay services, circuit simulation services, SMDS, and SVC services.
4, DXI (DATA Exchange Interface)
DXI defines between digital terminal devices DTE and digital connection devices DCE, DTE is connected to DCE through DXI, and then access the ATM network via the ATM UNI interface, DCE completes the adaptation process of data terminals that do not meet the ATM standard, Equipped with the terminal adapter.
5, FUNI (Frame Based Uni Interface)
Funi's meaning is similar to DXI. FUNI completely moves the ATM adaptation function into the switch between the switch, the terminal, and the ATM switch transmits the FUNI frame, so the FUNI has higher in the access line compared to the cell-based DXI interface. effectiveness.
Chapter VII Support IP in ATM Network
I. Introduction II, lane three, clip four, MPOA 5, IP exchange 6, marker exchange seven, conclude
In the past ten years, ATM has become an important technique for next-generation networks, which can provide unprecedented scalability and cost performance, and support for future real-time business, multimedia business, etc. In the future information system, ATM will play an important role. However, the current information system, LAN and WAN, based on network layer protocols such as IP, IPX, AppleTalk, etc.,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, And the key to achieving this is the same network layer protocol, such as IP, IPX, applied to existing networks and ATMs because providing a unified network perspective to high-level protocols and applications is the network layer task. So far, there have been many ways to run IP on ATM, such as: ATM Forum LANE and MPOA, IETF CLIP and NHRP, IPSilon Network IP Exchange, and Cisco's mark exchange, will be introduced one by one.
First, introduction
ATM and existing protocol systems, especially network layers, IPX, etc., coexisting, how to implement existing network protocols and ATMs on a single network, how to interconnect ATM and traditional networks, The majority of researchers, designers and industry research topics. However, ATM and IP are derived from different technical groups and foundations, with their own applications. The purpose of IP is to send the packet to the destination in an uncertain state. It is non-connected, there is no guarantee of quality of service; and the purpose of ATM is to provide a guaranteed integrated business, which is a connection, fast fixed length letter. Yuan exchange. The huge difference in ATM and IP makes it effective to integrate both into a problem.
Two different models have two different models in the ATM network, and both models look at the relationship between ATM protocol layers and IP at different angles.
The first is the peer model, in essentially peer-to-peer layers, which is recommended to use the same address scheme in the IP-based network in the ATM network, so the ATM endpoint will be identified by IP address The ATM signaling will carry such an address, and the route of the ATM signaling also enables the existing network layer routing protocol. Because the existing routing protocol is used, the peer model excludes the need for developing new ATM routing. The peer model is simply simplified the address management of the end system, and the complexity of the ATM switch is largely increased because the ATM switch must have a multi-protocol router function, support existing address schemes and routing protocols. In addition, the existing routing protocol is based on current LAN and WAN development, and cannot be well mapped to ATM and using ATM service quality characteristics. In the current solution, IP exchange and marking exchange are based on peer-based models.
