ITMI982555A1 - METHOD OF DEVICE FOR STATISTICAL MULTIPLEXING IN TECHNIQUEWEIGHTED FAIR QUEUING (WFQ) OF FLOWS THAT CONVEY ATM TRAFFIC - Google Patents

Description

(DESCRIPTION of the industrial invention in the name:

Field of application of the invention

The present invention relates in general to communication systems which employ the numerical technique of fast cell switching, called ATM (Asynchronous Transfer Mode), used for the transport of voice signals, video signals and data.

More particularly, the invention relates to a method and a device for statistical multiplexing in Weighted Fair Queuing (WFQ) technique of flows that convey ATM traffic.

The ATM technique assumes an increasingly important role in the integrated switching of digital signal flows belonging to services for the transmission of voice, video and data signals, with different bandwidth requirements and different traffic characteristics.

In particular, the ATM technique provides that the information relating to the various services is organized in contiguous units with a fixed length of 424 bits (53 bytes), called cells. These cells contain, in addition to the actual data, a header field which carries, among other things, the information necessary for routing the cell through the geographic network.

The connection network, which has the task of spatially switching the cells from an entrance door to an exit door, must be able to handle high volumes of traffic, in the order of hundreds of Gbit / s.

The introduction of service priority mechanisms on the flow of ATM cells, whether they are represented by single connections or by sets of connections, is referred to as ASN (ATM Switching Network). A completely specular discourse applies to the section located on the outgoing side of this matrix. In this case, the outgoing flows from the ASN matrix constitute as many LINK-IN input flows for identical statistical multiplexers, SMU1. SMUk, and the LINK-OUT flows outgoing from the latter reach the ATM multiplexers (demultiplexers) AMX1, ..., AMXk, each of which sorts the traffic conveyed to a respective group of Line Interface Cards, LIC1, ... LICr operating at different bit-rates.

In operation, the ASN matrix is assumed to be self-routing and non-blocking. As is known, the main task of statistical multiplexers is to organize the outgoing traffic on the basis of the variability of the incoming traffic, dynamically distributing the band available on the outgoing link among the various active sessions, with the express purpose of maximizing the overall manageable traffic. from the ASN matrix, at the same time guaranteeing to the various sessions the minimum bandwidth agreed in the negotiation phase of the Traffic Management protocol. These purposes are achieved through a scheduling function, while a shaping function limits the bandwidth of the various connections at the top, limiting the loss of ATM cells due to congestion.

To meet these needs, SMU multiplexers have large buffers, and therefore able to adequately handle the channels that carry non-real-time traffic (nrt), such as the one defined in the nrt-VBR (nrt Variable Bit Rate), ABR (Available Bit Rate), and UBR (Unspecified Bit Rate). On the other hand, the channels that carry real-time traffic, such as that defined in the CBR (Constant Bit Rate) and rt-VBR service classes, obviously do not need large buffers.

Figures 2 and 3 of the present application are high-level schematizations of the statistical multiplexers SMU valid in general. Fig. 2 provides a functional diagram that indicates how the ATM cells that make up a LINK-IN input link are divided into n transmission queues QU1, QUn and stored in a buffer managed by a Buffer Managing BM function. Each queue groups cells coming from different elementary flows which constitute as many connections within the network, supported by respective virtual channels VC (Virtual Channel). The latter are conveniently distributed within different virtual transmission paths VP (Virtual Path). The grouping of the VCs by transmission queue can take place according to different criteria, for example, according to the bit-rate of the LIC line interface, or according to the VP path within the network.

Fundamental for the implementation of the aforementioned BM functionality is the fact of keeping the knowledge of the memory locations of the n queues and of the cells present in each queue updated. For this purpose, the Buffer Manager BM, for each acquired cell, updates a table of indicators, Q-ID1, .... Q-IDn of the address of the last cell occupied in each queue of cells (assuming that the stacking mode is known within tail). To carry out the aforementioned operation, the BM block analyzes Γ identification of the connection present in the Header part of the ATM cell, precisely in the VC (Virtual Channel) and VP (Virtual Path) fields, and uses this information to calculate a Q-ID indicator at the transmission queue. . The acquisition, or discard, of a cell is decided on the basis of the traffic class and the degree of occupation of the queue and the buffer as a whole. In case of acceptance, the BM requests the writing of the cell in the buffer and decides whether or not the Q-ID indicator should be memorized for the further functionalities ending with the queue service and the cell transmission.

Downstream of the BM block there is a SORTER block which performs a service function, called Cell Scheduling, consisting in suitably selecting the ATM cells to be sequentially placed on the LINK-OUT output link. There are many algorithms developed to perform the aforementioned functionality, among these the Weighted Fair Queuing, later called WFQ, assumes particular importance for the purposes of the invention.

Fig. 3 highlights the circuit setting of the SMU multiplexers in fig. 2, a description of the BM and SORTER blocks with good circuit detail will be provided below.

Returning to the aforementioned article and precisely to fig. 3, better explained in fig. 4 of the present application, it is possible to note the technological evolution that a single statistical multiplexer SMU undergoes with respect to the generic functional representation of fig. 2, regardless of the fact that it is placed at the input or at the output of the switching matrix ASN of fig.1. In the representation of fig. 4, in fact, the clear separation made between the transmission queues that support real-time traffic and the queues that support non-real-time traffic is evident. The presence of a modular structure is also evident in which the logical queues of non-real-time type QU cells, and the queues of the related Q-ID queue indicators, are grouped into m logical blocks identified by respective indicators B-ID1, .... B-IDm, while the real-time QU cells are grouped into a single queue (CBR) visible at the lower end of the figure.

Each B-ID block includes its own SB (Scheduler Block) sub-block which performs the scheduling function on the n logical queues Q-ID1, ..., Q-IDn of the block itself, and provides a sequence of indicators to a further scheduling block SBS (Scheduler Block Scheduler) common to all B-ID blocks and to the queue of real-time connections. The SBS scheduler recomposes the two types of traffic in the outgoing LINK-OUT stream.

In operation, the ATM cell queues of each B-ID block share a configurable and deterministic cell-rate, in more explicit terms this means that each B-ID block is assigned a fixed band, possibly different from block to block. The criteria for assigning the logical queues to the individual blocks are indicated in the aforementioned article. To generate the services of the B-ID blocks, the SBS scheduler makes use of a calendar in which the B-ID indicators are continuously inserted and extracted according to a suitable known criterion. Similarly, the scheduler SB of a generic B-ID block generates the services of the individual queues by means of a calendar containing the Q-ID indicators of the queues belonging to the B-ID block considered. When a block is served, its scheduler SB determines the queue entitled to extract a cell from the buffer. How the calendars are used depends on one, or the other, of two scheduled scheduling modes, and precisely the so-called Rate Shaping mode for limiting the peak cell-rate, and the WFQ mode to guarantee individual connections the minimum bandwidth agreed in the negotiation phase of the Traffic Management protocol is not real-time.

With regard to the above, it can be concluded that the structure highlighted above operates the subdivision of the band between queues belonging to different logical blocks by means of a set of physically distinct m + 1 calendars organized in a double hierarchical level. This architecture has the following two limitations:

• Impossibility of dynamically distributing the overall bandwidth among the blocks, given that each block has a fixed bandwidth, so any “transmission gaps” of a block cannot be exploited by the others.

• Impossibility of increasing the number of blocks at will, since the circuit implementation of a large number of BS schedulers (one scheduler per block) with a calendar mechanism, which in practical applications can easily assume large dimensions, finds a limit set by the technology of semi-custom integrated circuits.

The need to scan the calendar of a BS scheduler at each cell time to determine the queue of cells to be served, this involves the execution of numerous operations in a short time, requiring the use of very fast and therefore expensive circuits.

The last two limitations of the prior art of fig. 4 are partially eliminated thanks to an invention developed in the Applicant's laboratories, and already the subject of an Italian patent application filed with number MI97A 002147 on 09/23/1997. This invention, schematized in Figures 5 and 6, allows to dynamically divide the band between the transmission queues served by a generic block of queues using an original application of the WFQ algorithm. Unfortunately, the impossibility of dynamically distributing the band between the blocks continues to exist.

Fig. 5 illustrates the functional arrangement of the blocks of a statistical multiplexer made according to the aforementioned invention. Also in this case the transmission queues of the non real-time connections are divided into m functional blocks addressed by the block indicators B-ID1, ..., B-IDm; the Q-ID indicators at the individual queues undergo a similar distribution. The Q-ID indicators intended for a generic block are then treated by an optional SHP / SKD block which performs the shaping and scheduling functions to limit the peak band; after which at each new occurrence of an indicator, depending on the type of queue of cells to which it is associated, it is inserted in one of three FIFO (First In First Out) lists, respectively HP-FIFO, LP-FIFO, and VLP -FIFO, indicated in order of service priority. For the sake of brevity, the indication of a flag (semaphore) placed at the service of a queue that groups all the connections served in real-time, which therefore bypasses the WFQ functionality, has been omitted, and it should also be noted that the VLP-FIFO lists are optional. The FIFOs are used to carry out the functionality of WFQ in the way claimed in claim 1 which is reproduced below:

"Method for the management of resources in ATM technique for Weighted Fair Queuing applications in a system for the transmission on an output link of ATM cell flows coming from an input link belonging to classes of service of different quality, to at least some of said flows being associated with a respective minimum guaranteed band assigned during the service establishment phase, and each of said cell flows being ensured the possibility of using any additional band available on said output link, characterized by the fact of assigning to each of said flows, in a whole multiple renewal period of the transmission time of a cell on said outbound link, a respective number of access rights to said outbound link, or token, proportional to said minimum band, and, for each renewal period, to include the following stages of:

- insert the cells of each of certain flows in a respective transmission queue;

- assign to each of said transmission queues its own number of tokens;

- inserting indicators of transmission detainees inside a high priority memorization structure;

- scan the high priority storage structure by serving the transmission queues indicated by decreasing the tokens available to them and transferring the indicators of the queues that have run out of their tokens within a low priority storage structure;

- once the scanning of said high priority storage structure has been completed, scan said low priority storage structure by serving the queues indicated by it until the end of the renewal period ".

As can be seen, the possibility of dividing the cell queues into functional blocks does not at all transpire from the content of the claim. In the text this possibility is taken for granted, as it is only vaguely hinted at. In any case, the description of the means which distribute the service between the different B-ID blocks is completely lacking, and the reader can therefore reasonably assume that they are implicitly similar to those previously known, that is of the type of the SBS block of fig. 4. The latter is represented in fig. 5 by a SHP / SKD block which receives the B-ID indicators for this purpose.

Fig. 6 highlights the more properly systemic aspect of the statistical multiplexer represented functionally in Fig. 5; this figure constitutes a reference point for comparison with subsequent similar figures which will illustrate the invention which is the subject of the present application, and of some variants. As can be seen, the buffer manager BM provides the block B-ID and queue Q-ID indicators to the SHP / SKD-Q block, and the only block indicator B-ID to the SHP / SKD-B block. These two blocks, as mentioned, perform the shaping and scheduling functions for the service provided to the blocks (SHP / SKD-B) and to the queues inside the blocks (SHP / SKD-Q). The Q-ID indicators extracted from the SHP / SKD-Q block are each time written in one of the three FIFOs (fig. 5) of a corresponding group of FIFOs included in a WFQ-Q block, which performs the aforementioned WFQ functionality using a B-ID indicator extracted from the SHP / SKD-B block to address the group of FIFO served. The Q-ID indicators extracted from the WFQ-Q block are ultimately supplied to the BM block to complete the service functionality on! LINK-OUT output link.

The examination of Figs. 5 and 6 clearly highlights the residual limitation of the aforementioned invention, due to the fact that there is no functionality that allows to dynamically vary the band assigned to the blocks identified by the B-ID indicators. If then, as in practice, the optional SHP / SKD blocks are actually present, the limitations of the known art of fig. 4 also subsist in part, otherwise overcome, as the same architecture is proposed again in m + 1 calendars. . Aims of the invention

Therefore, the purpose of the present invention is to overcome the drawbacks of statistical multiplexers for ATM applications made by means of a functional block structure of cell queues according to the known art. More precisely, the invention aims to overcome, firstly, the limitation due to the inability to dynamically divide the available bandwidth between the blocks, in order to better meet the needs of real traffic, and secondly the typical limitations of scheduler array structures, or rather their implementation difficulty.

Another object of the present invention is to indicate a method for statistical multiplexing in the ATM technique which allows to make the most of the available band without creating congestion and drastically reducing the overall number of operations necessary to achieve these purposes.

Summary of the invention

To achieve these purposes, the present invention relates to a statistical multiplexing method of ATM cell flows originating from connections having service classes of different quality, said flows generating transmission queues of cells divided into functional blocks, and to at least some of said flows being assigned a minimum guaranteed band on the multiplexed flow and the possibility of distributing a residual band that may be available on the same flow by means of a known function, later called Weighted Fair Queuing, or WFQ, which dynamically assigns appropriate quantities of tokens, evaluated on a statistical basis , to individual transmission queues, which spend said tokens to access the residual band on the multiplexed stream (LINK-OUT), characterized by the fact that said WFQ functionality includes dynamic assignment phases of appropriate quantities of tokens, evaluated on a statistical basis, also to said functional blocks, which spend said tokens for allow its own transmission queues to access the distribution of the residual band on the multiplexed flow (LINK-OUT), thus dividing the residual band on the multiplexed flow among said functional blocks, as also described in claim 1.

Another object of the invention is a statistical multiplexing method that possesses all the characteristics of the method described in claim 1, and the further characteristic of performing at any cell time:

- a first scheduling of the emission intervals of indicators of the functional blocks to be served, adapted to dilate said intervals to limit the cumulative peak band of the group of queues of cells belonging to the served blocks;

- a second scheduling of the emission intervals of indicators of the transmission queues to be served, adapted to dilate said intervals to limit the peak band of the queues served, as described in a method dependent claim.

Another object of the invention is a statistical multiplexer that implements the method described in claim 1, as described in independent claim 13.

Another object of the invention is a statistical multiplexer which possesses all the means of the multiplexer of claim 13, and further means for scheduling the service intervals of the functional blocks and the related transmission queues, as described in a device-dependent claim.

Further objects of the invention are method and device variants, as they result from the respective dependent claims.

Advantages of the invention

A first notable advantage of the present invention is due to the fact that the application of the WFQ algorithm both at the block level and at the code level allows to extend the dynamic distribution of the band also at the block level, allowing the transmission resources to be optimally exploited. . Having said this, a further advantage deriving from it is obtainable by grouping all the connections of the real-time service classes of a block, for example CBR, by placing them in a high priority queue within the block itself. In this way, the WFQ algorithm applied to the queues of the considered block guarantees the minimization of the variation of the cell delay, or CDV (Celi Delay Variation), caused to the CBR connections by the other connections belonging to the block. However, the delay of CBR connections is affected by the delay in the service of the block to which they belong. If the choice of the system parameters makes this dependence a problem, it is possible to group the CBR connections of all the blocks in a single high priority block dedicated to them. As you can see, the invention allows you to create two different bypass levels in an extremely natural way, particularly suitable for the treatment of real-time transmission flows.

A second significant advantage of the statistical multiplexer of the present invention is that it only needs two calendars to control the peak band, both at block level and at queue level, limiting the band to both levels of the logical hierarchy. A first calendar is for blocks, and a second for queues; the latter is made available to the queues of the blocks selected each time. Therefore, with respect to the architectures of the prior art which replicate calendars as many times as there are blocks, the logic necessary to control the peak rate undergoes a drastic reduction in complexity. As a result, with the same number of ports of the statistical multiplexer, a decrease in the size of the integrated circuits that make up the calendars and a consequent reduction in production costs. Furthermore, the removal of the technological constraint on the maximum number of blocks imposed by the replication of calendars makes it possible to build systems with a number of ports considerably greater than that currently possible, which is around a hundred. The architecture of the statistical multiplexer according to the present invention allows to impose the limitation of the peak in a selective way, since there can exist queues to which the pure WFQ algorithm is applied, i.e. with an unrestricted peak rate, and queues which are limited at the top in the band. . This property, which also applies to blocks, can be obtained by configuring appropriate parameters via software.

Brief description of the figures

Further objects and advantages of the present invention will become clear from the following detailed description of an example of its embodiment and from the annexed drawings given purely for explanatory and non-limiting purposes, in which:

- Fig. 1 shows an architecture commonly found in ATM network nodes; - Figs. 2 and 3 show high-level schematizations of the statistical multiplexers SMU of fig. 1 valid in general;

- fig. 4 shows a functional schematic of the SMU block of fig. 2, referring to the known art;

- fig. 5 shows a further functional schematic of the SMU block of fig. 2, as shown in a previous patent application by the same Applicant;

- fig.6 indicates a functional block diagram of the SMU block of fig.3 when the functional diagram of fig.5 is valid;

- fig.7 shows a functional schematic diagram of the SMU block of fig.2 realized according to the present invention;

- fig.8 shows a functional schematic of a SHP / SKD block of fig.7; - fig. 9 shows a functional block diagram of the SMU block of fig. 3 when

the functional diagram of fig.7 according to the invention is valid;

- fig.10 shows a circuit diagram of a block BM of fig.9;

- fig.11 shows a circuit diagram valid for both blocks WFQ-Q and WFQ-B of fig.9;

- fig.12 shows in detail a generic block TK-CNTj and a second block TK-CNT-B of fig.11;

- figures 13 to 16 show flow diagrams which schematize the operation of the WFQ-B block of fig. 9;

- fig.17 shows a functional block diagram of a variant of the block representation of fig.9;

- fig. 18 shows a flow diagram which schematizes the operation according to the variant of fig. 17; And

- figs. 19 and 20 show functional block diagrams of further variants of the block representation of fig. 9.

Detailed description of some preferred embodiments of the invention With reference to Fig. 7, the statistical multiplexer object of the present invention is now described. With respect to the general description of Fig. 2, the transmission queues of cells QU which group different connections, are here divided into m functional blocks addressed by the block indicators B-ID1. B-1Dm; the Q-1D indicators undergo a similar distribution to the individual transmission queues.

In fig. 7 two apparently distinct functions of WFQ can be seen, a first for the treatment of the B-ID indicators, and a second for the treatment of the Q-ID indicators. A pair of indicators, B-ID and Q-ID, is generated for each occurrence of a cell. The B-ID indicator, depending on the type of pre-established distribution between the transmission queues of ATM cells and blocks, is, if absent, inserted in one of three FIFO lists, respectively indicated in order of service priority with HP-FIFO, LP -FIFO, and VLP-FIFO. The Q-ID indicator, depending on the nature of the queue to which it is associated, is, if absent, inserted in similar FIFO lists combined with the block identified by the corresponding B-ID indicator. The two groups of FIFOs are "individually" used to perform the WFQ functionality in a similar way to that described, only for the transmission queues, in the aforementioned patent application in the name of the same Applicant. The B-ID indicators selected by the first function of WFQ are then processed by a SHP / SKD block that performs the shaping and scheduling functions to limit the peak band of the set of transmission queues assigned to the generic block, after which the B-ID indicators scheduled from time to time are placed available to the second WFQ functionality for the selection of the group of three FIFOs to be served. Similarly to what was done for the B-ID indicators, the Q-ID indicators provided by the second WFQ functionality are treated by a SHP / SKD block that performs the shaping and scheduling functions to limit the peak bandwidth of multiplexed flows.

It should also be noted that the VLP-FIFO lists are optional, as are the SHP / SKD blocks.

In a first variant embodiment of the present invention an empty queue flag is provided at the service of a queue which groups all the connections that transmit at the peak rate (CBR and rt-VBR) of a generic block addressed by B-ID, and which therefore need to be served in real-time. The indication of this flag in the figures illustrated so far has been omitted, for the sake of brevity.

In a second variant, all the connections that transmit at the peak rate of all the blocks are grouped into a single high priority block dedicated to them, in this case both the WFQ functions are bypassed and the Q-ID indicator is directly supplied to the relative SHP / SKD block for its scheduling and priority insertion of the cell on the output link. Also in this case, in the figures illustrated so far, the indication of the empty queue flag associated with the single queue that groups all the ATM cells of the real-time connections of all the blocks is missing.

With reference to fig. 8, it can be noted that the generic SHP / SKD block, whether it serves the transmission queues indicated by Q-ID or B-ID, includes a SHAPER-Q (SHAPER-B) block, which performs a shaping function, placed upstream of a SCHEDULER-Q block (SCHEDULER-B), which in turn performs a scheduling function. The shaping function calculates a waiting time, at the limit of zero, and converts the information into an N-SLOT-Q (N-SLOT-B) value which represents a slot of a calendar having a finite length (shown in the figure) included in the SCHEDULER-Q (SCHEDULER-B) block. The Q-ID (B-ID) indicator that reaches the shaper and the N-SLOT-Q (N-SLOT-B) value calculated by the shaper are transferred to the scheduling function which inserts the Q-ID (B- ID) in the calendar location indicated by N-SLOT-Q (N-SLOT-B). The scheduling function provides for the cyclic scanning of the calendar with cadence 1 / Tc, where Tc is the generic cell time, that is the time interval necessary for the serial transmission on the link of all the bits of the cell. At each scanned location, the indicators stored there are extracted from the calendar on a FIFO basis. In figure TNOW represents the current cell time, or the multiple time of Tc elapsed from the start of the scan cycle. Time TNOW indicates in the figure an empty location, while TV time indicates a previous calendar location.

The shaping functionality consists in limiting the band of queues and blocks at the top. For this reason it calculates the Q-ID insertion time according to the following law:

INSERTION TIME = max [PREVIOUS (N_SLOT_QxTc) 1 / PCR, TNOw],

where PREVIOUS (N_SLOT_Q) indicates the location of the calendar in which the Q-ID indicator was last entered, and POR (Peak Celi Rate) is the peak cadence of the queue indicated by Q-ID. The same law is valid for blocks as long as you replace N-SLOT-Q with N-SLOT-B and define a PCR peak rate for each block. The configuration of the PCR parameters in the SHAPER_Q and SHAPER_B blocks is done via software. As a result of the above mentioned insertion law, the calendar scan will appropriately delay the emission of the queue and / or block indicators, ensuring that they are not served too quickly, thus violating the band assigned to them.

With reference to fig. 9, it can be noted that the systemic representation shown there differs from that of the known art of fig. 6 for the following:

the presence of a WFQ-B block placed immediately upstream of the SHP / SKD-B block, here separated into its SHAPER-B and SCHEDULER-B components;

the WFQ-Q block is placed upstream of the SHP / SKD-Q block, here separated into its SHAPER-Q and SCHEDULER-Q components;

the B-ID indicator coming out of the SCHEDULER-B block is also supplied to the buffer manager BM as Q-ID.

The functional block diagram in fig. 9 highlights the presence of two indicator travel rings. The innermost ring is crossed by the Q-ID indicators that start from the BM block and cross in the order: WFQ-Q, SHAPER-Q, SCHEDULER-Q, and return to the starting BM block. The outermost ring is crossed by the B-ID indicators which start from the BM block and cross in the order: WFQ-B, SHAPER-B, SCHEDULER-B, and return to the starting BM block. The second ring interacts with the first at the level of the WFQ-Q block. The reference to the rings has a mostly descriptive value.

With reference to fig. 10, the circuit architecture of the BM buffer manager in fig. 9 is now illustrated at a general level. As can be seen in the figure, the BM block includes the following sub-blocks: a controller of the ACC-CONTR access functionality, a controller of the SER-CONTR service functionality, a CODE-CONT counter associated with the transmission queues, a buffer of the transmission queues BUF-CEL, a series / parallel S / P converter, a parallel / series P / S converter, and finally a SYNC distributor of the CK network clock. The controller blocks and the cell buffer communicate with a bidirectional data bus DATA-BUS, a BUS-IND address bus, and control lines. The LINK-IN input link reaches the S / P converter and from there to an input of the ACC-CONTR block, while the outgoing signal from the SER-CONTR block reaches the P / S converter for input on the output link LINK-OUT. The ACC-CONTR block sends a first WR-EN write enable signal to the BUF-CEL buffer for writing an ATM cell, and a second WR-EN-P signal to a WFQ-CONTR unit for execution of the WFQ (visible in fig. 11) for writing the B-ID and / or Q-ID pointers associated with the relative cell, a TR-TYPE value that characterizes the type of traffic supported by the connection to which the cell belongs, and a first and / or a second count value VAL-CNT of two counters respectively associated with the queue hosting the cell and the block hosting the queue. The SER-CONTR block sends an RD-EN read enable signal to the BUF-CEL block. The CODE-CONT counter receives from the ACC-CONTR block an INCR-CONT signal for increasing the count and a READ-CONT signal for reading the count, and from the SER-CONTR block a signal for decrementing the DECR-CONT count. In response to the READ-CONT signal, the counter provides the data read VAL-CNT. The SER-CONTR block receives two types of pointers (indicators) B-ID and Q-ID from the WFQ-Q and WFQ-B blocks (fig. 9), or from the SCHEDULER-Q and SCHEDULER-B blocks (fig. .9) if these last two blocks are not optional.

Before illustrating the operation of the buffer manager BM, which has already been largely anticipated in the introduction, it is worth describing figs. 11 and 12 and then connect them to fig. 10 in operation.

With reference to ftg.11, the circuit architecture of the WFQ-Q and WFQ-B blocks of fig.9 is now illustrated at a general level. As can be seen, and without any limitation for the invention, the architecture refers to the case of control of the two blocks WFQ-Q and WFQ-B through a single execution unit of the WFQ algorithm indicated with WFQ-CONTR. The above unit is connected to the data bus and the address bus of the! buffer manager BM and receives a WR-EN-P write enable signal from the ACC-CONTR block. The WFQ-CONTR unit communicates with the WFQ-Q and WFQ-B blocks via an internal BUS-DATA-WFQ bus and sends the read / write signals obtained from the decoding of the information received on the BUS-address bus to the aforementioned blocks. IND or self-generated. The signals sent to the WFQ-Q block are the following:

• W / R-ENH1, ..., W / R-ENHm; W / R-ENL1, ..., W / R-ENLm; W / R-ENV1, ..., W / R-ENVm; • W / R-ENQ1. W / R-ENQm;

• W / R-ENR1, .... W / R-ENRm.

The signals sent to the WFQ-B block are the following:

• W / R-ENH-B; W / R-ENL-B; W / R-ENV-B;

■ · W / R-ENB and W / R-ENR-B.

The WFQ-Q block comprises m groups B1. Bm, each of three FIFO, HP-FIFO, LP-FIFO, and VLP-FIFO memories. Block WFQ-B, on the other hand, comprises a single group of FIFO memories. The WFQ-Q block also includes m flags RT-FU \ G and m counters TK-CNT1. TK-CNTm, the function of which will be explained later, while block WFQ-B includes a single flag RT-FL ^ G and a single counter TK-CNT-B. Inside the WFQ-Q block, the read / write signals W / R-ENH reach the HP-FIFO memories, the W / R-ENL signals reach the LP-FIFO memories, the W / R-ENV signals reach the VLP- memories. FIFO, the W / R-ENR signals to the RT-FLAG flags, and the W / R-ENQ signals to the TK-CN counters in the corresponding numeral order. Inside the WFQ-B block the read / write signals W / R-ENH-B, W / R-ENL-B, W / R-ENV-B reach the HP-FIFO, LP-FIFO, VLP memories respectively -FIFO, the W / R-ENR-B signal to the RT-FLAG flag, and the W / R-ENB signal to the TK-CNT-B counter. The WFQ-CONTR unit places on the internal data bus BUS-DATA-WFQ the pointers B-ID and Q-ID, the values to be written, or read, in the RT-FLAG flags and in the TK-CNT counters selected at each cell time by carrying out the functionality of WFQ.

With reference to fig. 12, the format of the j-th counter TK-CNTj is shown between the counters TK-CNT1, ..., TK-CNTm of the block WFQ-Q, and of the counter TK-CNT-B of the block WFQ -B. It is useful to premise that for n queues within a block the following relation holds: z s = n, where z is the number of ABR connections, and s is the number of non-ABR connections. The counter TK-CNTj is a memory structure comprising z fields, N ° TOK (QU1). N ° TOK (QUz), each containing a number of '' tokens ", ie tokens, available to each transmission queue QU1, ... QUz belonging to the generic block Bj. Similarly, the counter TK-CNT-B includes m fields , N ° TOK (B1), ..., N ° TOK (Bm), each containing a number of tokens available to a corresponding block B1, Bm.

In operation, what has been said about fig. 2 applies to the BM block, to which reference is made, with the foresight that the B-IO block pointers are now added to the only Q-ID pointers. The previous illustration carried out in fig. 2, given the extreme generality of that figure, was not exhaustive about the fate of the pointers. This gap is now filled, thanks also to what has been said about the functionality of WFQ in fig. 5. This place the BM block performs the access function and ends the service one, essentially performed by the WFQ-CONTR unit. The transfer of a Q-ID pointer from the ACC-CONTR control block to the WFQ-CONTR unit is necessary if at least one cell belonging to the considered queue is not already present in the BUF-CEL buffer. This information is contained in suitable queue counters included in the CODE-CNT counting block, each of which indicates the number of cells belonging to a respective queue present in the BUF-CEL buffer. The counters are updated by signals INCR-CNT and DECR-CNT, respectively at each acceptance and service of the queue in question. The transfer of B-ID is necessary if at least one cell belonging to a transmission queue assigned to the block in question is not already present in the BUF-CEL buffer. This information is obtained from block counters included in the CODE-CNT block, the meaning of which in relation to the blocks is similar to that of the queue counters in relation to queues. Any transfer of B-ID and / or Q-ID is completed with the information of traffic type TR-TYPE. More specifically, there are three possible traffic type values that give rise to as many destinations in the transfer of Q-ID pointers:

- HPT (High Priority Traffic): the Q-ID pointer is inserted in a respective high priority HP-FIFO memory of the WFQ-Q, so as to guarantee its privileged extraction by minimizing the time spent in the FIFOs of the WFQ-Q blocks . A queue with a traffic type value of HPT is suitable for holding CBR or rt-VBR connections.

- LPT (Low Priority Traffic): the Q-ID pointer is inserted in a respective low priority LP-FIFO memory of the WFQ-Q as long as it has available tokens; once the permissions have been exhausted, the pointer is inserted into a very low priority VLP-FIFO memory. A queue characterized by a traffic type value equal to LPT is suitable for containing nrt-VBR connections of the ABR type (with available bandwidth). This traffic type value allows to guarantee a minimum bandwidth to the queues it characterizes.

- VLPT (Very Low Priority Traffic): the Q-ID pointer is inserted in a respective very low priority VLP-FIFO memory of the WFQ-Q. A VLP-FIFO memory is only served when the corresponding higher priority FIFOs are empty. A queue characterized by a traffic type value equal to VLPT is suitable for containing UBR connections, therefore a minimum bandwidth is not guaranteed.

As for the subdivision of the B-ID pointers within the HP-FIFO, LP-FIFO, and VLP-FIFO memories of the WFQ-B block, the situation is not as well defined as in the case of the Q-ID pointers to the individual code. In fact, having to decide in which FIFO to store a Q-ID pointer, it is only necessary to know the type of queue to which it is associated, the information is provided by the traffic-type parameter. The distinction between flows with a minimum unsecured band and flows with a minimum guaranteed band therefore allows the seconds to be assigned to the LP-FIFO whose service involves the management of the tokens. In a block, on the other hand, different types of queues coexist as regards the class of service. It is therefore necessary to establish a criterion to divide the B-ID pointers among the three FIFOs of the WFQ-B block. A first criterion could be to divide the blocks (and therefore their B-ID pointers) into three groups of equal aggregate bandwidth and to assign the groups to the FIFOs, in order of priority, based on the number of real-time queues contained. in each group. This choice could cause an excessive increase in the CDV of the queues included in the blocks that are part of the group associated with the VLP-FIFO memory, so a second distribution criterion could no longer consider the three groups of blocks equibanda but rather of increasing bandwidth in the sense of decreasing priority. In this way, the monopoly of the service by the group with higher priority is avoided, and, for the same reasons, by the group with intermediate priority in relation to the one with lower priority. Whatever the block distribution criterion, and if you want to ignore the second embodiment variant, it is advisable to make sure that the block WFQ has the same operation as the queue WFQ, allowing equal flexibility in the treatment of the blocks. More details will be provided by illustrating the operating flow-charts of the block WFQ.

The previous considerations on the conceptual and real difficulty of dividing the blocks into decreasing priority groups, together with the difficulty of dividing the transmission queues among the different blocks in order to make the chosen priority effective, are clear indications of the non-obviousness of the present invention, to be opposed to possible conclusions of an analysis which, very reductively, could see the present statistical multiplexing as the simple duplication of a known WFQ functionality. The lack of arguments in favor of this erroneous conclusion will also appear by continuing to illustrate the functioning and the tenor itself of the claims, where it will be noted how the phases that make the statistical multiplexing of flows dynamic, namely the phases of assignment and renewal of tokens, are carried out jointly for the blocks and for the respective queues. This means that there is a more extensive WFQ functionality capable of integrating the two complementary treatments reserved for transmission blocks and queues, while maintaining distinct the phases that lead to the emission of block and queue indicators. However, these phases are not independent, since the phases that lead to the emission of a transmission queue indicator can start only after a functional block has been selected, in this way the distribution of the residual band between the transmission queues is conditioned. The interrelationships between the block and queue WFQ are too significant for one to think of a simple replication of a known algorithm, it would seem much more logical to recognize the originality of the extended functionality of WFQ used in the present invention.

Continuing in the illustration of the operation, it can be seen that the acceptance of a cell from a transmission queue of cells extracted from the LINK-IN input stream (fig. 1) ends with the insertion of pointers B-ID and / or Q -ID in the appropriate FIFO of the WFQ. Compliance with the previously described insertion condition ensures that each block and queue pointer appears at most once in the FIFOs of the WFQ. If necessary, the insertions of the Q-ID and B-ID pointers are made in parallel.

The service of a transmission queue of cells to be inserted in the LINK-OUT output flow (fig. 1) involves, at each cell time, the following operational phases carried out sequentially:

a) extraction of a B-ID pointer from the FIFO group of the WFQ-B block (by running the WFQ algorithm);

b) possible transfer of the B-ID pointer to the SHAPER-B block (fig. 9) and from these to the SCHEDULER-B block for the emission of a B-ID pointer to a block entitled to the service with a limited peak rate ;

c) extraction of a Q-ID pointer from one of the FIFO groups of the WFQ-Q block selected by the B-ID pointer issued in the previous phase (by running the WFQ algorithm);

d) possible transfer of the Q-ID pointer to the SHAPER-Q block (fig. 9) and from these to the SCHEDULER-Q block for the emission of a Q-ID pointer to a transmission queue entitled to the service with peak rate limited;

e) transfer of the pointers B-ID and Q-ID thus determined to the buffer manager block BM for carrying out the following operations:

- extraction of a cell from the BUF-CEL buffer by the SER-CONTR block for its insertion in the LINK-OUT output link;

- generation of the DECR-CNT signal for decrementing the block and queue counters of the CODE-CNT count block;

- generation of the READ-CNT signal for reading the VAL-CNT value of the block and queue counters of CODE-CNT;

- possible generation of the WR-EN-P signal for the rewriting of B-ID and / or Q-ID in the respective FIFO group if the respective VAL-CNT value is not null.

As will be better seen shortly, the B-ID and Q-ID pointers are issued by the WFQ-B and WFQ-Q blocks according to the following law:

• if the HP-FIFO memory contains at least one pointer, it is extracted; • if the HP-FIFO memory is empty and the LP-FIFO memory contains at least one pointer, it is extracted;

• if the HP-FIFO and LP-FIFO memories are both empty and the VLP-FIFO memory contains at least one pointer, it is extracted.

With reference to figs. 13, 14, 15, and 16, and to what has been said in the previously cited patent application filed by the same Applicant about the methods of execution of the WFQ algorithm applied only to the transmission queues, the methods are now better specified. execution of the same in the context of the statistical multiplexing method object of the present invention, specifying that in the new context the aforementioned WFQ algorithm is applied in an innovative way to select the blocks, as well as to serve the respective transmission queues.

The extension of the aforementioned algorithm also to blocks of transmission queues is an absolute novelty of the present invention, and therefore the methods of such an application will be accurately illustrated with reference to the main resources included in the WFQ-B block. With the appropriate distinctions, the notion set out will continue to apply to the WFQ-Q block as well.

The cases relating to the first and second embodiment variant of the present invention, which provide for the use of RT-FL4G to establish an absolute priority on isolated real-time flows in dedicated queues, or in a block, are a direct consequence of the complex of concepts exhibited, or cited. Therefore, only the second variant will be briefly illustrated, which is also a reason for novelty. Having said this, please refer to what has already been said about the subdivision into service classes of the different flows entering a network node and multiplexed in the single LINK-IN flow (fig. 1) at the entrance of the generic statistical multiplexer SMU. It should be noted that there is no a priori relationship between the band of the incoming link and the band of the outgoing link LINK-OUT, and to summarize that the incoming flow is made up of cells belonging to a certain number of flows with different levels of QoS . These flows consist of a number z of transmission flows of available bandwidth or ABR applications (such as, for example, the non "real time" flows relating to a data file) for which the system still guarantees a minimum bandwidth and maximum availability of the excess band, and in a number s of transmission flows of non-ABR applications (such as, for example, the "real time" flows relating to television signals, etc.), for which the system must in any case guarantee the established peak band ( CBR connections or of highest priority). To the aforesaid flows are added the flows of the UBR type, for which the service does not guarantee a minimum bandwidth. The QoS classes mentioned, together with what has already been said for the blocks, suggest the criterion for distributing the B-ID and Q-ID pointers among the FIFO memories of the WFQ.

The assumption on which the WFQ algorithm is based is that there is available bandwidth to be distributed, therefore, called Rtot the total band available on the LINK-OUT link, called RAj the minimum contracted bands of the z transmission flows ABR (or more precisely with minimum guaranteed band) and called RNAj the fixed transmission bands of s non-ABR transmission flows (excluding UBR) the relation ∑, RAi + ∑, RNAi S Rtot must apply. correct assignment of RAi and RNAj values to individual applications.

The token mechanism was created to guarantee ABR queues the guarantee of a minimum bandwidth. For this purpose, the number of tokens assigned in a certain TR period, called renewal period, to each transmission queue belonging to an ABR transmission flow, is proportional to the minimum band assigned during the negotiation. The renewal period TR is an integer multiple of the cell time Tc. It is defined as: TR = (Rtot / G) * Tc where the G factor characterizes the granularity linked to the minimum manageable band. The number of tokens assigned to each queue belonging to ABR transmission flows is equal to RAj / G. The variation of the TR period, that is of the granularity with which the tokens are distributed, allows to optimize in every situation the performances of the statistical multiplexer of figs. 10 and 11.

Advantageously, the contracted bands can be easily varied simply by changing the number of tokens available to each application. When negotiating, it is also possible to assign tokens in such a way that the sum of all the minimum bands of all the ABR flows and of all the peak bands of the non-ABR flows (excluding UBR) is always less than the total band of the link in exit. This would make it possible to always have a certain excess bandwidth available to distribute. In fact, tokens can be seen as a unit of measurement of bandwidth, so who decides whether or not to accept a connection, must verify that the traffic conditions on the network allow to guarantee compliance with the parameters for that connection and the others already present. of traffic that characterize them. If this evaluation is made using the "token" unit of measure, then tokens must be assigned to each connection with fixed band and to each connection with minimum guaranteed band. From the above it follows that during the acceptance phase token also to the QUj queues whose pointers are contained in the HP-FIFO memories of the WFQ-Q block. This same reasoning also applies to the blocks, since the band parameters of each block result from the bands of the respective queues, so in phase tokens are also assigned to the blocks whose B-ID pointers are contained in the HP-FIFO memory of the WFQ-B block. the blocks never exceed the total output band (LINK-OUT).

However, the fact of using the tokens exclusively to guarantee the minimum bands remains valid for the statistical multipiexer according to the present invention, which is why fixed band connections are always served with high priority, regardless of the value of the tokens that those who decide to accept assign to each of them. The question then arises spontaneously: and if a fixed band connection (eg CBR) transmitted at a higher bandwidth than its PCR (Peak Celi Rate), would it not actually receive more than it should? The answer is affirmative, but this cannot happen because, as is known, in the input nodes of the network (fig. 1) there are components called "policers" (not shown in the figures) which verify and ensure that each CBR (VBR) connection does not transmit at a higher bandwidth than its PCR (SCR).

With reference to Fig. 13, the method used in the present invention for the distribution of the available band between the different B-ID blocks is illustrated. The method is described for a single TR renewal period.

At the beginning of the TR period, the WFQ-CONTR control unit (fig.11) goes into an INITIALIZE TR-CNT phase in which a TR-CNT counter is initialized with the TR value of the token renewal period.

An INITIALIZE TK-CNT-B, TK-CNT1, ..., TK-CNTm phase follows, during which the storage devices of the tokens indicated above are initialized (fig. 12) with the token values N ° TOK (B1) . N ° TOK (Bm) assigned to the respective fields of the counter TK-CNT-B, and with the token values N ° TOK (QU1), ..., N <D> TOK (QUz) assigned to the respective fields of the m counters TK-CNTj. The number of tokens N ° TOK (Bj) that a generic block actually uses to guarantee the minimum block band is given by the following expression:

where z is the number of ABR-type transmission queues QU assigned to the j-th block.

Subsequently the WFQ-CONTR unit goes into a phase INITIALIZE THE FIFO OF THE WFQ-Q, WFQ-B BLOCKS during which any pointers contained in the aforementioned memories are removed, and are suitably inserted in the groups of memories HP-FIFO, LP- FIFO, VLP-FIFO of the block WFQ-Q of the Q-ID pointers to transmission queues which have cells to be transmitted, and in the homonymous memories of the WFQ-B block of the B-ID pointers to the blocks which include the aforesaid queues.

The initialization steps illustrated so far clearly show how up to this point two independent WFQ functions are not distinguishable at all, respectively for the transmission blocks and queues, since in fact the relative treatments are carried out jointly. From this point forward it is possible to illustrate the two functions separately, but this does not mean that the second is independent from the first.

The WFQ-CONTR unit in the TR-CNT - Tc → TR-CNT phase decreases the counter TR-CNT by a quantity corresponding to the cell time Tc, and goes into a test phase TR-CNT = 0 in which it checks whether the TR time has not ended.

If the TR time has not yet expired, the WFQ-CONTR unit goes into a reading phase READ HP-FIFO during which a B-ID pointer is read, followed by a test phase IS HP-FIFO EMPTY? in which it checks whether in the aforementioned FIFO there are no more B-ID block pointers with high service priority yet to be emitted.

If the test indicates that these pointers are present, the WFQ-CONTR unit moves into a SERVI HP-FIFO phase during which the block whose B-ID pointer was read in the READ HP-FIFO phase is served. At the end of the SERVI HP-FIFO phase, which will be illustrated in fig. 14, the WFQ-CONTR unit returns to the phase in which TR-CNT is decreased and cyclically repeats the phases between the latter and the service phase until until the time TR is exhausted, or until the HP-FIFO memory itself has not been emptied.

When this happens, the test IS HP-FIFO EMPTY? is true and the WFQ-CONTR unit goes into a reading phase READ LP-FIFO during which a B-ID pointer is read, followed by a test phase IS LP-FIFO EMPTY? in which it checks whether in the aforementioned FIFO there are no more low service priority block B-ID pointers yet to be emitted.

If the test indicates that these pointers are present, the WFQ-CONTR unit goes into a SERVI LP-FIFO phase during which the block whose B-ID pointer was read in the READ LP-FIFO phase is served. At the end of the SERVI LP-FIFO phase, which will be illustrated in fig. 15, the WFQ-CONTR unit returns to the phase in which TR-CNT is decreased and cyclically repeats the reading and service phases of the LP-FIFO memory until until the time TR is exhausted, or until the LP-FIFO memory itself has not been emptied.

When this happens, is the test LP-FIFO EMPTY? is true and the WFQ-CONTR unit goes into a reading phase READ VLP-FIFO during which a B-ID pointer is read, followed by a test phase IS LP-FIFO EMPTY? (transferred for ease of representation at the beginning of fig. 16) in which it checks whether in the aforementioned FIFO there are no more B-ID block pointers with very low service priority still to be emitted.

If the test indicates that these pointers are present, the WFQ-CONTR unit goes into a SERVI VLP-FIFO phase during which the block whose B-ID pointer was read in the READ VLP-FIFO phase is served. At the end of the SERVI VLP-FIFO phase, which will be illustrated in fig. 16, the WFQ-CONTR unit returns to the phase in which TR-CNT is decreased and cyclically repeats the reading and service phases of the VLP-FIFO memory until until the time TR is exhausted, or until the VLP-FIFO memory itself has not been emptied. In both cases the START phase is restarted and the WFQ-CONTR unit goes back to the initialisation phase of the TR-CN counter in which the original value of the TR renewal period is restored in the TR-CNT counter, to which it followed by the phase in which the tokens in the counters TK-CNT1, ..., TK-CNTm and TK-CNT-B, FIFO are re-calculated and redistributed, and the phase in which the contents of all the FIFOs of the WFQ blocks are reinitialized -Q and WFQ-B.

With reference to fig. 14, the SERVI HP-FIFO phase of fig. 13 is illustrated which begins with a phase EMIT B-ID ON BUS-DATA-WFQ whose meaning is obvious.

A phase follows in which the VAL-CNT value referred to the occurrences of the block indicated by B-ID is read, and then a test is performed on the value read to determine if there are still blocks of that type to be served. This will be true if there is at least one cell still to be transmitted in at least one queue assigned to the block just served.

If the VAL-CNT test = 0? is false, the WRITE B-ID IN HP-FIFO phase is performed to reinsert the B-ID pointer in the HP-FIFO memory, otherwise the WFQ-CONTR unit returns to the TR time decrease phase and repeats the emptying cycle of HP-FIFO, thus ending the service phase.

With reference to fig. 15, the SERVI LP-FIFO phase of fig. 13 is illustrated which begins with a phase EMIT B-ID ON BUS-DATA-WFQ whose meaning is obvious. The VAL-CNT value referred to the occurrences of the block indicated by B-ID is then read, and then a test is performed on the value read to determine if there are still blocks of that type to be served.

If the test VAL-CNT = 0 is false, the WFQ-CONTR unit performs in phase IS TK-CNT-B = 0 a test on the value of the field of the token counter TK-CNT-B referred to the block indicated by pointer B- ID.

If the test shows that the block in question still has tokens to spend, the B-ID pointer is reinserted into the LP-FIFO memory and the value in the field of the token counter TK-CNT-B is decreased by one unit. Then the WFQ-CONTR unit returns to the phase of decreasing the TR time and repeats the LP-FIFO emptying cycle.

If the test on the token counter TK-CNT-B shows that the block indicated by B-ID has already spent all its tokens, the B-ID pointer is inserted into the VLP-FIFO memory, after which the time TR is decreased by one unit the LP-FIFO emptying cycle is repeated, thus ending the service phase.

Fig. 16 illustrates the SERVI VLP-FIFO phase of Fig. 13 which is completely analogous to the SERVI HP-FIFO phase of Fig. 14, except that the B-ID pointer is now inserted into the VLP-FIFO memory.

It is quite clear that, temporarily disregarding subsequent scheduling operations of the B-ID and Q-ID indicators, once a phase is completed EMIT B-ID ON THE DATA-BUS-WFQ, the execution of those phases of the WFQ applied to the transmission queues which lead to the emission of a Q-ID indicator and which are completely analogous to the corresponding steps illustrated for the blocks.

From the flow diagrams illustrated above, the salient features of the block WFQ can be summarized, saying that the high-priority FIFO HP-FIFO contains B-ID pointers to functional blocks that are always served, regardless of the tokens assigned to them during the negotiation phase. .

The low-priority FIFO LP-FIFO contains B-ID pointers to functional blocks which still have tokens available and which, therefore, are definitely entitled to be served.

The very low priority FIFO VLP-FIFO contains B-ID pointers to functional blocks that have already spent all the tokens at their disposal and that, therefore, have certainly already fully exploited the minimum bandwidth granted to them. These blocks therefore have the same right to be served, so that the residual band is evenly distributed among them.

During the time interval TR the LP-FIFO memory is scanned cyclically and the blocks indicated by it are uniformly served. These blocks, due to the decrease in the number of tokens and their call for the service of the respective transmission queues, can come to be in different states.

A block can simultaneously run out of its tokens and have no more queues to serve. This situation will occur when, during the TR period, the aggregate cell-rate was exactly equal to the minimum band assigned to the block.

On the other hand, a block may no longer have queues to serve before it has run out of its tokens. This means that this block will include transmission queues whose aggregate cell-rate, during the TR period, was lower than the minimum band assigned to the block.

A block may eventually run out of its tokens before the TR period expires. This means that this block will include transmission queues whose aggregate cell-rate, during the TR period, was higher than the minimum band assigned to the block.

During the TR period, due to the LP-FIFO memory emptying activity, the following two cases may occur:

In the first case, the emptying of the LP-FIFO coincides with the end of the TR period. This means that all functional blocks have transmitted, during the TR period, with an aggregate cell-rate equal to the speed determined by the agreed minimum band. In this case, during that period, there will be no excess bandwidth available to distribute between the blocks.

In the second case, the emptying of the LP-FIFO arrives before the end of the TR period. This means that at least one of the functional blocks transmitted, during the TR period, with an aggregate cell-rate lower than the speed determined by the agreed minimum band. In this case, during this period, there will be excess bandwidth available to eventually be distributed among the blocks. The VLP-FIFO memory is scanned cyclically so that all the blocks whose indicator is present are uniformly served, regardless of the number of tokens they originally owned and regardless of the number of queues that they still have to serve (Round-Robin ).

It will be appreciated that, in this way, even in the worst case in which all the blocks have queues whose aggregate cell-rate is higher than the minimum bandwidth assigned to the respective block, the same blocks are correctly served at least with the agreed minimum bandwidth. It is then further appreciable. the fact that, thanks to the VLP-FIFO memory scanning mode, no block is able to monopolize the service.

In the specific case of the second implementation variant, in addition to the three FIFO memories, the RT-FLAG flag is used to control the single block that groups all the transmission queues with the need for real-time service. In this case the WF-CONTR unit, once the initial phase has been completed, immediately proceeds to carry out a test on the RT-FL4G value to see if there is a block to be served with absolute precedence, even with respect to the high priority FIFO HP- FIFO. Once the aforementioned block has been emptied and the flag deactivated, the WF-CONTR unit proceeds as illustrated in fig-13.

With reference to FIG. 8, we now discuss i! case in which the SHAPER-B, SCHEDULER-B and SHAPER-Q, SCHEDULER-Q blocks of fig. 9 are not to be considered optional, but rather effective. This is the most probable case as generally blocks and transmission queues with non-real-time traffic-type require a prior check of the peak band before their admission to the service. Please note that for the real-time correspondents the peak is guaranteed in advance when the connection is accepted. In the case presented, the main flow diagram of fig. 13, as well as the flow diagrams of the service phases illustrated in figures 14, 15 and 16 remain unchanged, since the service they produce is to issue a B-ID pointer and / or Q-ID on the BUS-DATA-WFQ data bus of the WFQ-CONTR control unit. The aforementioned unit, from the analysis of the TR-TYPE, will decide whether to extend the BUS-DATA-WFQ bus towards the SHAPER-B and SHAPER-Q blocks, or otherwise transfer the B-ID and / or Q-ID pointer directly to the SER-CONTR unit (fig. 10) which controls the terminal phase of the service function. In a completely analogous way, it is possible to selectively disable, via software, the control of the peak on the band of single blocks, and / or on the band of single transmission queues, which will continue to have a band regulated by the pure WFQ algorithm. The proposed structure therefore allows the coexistence of blocks and / or queues with or without the control of the peak cell-rate.

Fig. 17 shows the functional block diagram of a third embodiment variant in which the search for a B-ID indicator to be emitted is carried out in parallel by the SCHEDULER-B blocks and by the tandem WFQ-B, SHAPER-B, and the search for a Q-ID indicator to be emitted is carried out in parallel by the SCHEDULER-Q blocks and by the WFQ-Q, SHAPER-Q tandem. The resulting advantage is to reduce the CDV that the indicator undergoes by covering its entire ring.

With reference to fig. 17, it can be noted that the block representation shown there differs from that of fig. 9 for the following:

- the output of the SHAPER-Q block is no longer connected to the SCHEDULER-Q block but rather to a first input of a SELECT Q-ID block;

- the output of the SCHEDULER-Q block is no longer connected to the buffer manager BM but rather to a second input of the SELECT Q-ID block;

- a first output of the SELECT Q-ID block is connected to the buffer manager BM;

- a second output of the SELECT Q-ID block is connected to the input of the SCHEDULER-Q block.

Dual considerations apply to the WFQ-B, SHAPER-B, SCHEDULER-B blocks; in this case there is a SELECT B-ID block whose output connected to the buffer manager BM is also connected to a third input of the WFQ-Q block. For the sake of brevity, the indication of a respective control unit of the SHAPER-Q blocks has been omitted in the figures. SCHEDULER-Q, and SHAPER-B, SCHEDULER-B.

With reference to fig. 18, the operation of the statistical multiplexer shown in the block diagram of FIG. 17 regarding the shaping and scheduling functions only. For the sake of brevity, the flow chart of the aforementioned functions is indicated which concern only the B-ID indicators, since the flow chart that deals with the Q-ID indicators is completely identical to that which will now be illustrated. In fig. 18 after the START indication, the processing flow is divided into two branches performed in parallel. On the first branch, the B-ID indicator extracted from the WFQ-B block is sent, at the end of the RUN WFQ-B phase, to the SHAPER-B block. The latter calculates the insertion time in the calendar, expressed as N-SLOT-B, and in the following phase the control unit inserts B-IO in the calendar at the location indicated by N-SLOT-B, increases by a cell time Tc a TNOW current time counter used to scan the calendar, and then returns to the starting point immediately after the START. Looking at the block diagram in fig. 17, phase B-ID → CALEND (N-SLOT-B) is equivalent to the path taken by B-ID from the second output of the SELECT B-ID block to the input of the SCHEDULER-B block.

Simultaneously with the processing steps above, the control unit performs a test phase IS B-ID FROM SCHED-B to check for the presence of a B-ID indicator issued by SCHEDULER-B.

If the answer is yes, the B-ID pointer is emitted on the first output of the SELECT B-ID block from which it continues towards the buffer manager BM, and the control returns to the execution of the TNOW + Tc → TNOW phase and to the subsequent repetition of the elaborations on the two branches.

If, on the other hand, the answer is no, it means that the search algorithm governing SCHEDULER-B has not identified any B-ID indicator to be extracted, the control unit then performs a second test phase IS B-ID DA WFQ-B to check for the presence of a B-ID indicator issued by the WFQ-B block.

If the answer is no, no B-ID indicator has been issued by WFQ-B. In this case, no B-ID indicator (and therefore Q-ID) can be emitted from the ring in the present cell time and no further operation must be carried out, therefore the control returns to the execution of the TNOW + Tc → TNOW phase and to the following one. repetition of the processing on the two branches.

If in the second test phase the answer is yes, it means that a B-ID indicator has been issued by the WFQ-B block and it is necessary to establish if, and when, it can be re-issued towards the BM block. For this purpose, a third test phase is carried out È N-SLOT-B> TNOW / Tc in which the scheduler control unit checks whether the insertion location in the calendar calculated by the SHAPER-B block is greater than that (TNOW / Tc) corresponding to the current time TNOW.

If so, no B-ID indicator can be extracted from the ring in the present time cell TNOW, and the control passes to a subsequent phase INSERT B-ID IN THE CALENDAR TO LOCATION N-SLOT-B, the meaning of which is obvious; then the control returns to the execution of the TNOW + Tc → TNOW phase and to the subsequent repetition of the processing on the two branches.

If in the third test phase the answer is negative, it means that the N-SLOT-B location is equal to that calculated by the shaper at the current time TNOW, the minor sign would in fact represent a condition made impossible by the particular algorithm adopted by the shaper. In this case an EMIT B-ID phase is performed in which the B-ID indicator is emitted on the first output of the SELECT B-ID block, from where it continues towards the buffer manager BM. No entry of B-ID in the calendar is necessary. Then the control returns to the execution of the TNOW + Tc → TNOW phase and to the subsequent repetition of the processing on the two branches.

Advantageously, a single integrated circuit of the ASIC type (Application Specific Integrated Circuit) has been realized which includes all the functional blocks indicated in fig. 9.

Architecture with reduced functionality can be obtained by removing from the architecture of fig.9 some blocks that implement the functions not considered. In this case it will be necessary to modify the flowcharts illustrated so far in a congruent way. For example, in the multiplexer illustrated in the third variant of fig.17 it is possible to include the parallel processing offered by the SELECT block in one or the other, or in both, the emission rings of the B-ID and Q-ID pointers.

Another example of such structures can be seen in fig.19 relating to a fourth embodiment variant of the present invention in which the SHAPER-Q and SCHEDULER-Q blocks have been removed. In this case, the queue indicator Q-ID extracted from the WFQ-Q block is not inserted in the calendar but passed directly to the buffer manager BM, which must request the reading of an ATM cell stored in the BUF-CEL buffer (fig. 12) and, if necessary, transfer the Q-ID indicator back to the WFQ-Q functionality. The solution indicated in fig 19, which does not allow to control the peak band of each transmission queue, nevertheless guarantees a better use of the band and less use of logic.

A further example of a structure with reduced functionality is visible in fig. 20, relating to a fifth embodiment variant of the present invention in which the WFQ-B block has been removed. In this case the BM block transfers the B-ID indicators directly to the SHAPER-B block which, having performed its function, inserts it in the calendar scanned by the SCHEDULER-B block. This choice is functionally equivalent to the way of operating according to the known art, as it assigns a constant band to the blocks and reserves the application of the WFQ functionality only to the transmission queues.

Claims (27)

CLAIMS 1. Method of statistical multiplexing of ATM cell flows originating from connections having service classes of different quality, said flows generating transmission queues of cells (QUj) divided into functional blocks (Bi, ..., Bm), and to at least some of said flows being assigned a minimum guaranteed band on the multiplexed flow (LINK-OUT) and the possibility of distributing a residual band that may be available on the same flow by means of a function, later called Weighted Fair Queuing, or WFQ, which dynamically assigns appropriate quantities of tokens, evaluated on a statistical basis, to individual transmission queues (QUj), which spend said tokens to access the residual band on! multiplexed flow (LINK-OUT), characterized by the fact that said WFQ functionality includes phases of dynamic assignment of appropriate quantities of tokens, evaluated on a statistical basis, also to said functional blocks (Bi, ..., Bm), which spend said tokens to allow their own transmission queues (QUj) to access the distribution of the residual band on the multiplexed flow (LINK-OUT), thus dividing the residual band on the multiplexed flow between said functional blocks (Bj). 2. Statistical multiplexing method according to claim 1, characterized in that the steps of said WFQ functionality are repeated cyclically with cycle renewal period (TR) which is an integer multiple of the transmission time of a cell (Tc) on said multiplexed flow (LINK-OUT), each cycle including the phases of: a) assign (INITIALIZE TK-CNT1, ..., m) to at least some of said transmission queues (QU1, .... QUz) a respective number of tokens (N ^ KfQUI), ..., N ° TK ( QUz)) proportional to the band of the respective said flows; b) assign (INITIALIZE TK-CNT-B) to at least some of said blocks (B1, .... Bm) a respective number of tokens, (N ° TK (B1) N ° TK (Bm)) proportional to the cumulative band of the respective component flows; c) distribute (INITIALIZE FIFO WFQ-B) the indicators (B-ID) of blocks that include transmission queues waiting to be served within a high priority list (HP-FIFO), a low priority list (LP-FIFO) , and a very low priority list (VLP-FIFO) belonging to a first set of lists (WFQ-B) managed on a FIFO basis; d) distribute (INITIALIZE FIFO WFQ-Q) the indicators (Q-ID) of said transmission queues that are waiting to be served within a high priority list (HP-FIFO), a low priority list (LP-FIFO), and a very low priority list (VLP-FIFO), individually belonging to second sets of lists (WFQ-Q) managed on a FIFO basis, associated with the functional blocks; e) scan, at cell frequency (Tc) and on the basis of the available tokens, the lists of indicators (B-ID) of said first set (SERVI HP-FIFO, SERVI LP-FIFO, SERVI VLP-FIFO) for emission of indicators (B-ID) which select functional blocks (Bj) having queues to be served; f) scanning, at cell rate (Tc) and on the basis of the available tokens, the lists belonging to said second sets of lists (WFQ-Q) selected, for the emission of respective indicators (Q-ID) of transmission queues ( QUj) enabled to transmit a cell on the multiplexed stream (LINK-OUT). 3. Statistical multiplexing method according to claim 2, characterized in that said step of scanning e) comprises the steps of; - read (READ HP-FIFO) an indicator (B-ID) from a high priority list (HP-FIFO) and serve (SERVI HP-FIFO) the indicated block, ending the scan (IS HP-FIFO EMPTY) with the emptying of said high priority list; - read (READ LP-FIFO) an indicator (B-ID) from a low priority list (LP-FIFO) and serve (SERVI LP-FIFO) the indicated block, decreasing, if necessary, the available tokens and transferring the 'indicator (B-ID) of the block that has exhausted its tokens within a very low priority list (VLP-FIFO), ending the scan (IS LP-FIFO EMPTY) with the emptying of said low priority list, o with the end of the renewal period (TR); - read (READ VLP-FIFO) an indicator (B-ID) from a very low priority list (VLP-FIFO) and serve (SERVI VLP-FIFO) the indicated block, ending the scan (IS VLP-FIFO EMPTY) with the emptying of said very low priority list, or with the end of the renewal period (TR). 4. Method of statistical multiplexing according to claim 3, characterized in that each said phase of serving (SERVI HP-FIFO, SERVI LP-FIFO, SERVI VLP-FIFO) the blocks indicated in said lists having different priorities, furthermore comprises the phase re-inserting said block indicator (B-ID) in the source list in the event of the presence of at least one cell still to be transmitted in at least one queue (QUj) assigned to the selected block. 5. Statistical multiplexing method according to one of the preceding claims, characterized in that said connections having service classes of different quality include some for which the peak cell-rate is guaranteed, and which therefore involve flows which do not have the opportunity to make use of said further band possibly available on the multiplexed flow (LINK-OUT), said flows at the peak cell-rate generating transmission queues allocated in a single block which are served as a priority until it is emptied. 6. Method of statistical multiplexing according to one of the preceding claims, where some of said connections are supported by flows of ATM cells which require a control of the peak band such as not to allow a certain maximum bandwidth value to be exceeded within said flow multiplexed (LINK-OUT), characterized by the fact of resorting to a first timing that expands the emission intervals of at least some said indicators (B-ID) of the functional blocks (Bj) to limit the cumulative peak band of the group of connections to the selected blocks. 7. Statistical multiplexing method according to claim 6, characterized in that said first timing is obtained by calculating at each cell time (Tc) an insertion location (N-SLOT-B) in a first calendar of an indicator (B- ID) of a selected block, the first calendar being read with the same writing rate to extract an indicator of the block (Bj) which will actually be selected at the current cell time (TNOW). 8. Method of statistical multiplexing according to claim 6, or 7, characterized by the fact of resorting to a second timing which dilates the emission intervals of at least some of said transmission queue indicators (Q-ID) to limit the peak band of the queues served. 9. Statistical multiplexing method according to claim 8, characterized in that said second timing is obtained by calculating at each cell time (Tc) an insertion location (N-SLOT-Q) in a second calendar of said queue indicator to be served (Q-ID), the second calendar being read with the same writing rate to extract an indicator of the transmission queue (QUj) that will actually be served at the current cell time (TNOW). . 10. Statistical multiplexing method according to claim 7, characterized in that said insertion in a first calendar of a block indicator (B-ID) is carried out in parallel with said extraction. Method of statistical multiplexing according to claim 9 or 10, characterized in that said insertion in a second calendar of a transmission queue indicator (Q-ID) is carried out in parallel with said extraction. 12. Statistical multiplexing method according to one of claims 7 to 11, characterized in that said indicators (B-ID, Q-ID) extracted from a common location (N-SLOT-B, N-SLOT-Q) of a respective calendar, are extracted according to the FIFO methodology. 13. Statistical multiplexer of ATM cell flows originating from connections having service classes of different quality, said flows generating cell transmission queues (QUj) divided into functional blocks selected for the service, and at least some of said flows being assigned a band guaranteed minimum on the multiplexed flow (LINK-OUT), and the possibility of distributing a residual band that may be available on the same flow, including: a) a buffer for storing said transmission queues (QUj) of ATM cells: b) means that control access to the buffer for the insertion or extraction of said cells; c) means for counting the number of cells included in each said transmission queue; d) means for carrying out a known technique, hereinafter called Weighted Fair Queuing, or WFQ, which dynamically assigns appropriate quantities of tokens, evaluated on a statistical basis, to individual transmission queues (QUj), which spend said tokens to access the residual band on the multiplexed flow (LINK-OUT), characterized by the fact that said means of execution of the WFQ technique are modified (WFQ-CONTR, WFQ-B, WFQ-Q) so as to dynamically assign appropriate quantities of tokens, evaluated on a statistics, also to said functional blocks (Bi. Bm), which spend said tokens to allow their own transmission queues (QUj) to access the distribution of the residual band on the multiplexed stream (LINK-OUT), thus dividing between said blocks functional (Bj) the residual band on the multiplexed stream. 14. Statistical multiplexer according to claim 13, characterized in that said modified means for carrying out the WFQ technique (WFQ-CONTR, WFQ-B, WFQ-Q) comprise: a) a first storage structure (TK-CNT1, ..., TK-CNTm) available to each transmission queue (QUj) to store a respective number of tokens (N ° TK (QU1). N ° TK (QUz) ) proportional to the bandwidth of the respective said flows, said first storage structure (TK-CNT1, ..., TK-CNTm) being usable as a set of counters; b) a second storage structure (TK-CNT-B) available to each of said functional blocks (B1, ..., Bm) to store a respective number of said tokens (N ° TK (B1), ... , N ° TK (Bm)) proportional to the cumulative band of the respective component flows, said second storage structure (TK-CNT-B) being usable as a set of counters; c) a first set of FIFO memories having different service priorities (WFQ-B), usable for storing indicators (B-ID) of the functional blocks (Bi, ..., Bm); d) second sets of FIFO memories having different service priorities (WFQ-Q), individually usable for storing indicators (Q-ID) of the transmission queues (QUj) belonging to a respective said functional block (Bj); e) a control unit (WFQ-CONTR) connected to the first (TK-CNT1, ..., TK-CNTm) and second (TK-CNT-B) storage structure reserved for tokens, to said first (WFQ-B ) and second sets of FIFO memories (WFQ-Q) reserved for the indicators, said control unit (WFQ-CONTR) being governed by a program that assigns the respective said quantities of tokens and regulates their consumption. 15. Statistical multiplexer according to claim 14, characterized in that said first set of FIFO memories (WFQ-B) and each second set of FIFO memories (WFQ-Q) having different service priorities, individually comprise a high priority FIFO memory (HP-FIFO), a low priority FIFO memory (LP-FIFO), and a very low priority FIFO memory (VLP-FIFO). 16. Statistical multiplexer according to one of claims 13 to 15, characterized in that said modified means of carrying out the WFQ technique (WFQ-CONTR, WFQ-B, WFQ-Q) further comprise a real-time counter which is initialized by said control unit (WFQ-CONTR) with a predetermined value, decreased at each cell-time (Tc), and then reinitialized after each reset, by cyclically repeating the aforementioned steps; the real time counter being used by the control unit (WFQ-CONTR) to renew, at each initialization of the aforementioned counter, the writing of the initial number of said tokens reserved for the functional blocks (Bi, ..., Bm) and for the transmission queues (QUj), respectively within the first (TK-CNT1, ..., TK-CNTm) and second (TK-CNTB) storage and counting structure. 17. Statistical multiplexer according to claim 16, characterized in that at each initialization of said real time counter the said control unit (WFQ-CONTR) renews the content of said high (HP-FIFO), low (LP -FIFO), and very low (VLP-FIFO) priorities belonging to said first (WFQ-B) and second sets (WFQ-Q) of FIFO memories. 18. Statistical multiplexer according to claim 17, characterized in that during each said renewal period (Tr) the said unit. control (WFQ-CONTR) scans, with cell frequency (Tc) and on the basis of the available tokens, the said first set of FIFO memories (WFQ-B) for the emission of an indicator (B-ID) for the selection of a functional block (Bj) having at least one queue to be served. 19. Statistical multiplexer according to claim 18, characterized by the fact that during each said renewal period (Tr) the said control unit (WFQ-CONTR) scans, at a rate of clock (Tc) and on the basis of the available tokens, a said second set of FIFO memories (WFQ-Q) associated with a selected functional block (Bj), for the emission of an indicator (Q-ID) of a transmission queue (QUj) enabled to transmit a cell on the multiplexed flow (LINK- OUT). 20. Statistical multiplexer according to one of claims 13 to 19, where said connections having service classes of different quality include some for which the peak cell-rate is guaranteed, and which therefore involve flows which cannot be used of said further band possibly available on said multiplexed flow (LINK-OUT), said flows at the peak rate generating transmission queues allocated in a single block, characterized by the fact that said control unit (WFQ-CONTR) is further connected to a device (RT-FLAG) indicator of the filling status of said single block in such a way that the transmission queues of said single block are primarily served as long as said indicator device (RT-FLAG) is active. 21. Statistical multiplexer according to one of claims 13 to 20, where some of said connections are supported by flows of ATM cells which require a control of the peak band such as not to allow a certain maximum band value to be exceeded within said multiplexed flow (LINK-OUT), characterized in that it comprises first timing means (SHAPER-B, SCHEDULER-B) which dilate the emission intervals of at least some said indicators (B-ID) of the functional blocks (Bj) to limit the cumulative peak band of the group of connections belonging to the selected blocks. 22. Statistical multiplexer according to claim 21, characterized in that said first timing means (SHAPER-B, SCHEDULER-B) comprise a first traffic shaper (SHAPER-B) placed upstream of a first scheduler with calendar mechanism ( SCHEDULER-B), called first shaper (SHAPER-B) calculating at each cell time (Tc) an insertion location (N-SLOT-B) in the first calendar of an indicator (B-ID) of a selected block and writing the indicator (B-ID) in the first calendar at the calculated location, and said first scheduler (SCHEDULER-B) reading the first calendar with the same writing frequency to extract an indicator (B-ID) of the blocks (Bj) that will be actually served at the current cell time (TNOW). 23. Statistical multiplexer according to claim 21, or 22, characterized in that it also comprises second timing means (SHAPER-Q, SCHEDULER-Q) which dilate the emission intervals of at least some of said indicators (Q-ID) of the transmission queues (QUj) to limit the peak band of the queues served. 24. Statistical multiplexer according to claim 23, characterized in that said second timing means (SHAPER-Q, SCHEDULER-Q) comprise a second traffic shaper (SHAPER-Q) placed upstream of a second scheduler with calendar mechanism ( SCHEDULER-Q), called second shaper (SHAPER-Q) calculating at each cell time (Tc) an insertion location (N-SLOT-Q) in the second calendar of a transmission queue indicator (Q-ID) to be served and writing the indicator (Q-ID) in the second calendar at the calculated location, and called the second scheduler (SCHEDULER-Q). reading the second calendar with the same writing frequency to extract an indicator (Q-ID) of the transmission queue (QUj) that will be effectively served at the current cell time (TNOW). 25. Statistical multiplexer according to claim 22, characterized in that said first shaper (SHAPER-B) operates in parallel to said first scheduler (SCHEDULER-B). 26. Statistical multiplexer according to claim 24, or 25, characterized in that said second shaper (SHAPER-Q) operates in parallel to said second scheduler (SCHEDULER-Q). 27. Statistical multiplexer according to one of claims 22 to 26, characterized in that said indicators (B-ID, Q-ID) extracted from a common location (N-SLOT-B, N-SLOT-Q) of a respective calendar , are extracted according to the FIFO methodology.