GB2299241A - Configuration of digital switch - Google Patents

Description

Method for connecting path protection signals in digital switch The invention relates to a method according to the preamble of claim 1 for the configuration of the switching of digital signals connected to a digital time-spacetime (TST) switch in situations requiring a change in the switching, and for the path protection of such signals in a TST switch. The invention also relates to a circuit arrangement realizing the method.

Synchronous digital hierarchy (SDH) comprises a large entity for transmitting timedivision signals in a telecommunication network the trunk network of which is developing into a remote-controlled switching network. The first level of SDH signals is the synchronous transport module (STM-I) the transmission rate of which is 155.520 Mbit/s. The basic STM-1 frame consists of bytes (8-bit), which there are 2430, including control blocks. Thus the STM- 1 frame transnits 63 TU 12 (tributary unit) 2-Mbit/s signals, which can contain a 2-Mbit/s signal of a usual 30-channel PCM system. Each byte in the frame constitutes a 64-kbit/s channel. SDH signals, or transport modules, are formed from the subsystem signals by means of byte interlacing.

A digital switch (DXC) in the SDH can switch traffic between the various SDH levels and between different signals. In addition, it has to be capable of flexibly reconfiguring the network, ie. rerouting the connections, and to guarantee a quick switch-over to backup connections in network fault situations.

Digital switching has been extensively studied in order to find an optimal architecture. A structure which is nonblocking and meets the criteria concerning capacity and feasibility is the TST structure, or time-space-time switching, as is disclosed e.g. in our patent PCT/FV00174 (or corresponding F1-92 1834). The patent discloses the general principles of a TST switch in considerable detail. Although the method disclosed in patent PCT/FI/00174 functions quite well, there is, especially in bigger switches, need for an even more efficient switch control method.

The object of this invention is to provide a method and an algorithm according to that method to calculate a configuration of a TST switching structure both for pointto-point connections and for subnetwork back-up connections.

The invention proposes a method according to claim 1, in which the problem is solved by means of restructuring it, in practice with 63 connection matrices, and the method concentrates on solving only the connections of the first TS stage and in addition, only those signals that need rearranging.

Other advantageous embodiments of the invention are described in the subclaims.

Advantageous solutions for the circuit arrangement realizing the method are described in claims 9 to 11.

The signals to be connected are advantageously multiplexed subsignals of highcapacity signals, which in the SDH system means that the subsignals are mainly 2-Mbit/s VC-12 virtual containers, the main signals then being 1 55-Mbit/s STM- 1 signals.

The strength of the algorithm lies in fact in that it only rearranges connections needing rearrangement. This is partly realized by setting up the switching problem in a novel manner. The switching problem is presented as an imaginary matrix. The algorithm records all rearrangements performed to the connections to realize the switching. When all rearrangements have been completed the records provide all necessary information needed to realize the switching according to the switch control information.

In the method according to the invention, difficult switching situations are solved in the recursion stage, wherein the location of a connection in the imaginary matrix is randomly forced to a new location, whereafter operation is continued according to the basic method.

The connection field of the switch requires a calculation algorithm to configure the connections. Point-to-point connections can be configured using known algorithms.

But the PPTST algorithm according to the invention goes a step further and provides a solution also for the path protection (PP). The PPTST algorithm calculates, when necessary, also the point-to-doublepoint connections (broadcast).

Tests have shown that the method according to the invention is faster than or at least as fast as known calculation methods, also when the signals to be connected include path protection signals.

The algorithm according to the invention can also be used to route point-to-point signals through a three-stage Benes switching network.

The invention is described by means of examples and referring to the attached drawing.

Fig. 1 shows a general depiction of an imaginar) matrix used as a starting point in the algorithm according to the invention.

Fig. 2 illustrates the connection of path protection signals through a space switch.

Fig. 3 shows a general depiction of an input table.

Fig. 4 shows the algorithm data structure for a selected time slot (time = T) with the auxiliary tables of the algorithm shown around the imaginary connection matrix.

Fig. 5 is a simplified flow chart of the algorithm used in the method according to the invention.

Fig. 6 shows the flow chart for the first processing stage for path protection signals.

Figs. 7a and 7b illustrate the processing of ordinary point-to-point signals in the FindEven process.

Fig. 8 shows the flow chart of the FindEven process.

Fig. 9 illustrates the BestSwap process used in the pull-push process by means of connection matrices.

Fig. 10 shows the flow chart of the pull-push process.

Fig. 11 illustrates the HallSwap case in the pull-push process.

Fig. 12 illustrates the FillerSwap case in the pull-push process.

Fig. 13 illustrates the BestReplace case in the pull-push process.

Fig. 14 illustrates the ReplaceSwap case in the pull-push process.

Fig. 15 illustrates the ReleaseSwap case in the pull-push process.

Fig. 16 illustrates the PPSwap case in the pull-push process.

Fig. 17 shows a simplified flow chart of the PPSwap process.

Next we will discuss an implementation example concerning the processing of VC12 signals. Note, however, that the method according to the invention can be used to process higher-capacity signals which are cut into multiple VC-12 signals connected to a TST switch. The last stage of the TST switch is then used to solve the reconnection of the higher-capacity signal.

The invention defines for a TST switching field a calculation algorithm to define path protection (pp) signals in a digital switch DXC. The calculation algorithm uses a timer function to end execution if no solution is found in the calculation. Below, the following abbreviations will be used: PP means the same as SNCP, or subnetwork connection protection, TST means time-space-time switching architecture, and PPTST means path-protected TST connection. DXC means a digital switch.

In this example, the PPTST algorithm handles only VC-12 signals (VC = virtual container), because VC-4 signals and VC-4 protection algorithms do not use time switches and, therefore, are not problematic from the blocking standpoint. However, the TST structure requires a special treatment of VC-12 signals in the SNCP case, which is also called path protection PP in this document. The PPTST algorithm is used to solve the difficulties experienced earlier in the configuration of a switch.

Generally speaking, the algorithm is restricted, in addition to the TST connection, by time factors as well. The configuration has to be calculated by a real processor, which with the current technology is slow, and yet the requirements on connection setup times by network operators have to be met. The task of the timer function is to interrupt the calculation if time limits are exceeded.

The algorithm needs a transform layer which can also be applied to the general DXC-TST algorithm and the Game algorithm. The function of the transform layer is to handle configuration requests and ASIC control requirements. The transform layer also filters out unnecessary requests, such as VC-4 connections as no algorithm is needed in their configuration. The other requests constitute the TST connection configuration problem of VC-12 signals, which has to be solved using the PPTST algorithm.

In principle, the problem to be solved by the algorithm is to find 63 configurations for the space (S) and time (T) switches through a TST connection field. These 63 solutions correspond to the number of VC-12 signals in one STM- 1 frame. The PPTST algorithm aims at solving the main TST restriction. The connection through the space switch S has to be quite distinct, ie. a signal has to be directed through ports not used by other signals. The easiest way of presenting this S switch problem is to use 63 matrices of the size 16*16. An input port corresponds to a column and an output port to a row in a matrix Then the structure of the algorithm can be described simply as follows.

The signals are presented by means of 63 matrices. in which the time slot distribution corresponds to the original. The algorithm is used to rearrange the contents of these 63 matrices so that each signal uses the ports in a distinct manner. Since the ports mean rows and columns, the restriction can be formulated as follows: a signal must use a distinct row and column.

The space connection of a particular time slot is nonblocking when all the outputs of the space switch are in use. In other words, the input side time switch has to connect all time slots, or channels, in a way such that all time-switched time slots of the inputs of the space switch are directed to different outputs. If an input has no time slot that should be routed to a certain output, then, of course, that output need not be used. Thus, the function of the input side time switch before the space switch is to evenly divide the channels in the time slots so that the space switch can connect them to the appropriate, desired outputs. To be more specific, there should not occur in the space switch a congestion, in which a time slot contains more than one channel routed to a particular output time slot.Using the method according to the invention the connection matrix is set up such that switching is performed in a nonblocking manner.

Fig. 1 shows the basic structure of the PPTST algorithm: a table with 16 rows corresponding to the outputs, ie. outputl to outputl6, and 16*63 columns corresponding to the time slots, ie. (inputl to inputl6)*(timel to time63). This structure is however complemented with other structures to speed up the execution of the algorithm. in practice, each connection can be thought of as a crosspoint (x) in the matrix. The task of the algorithm is then to see to that there is only one crosspoint (x) on each row and in each column for a normal point-to-point connection, ie.

a connection from one input to one output.

The matrix in Fig. 1 is used for rearranging two types of signals: point-to-point signals and path protection, or pp signals.

A point-to-point signal has to be connected between one input port and one output port, which means it occupies a crosspoint of one column and one row. (In a bidirectional connection, the return direction naturally occupies another crosspoint.

For the sake of simplicity this is not discussed separately in this document, because both directions can be solved using the same principle and, in practice, as mirror images.) A pp signal in its entirety is a bidirectional connection which is protected.

Therefore, a pp signal uses three input ports and three output ports, ie. three crosspoints. Fig. 2 shows the structure of the pp signal. Here we will only briefly note that the pp selection signals appear simultaneously in the ports of the space switch so that the protection switching can be performed quickly without special calcu latin, ie. only by updating the switch control data. The selection and broadcasting functions are combined in one pp block so that the connection capacity required is minimized.

Pp signals usually utilize symmetric signal ports since a bidirectional connection uses same ports both in the transmit distribution stage (1 o 2) and in the receive selection stage (2 o 1). Using the markings in Fig. 2, this can be expressed for the space switch S as a condition: Xin = Xout; Yin = Yout; and Zin = Zout. In the case of a unidirectional protected signal only one side of the pp signal is used, ie. the switch realizes only point-to-doublepoint connections. Two unidirectional protected signals can be combined to make one complete pp signal if the ports used meet the above-mentioned condition Xin = Xout; Yin = Yout; and Zin = Zout.

The algorithm according to the invention is based on the basic structure shown in Fig. 1. The algorithm manipulates signals coming to the input ports in the T switch so that the signals are rearranged. The input ports are presented as columns and the calculation method builds 63 matrices which conform to the TST restriction mentioned above. The algorithm can be divided into three main tasks: 1. solution of path protection, or pp signals; 2. solution of easy normal point-to-point signals (find even solutions); 3. solution of difficult normal point-to-point signals using pull-push routine.

in the first task, all pp signals are handled. The pp signal parts are collected in a same time matrix and the individual pp signals are located in the 63 time matrices.

In this task, normal point-to-point signals are ignored.

in the second task, a major part of the normal connections are solved. Most of the normal connections can be located directly into a time slot matrix in which the ports in question are not already reserved. This is best depicted so that even swaps are performed in the input ports between the two signals in question in a manner such that both signals are located in the most probable final time slot in the configuration solution. This can be called "finding even solutions".

The third task of the algorithm concentrates on difficult normal connections. These signals do not have any obvious time slots in the input ports that have a free output port, ie. the output of the S switch is already reserved. Another connection has then to be made in the desired output port, and that requires rearrangement of already solved normal input signals to find a free output port for the difficult connection.

This reordering applies a so-called pull-push routine according to Hall's statement wherein overlapping connections are moved from a time matrix to another until a free output port is found. This task handles only normal connections, because incorporation of pp signals in the pull-push routine would make the task very difficult to converge, and it could result in an avalanche-like increase in the number of moves. For each move, a pp signal involves not only one port, but it can hit several connections. In addition, the avalanche can spread and result in the need for moving both normal and pp connections so that more overlapping, or unsolved cases, follows since in the pull-push routine only primary overlapping is relevant.

Such a process would also consume much time and the time factor actually prohibits the use of a more complex algorithm in the pp pull-push routine. Therefore, pp signals are included in this pull-push process only when no other solutions are found.

By keeping a record of the various pull-push actions (swaps) in the basic structure shown in Fig. 1 the target configuration is easier to indicate. The first time switch stage uses that record for the final reordering of signals. The target port information of the final order is then used to control the space switch. The last time switch stage uses the target time slot information to set the requested final signal distribution.

The crosspoints in the basic data structure containing 63 time matrices, shown in Fig. 1, are illustrated with the input table example in Fig. 3. The size of the input table is 16*63 and originally it contains all input signals, which are distributed according to the time of arrival. The targets of the virtual containers are presented by means of port and time slot numbers. The input table defines the basic structure as it forms the x-axis of the basic structure. This is accomplished through lining up the input table next to the structure, whereby the input table is placed according to the time on the x-axis as an input row in the basic structure. The contents of the input row then indicate the target port. The connection can be thought of as a crosspoint of a column and a row in the basic data structure.The configuration is calculated manipulating the input rows. The manipulation does not erase table inputs but only rearranges the positions of the values.

Fig. 3 shows example signals. Input 4 contains an incoming signal of a normal point-to-point connection arriving in time slot 4 and targeted to time slot 41 in output port 3. In the figure, the crosspoint is marked 3,41. The pp signal is in three time slots. In Fig. 3, the broadcast signal (1 o 2) is in time slot 11 of input row 1, and its targets are outputs 5 and 8, time slots 11 and 61, respectively; the crosspoint is marked 5/8,11/61. The selection signal parts are on input rows 5 and 8, time slots 14 and 6. As shown in Fig. 3, the selection signals are targeted to output port 1 and time slot 32 (1,32).

The central data structure of the algorithm is the input table according to Fig. 3. It is the same as the basic data structure shown in Fig. 1. The basic data structure as such is not physically necessary and it can be thought of as an imaginary data structure.

However, fast execution of the algorithm requires a control data field as large as the basic data structure in Fig. 1 but with another meaning, which will be discussed in conjunction with the OutputQueue. Next, we will discuss signal representation and structures related to the input table. Generally, the port value in this example is 0 to 16 and the time value 0 to 63, the zero value indicating that no signal is present, and the numbers from 1 to 16 representing the ports. Note, however, that the port value is only a logical value and as such it has nothing to do with the hardware location, e.g. the position of printed boards.

A unidirectional point-to-point signal is represented in the input table by two numbers, as was mentioned, and shown in the form "porf time slot". A pp signal, however, requires a two-part representation in the input table. The first part concerns incoming selection signals and the other part the outgoing broadcast signal. The selection signals are identical as they have the same target, ie. the target output port and time slot. The broadcast part of a pp signal has two targets and is therefore represented by two port numbers and two time slot numbers as shown in Fig. 3. In the input table, the representation also requires a recognition field PPNumber, which quickly indicates the difference between the point-to-point signal and pp signal.This data field contains a number: when it is other than zero, from 1 to 315, it indicates a pp signal, and zero indicates a normal signal. Numbering the pp signals also makes it easier to handle the pp signals in switching. For selection signals, the marking is "pp number, port, time slot", and for broadcast signals, "pp number, active port, active time slot, backup point, backup time slot".

The transform layer has to provide the receive interface with an input structure which corresponds to the input table and contains all the signal connection requests.

Furthermore, a table is needed in which all pp signals are in numerical order. The format of a pp queue element PP is: "pp number, signal port signal input time slot, signal output time slot, active port, active input time slot, active output time slot, backup port, backup input time slot, backup output time slot".

An imaginary matrix is not necessary for the algorithm, and the algorithm does not use the matrix, but the underlying principle of the algorithm is better illustrated with a large matrix.

Let us next consider the structure of the algorithm, illustrated in Fig. 4. It shows an imaginary connection matrix, which in this case represents connections directed via the space switch (S) in time slot T. When examining Fig. 4. please note that the square matrix in the middle is entirely imaginary, and the algorithm according to the invention requires only the separately named structures.

The pp collection routine uses a book-keeping table PPBooking, which is of the same size as the input table (InputTable). The elements in the book-keeping table are numbers that correspond to the pp numbers. Compared to the basic structure, which has both input and output axes, the book-keeping table can be onedimensional, for the pp signals utilize the ports in a symmetrical manner. The three input ports in question have the same numbers as the output ports. The bookkeeping table is shown in Fig. 4, its size is 16*63, and its elements have values ranging from 0 to 315.

Fig. 4 also shows an output reference table (OutputReference) of the size 16*63. Its elements indicate how many signals are targeted to an output port in a time slot.

The output queue table (OutputQueue) is of the size ( 16*16)*63. Its elements indicate the input ports of the signals targeted to an output port in a time slot. When the signal is a pp selection signal, the element contains the numbers of two input ports. The output queue table also contains the number of the pp signal (PPNumber). The element format is "pp number, input portl, input port2".

The overlap variable (Overlap) indicates the number of overlapping connections, or how many conflicting requests for an output port there are. This value indicates the total sum of overlaps for all time slots, even though Fig. 4 otherwise shows only one time slot (Time = T) for reasons of clarity.

A path protection pointer TimePP indicates whether a particular time slot contains pp signals. Yes-value indicates a pp signal and no-value indicates that the time slot contains no pp signals. The TimePP values are collected in a TimePPQueue table the length of which is 63. The SwapRecord table contains numbers which indicate the original time slot location of the signal in the corresponding input table. The size of the SwapRecord is 16*63 and the values of its elements range between 0 and 63.

Let us consider the situation of time slot T, shown in Fig. 4. There is a normal connection between input port 3 and output port 4, the pp number of which is 0 both in the input table and in the output table, and the structures contain references to each other. This connection can be seen as a crosspoint (x) between the input table and output queue in the imaginary matrix. The OutputReference value 1 of output port 4 indicates that only one connection is using this output port, ie. no overlapping occurs at this output and thus there is only one element in the output queue.

The only pp signal of the InputTable is easy to detect on the basis of the pp number: at input ports 2, 7 and 9 of time slot T there is a pp signal with the ordinal number 4. The broadcast part at input port 2 is targeted to output ports 7 and 9, as indicated by the Port value (7/9) in the table. The target port for the selection signal is output port 2. However, the selection signals do not increment the OutputReference value because only one of these signals is connected through.

The OutputReference indicates overlapping for output port 5. The output queue contains two elements which indicate the two input signals in question, ie. input ports 4 and 14. In this time slot T, there are two more overlaps. In all 63 time slots, there are altogether 24 overlaps, as indicated by variable Overlap in Fig. 4, ie. there are 21 more overlaps in time slots other than time slot T in Fig. 4.

The book-keeping table PPBooking indicates that pp signal number 4 is collected and placed in time slot T. The zeros in the table indicate locations for other possible pp signals. The SwapRecord indicates the time slots wherefrom the signals in question have been moved. Thus, the signal at input port 1 comes originally from time slot T and the pp broadcast signal at input port 2 comes originally from time slot 1, etc.

The result, or output, of the algorithm is now the rearranged InputTable and the rearrangement table SwapRecord. In other words, the SwapRecord indicates how each signal has to be moved by means of the first or input side time switch (T) in order to be connected through the space stage (S). The input table indicates the output port of the space switch to which the signal has to be connected. The input table also indicates the final output time slot. The contents of these tables are used as control data for the transform layer. In addition, the algorithm should return an indication about success or failure of the calculation to the transform layer.

All three procedures of the algorithm attempt to solve the configuration with the least possible execution, and they cover most of the possible configuration types.

Each part of the algorithm only processes data that need rearrangement. Data collection and the three main parts of the algorithm are depicted in Fig. 5.

The algorithm begins at block 51, whereafter path protection, or pp, signals are collected in block 52. Decision block 53 examines whether all pp routings are solved. After the pp solution, block 54 forms the basic data for rearranging overlapping connections. Then follows block 55, in which even solutions of easy normal connections are calculated. Then, difficult overlapping connections are solved in block 56 using pull-push moves.

Decision block 57 examines whether all solutions have been found, and the algorithm ends at block 59. If decision blocks 53 and 57 find that solutions are not found, the process jumps to block 58 which sends an error message to the transform layer, or the requester of the PPTST algorithm execution.

Collection 52 of path protection signals, shown in Fig. 5, is depicted in more detail in Fig. 6. At first, the pp signals are randomly distributed over the InputTable. The purpose of the PPCollection is now to collect the pp signal parts, or signals, in one and the same time slot. The location process is realized by first filling the time slots 1 to 63 that contain no pp signals. Then, after the start block 60, all the signals marked in the PPQueue are processed in the main loop of Fig. 6, denoted by the looping arrow B.

In block 62, the first pp signal pp number 1 is selected from the pp queue. In block 63, the time slot is selected. In decision block 64, it is checked from the PPBooking table whether the selected time slot is free. If not, execution jumps to loop A, in which decision block 67 checks whether all 63 time slots have been processed, and if not, block 68 selects the next time slot. When a free time slot is found in block 64, the execution moves on to block 65, in which the pp signal is moved to a free time slot and a corresponding marking is made on the book-keeping table PPBooking. Decision block 66 checks whether all pp signals have been processed.

If there are signals left, execution jumps to block 79 in which the next pp signal in the pp queue, pp number+l, is selected, and execution continues in loop B. When all pp signals have been processed, the process ends at block 61.

In the process described above, only pp signals compete for space in the selected time slot, for at this stage other normal signals are not processed. When free space has been found for the whole pp signal, the corresponding maricings will be made in block 65 and at the same time the pp signal parts are moved to the target time slot in the input table. At the same time, possible original signals in the target time slot are moved in the locations of the pp parts.

In loops A and B there may be cases in which free space is not directly found in the 63 time slots available. The execution then moves on to a permutation process, which is denoted by the looping arrow C in Fig. 6. In this first-degree permutation, the execution first moves from block 67 to block 69, and a new time slot is selected.

Block 70 then checks whether moving an already arranged pp signal can make room for another pp signal. If no space is found, it is checked in step 72 whether all time slots have been processed, and a new time slot is selected in block 73. A first-degree permutation means that a move is performed when an already arranged pp signal and a pp signal to be processed overlap at a port. In block 71, the signal already processed is removed and moved to a new free time slot (ie. time slots), and the pp book-keeping table is updated correspondingly. The signal to be processed is now inserted in the released time slot. From block 71 the process continues to block 65.

If the first-degree permutation fails, there follows a second-degree permutation, which is denoted by the looping arrow D in Fig. 6.

The second-order permutation functions in the loop formed by blocks 74-75-76-77 in a manner similar to that described for blocks 69-70-72-73, but now there is no immediately free time slot available, and the pp signals hit not only one common port, as before, but two common ports. These two common output ports can belong to one or two pp signals. Now, a free time slot is searched for these signals. If new free locations are found, they can be emptied for the pp signal processed, which is then moved to the free time matrix. If no solution is found in loop D, the process moves on to block 78, which sends out an error message, and the process is then ended in step 61.

After the pp collection routine. the input table time slots have been changed. These changes have to be synchronized for the different signals and they have to correspond to the contents of the SwapRecord table. The new positions of the pp signals also have to be updated in the pp queue so that backup connections and switching changes can be performed swiftly.

After the pp signals have been collected, the algorithm calculates the easy normal connections, for which the input table contains only one signal for each input port and time slot. Overlapping normal connections have to be moved so that each output port is used only once. The ControlSetup process (54 in Fig. 5) first creates the rearrangement control data. The number of target output ports in the input table is marked for every time slot in the OutputReference table. For pp selection signals, only one is taken into account. At the same time, the OutputQueue of each output port in the output queue is filled with the corresponding input port numbers. If the connection belongs to a pp signal, also the pp number is marked in the output queue. The Overlap counter is incremented at every overlap. The TimePP path protection pointer is set to ON when the time slot contains a pp signal, as was mentioned in conjunction with Fig. 4.

Then the algorithm moves on to process FindEven (55 in Fig. 5), which is used for solving easy normal point-to-point signals, ie. finding the even solutions, as illustrated by means of the two time slots Time=T and Time=S in Figs. 7a and 7b.

Because of symmetry, normal signals have a corresponding free output port in another time slot. This is due to the fact that when considering all time slots, only 63 VC-12 signals can reserve a particular port, whereby one overlap in an output port means that the same output port has at least one unused time slot. Because of the overlapping time slot, there is at least one free output port. The free output port in the other time slot S can be used directly, or the free output port in the overlapping time slot T indirectly, for another overlapping signal by means of a FindEven swap. The use of the free port is even only if one swap decrements the Overlap counter by at least one.

The FindEven process (55 in Fig. 5) for even solutions is depicted in more detail in the block diagram of Fig. 8. The process begins at step 80, and first' in block 82, a first element is taken from the OutputReference table, and all elements from 1 to 1008 (= 16*63) are processed in loop A. When block 83 indicates that the element number of the OutputReference table > 1, this indicates a possibility of a FindEven action for the overlapping output, whereby the process moves on to loop B. In loop B, the other time slots are checked for an empty (=0) OutputReference [Output][Time] element in the OutputReference table (Fig. 7, Time = S). In block 86 it is examined which inputs (Overlaplnputs) cause overlapping, and time slot = is set as the first time slot to be examined.In decision block 87 it is examined whether all time slots have been processed. If not, execution moves on to block 88, in which it is checked whether the value of the table element = 0. If the element is not empty, the next time slot is selected in block 89.

When an empty time slot is found in block 88, in this example time slot S, block 90 examines whether the input port Input is free in the new time slot S. In block 91 it is examined whether element OutputReference [Output][T] is empty, which means that a direct swap is possible between time slots T and S in block 92. This situation of the FindEven process is depicted in Fig. 7a, in which input ports 6 and 10 are targeted to output port 4 in time slot T. Input port 6 in time slot S is empty. If the input ports in the new time slot S are not empty, an indirect FindEven swap is possible, if the contents point to an output port which is not used in the original time slot T. The latter situation is depicted in Fig. 7b, in which the contents of port 6 of the InputTable in time slot S point to an empty output port 8.The process is ended in step 81 when the whole OutputReference table has been processed.

The swap function in the FindEven process is a little different from that of the PPCollection process. Depending on the nature of the swapped signals, now also the control data fields have to be updated simultaneously with the swap. This is also true for the PullPush process (56 in Fig. 5), which is below described with reference to the block diagram of Fig. 10.

The third main part of the algorithm concerns processing of difficult signals, which requires a bit more complex process, even though its principle is fairly straight forward. Hall's statement ("A distinct representation is present in the sets st if for all k there is k different representatives found in any union of k sets s") implies that when an overlapping connection is moved out of a time slot, a non-overlapping connection will be found for it after a finite number of moves, when examining the output ports of the first time slot. This, however, requires that the signals in question are point-to-point signals. Hall's statement defines the push swap rules. The push swaps as such do not necessarily use the shortest way to a solution.Although the data structure selected for the PPTST algorithm, ie. the InputTable, facilitates a short path, it may become difficult to uphold that short path. Because of pp signals, finding a solution is difficult, and in some cases a solution is impossible, in which case blocking will occur. The protection concept in which backup is realized by a space switch, inherently contains a possibility of blocking when operating in the TST architecture. But in the case of a point-to-point connection the pure use of the push sequence is easier to control, even if with a longer solution path. The pull swap manipulation is needed in the selected PPTST algorithm to help the push sequence stay on the good, short solution-seeking path in spite of the pp signals.

The purpose of the combined push and pull procedure is to identify the best final time slot where the route of the replacement connection to the starting point, or the overlapping connection, can be found. The replacing connection, which is usually connected to another output, is then moved to the overlapping time slot, and the connections between the first and the last connection can be rearranged by means of Hall's push sequence, until the last overlapping output is solved. An effective Halltype application is based on knowing the first and last connections of the sequence.

If the last, or the final, connection is not known, the rearrangements of the Hall-type push sequence will go round in local loops until one final solution is suddenly found by accident, whereby a solution is found for the reordering sequence.

A straightforward implementation of Hall's statement, ie. the push sequence, is also impossible because of the pp signals. Therefore, the algorithm must include an escape routine, with which the pull-push procedure can break free of a deadlock caused by seldomly moved pp signals. Another alternative would be to allow easy moving of pp signals, ic. to give them the same status as normal connections. Then, however, it might become difficult to converge the sequence because of the avalanche effect, which was discussed before.

The pull-push routine is continued until variable Overlap =0, ie. all overlapping ports have been solved. The process in Fig. 10 begins with step 100, whereafter the first element is selected in block 102 from the OutputReference table. Then it is examined whether the OutputReference[Output][T] element > 1. If not, the process goes on in loop A and the next element is selected in block 105, until in block 104 Overlap = 0, and the process is ended in step 101. If the situation is unsolvable, the process examines different swap options or it is ended when time-out occurs.

If in block 103 the OutputReference[Output](T] element > 1, it indicates an overlap, and the process will start the actual pull-push application, denoted by looping arrow C in the figure. For the pull-push loop there may be some crucial control values in an overlapping time slot: the outputs that are not used and the inputs that cause the overlap. These inputs and outputs constitute a set of connections. which will be subject to selections in the pull-push sequence. In block 106 it is examined whether the locked mode is set, and if it is, the status is changed in block 107, ie. the time slot is swapped, and if not, a new time slot T+1 is selected in block 108. Then the pull-push process selects a connection option for one signal, using a priority table described below with 8 different priority types. These cases are depicted in Figs. 7a, 7b, 9, and 11 to 16.Thus the process has moved on to subloop B shown in Fig. 10.

If the pp signals cause no interference, the sequence finds an option which finally decrements the Overlap value. In the case of interference caused by a pp signal, some of the options allow divergence from the best sequence path, and the sequence will be ended so that a better starting point can be found for the straightforward pull-push sequence. All options have different priorities, and the algorithm uses different options depending on their availability. These eight options are also searched for in time slots other than the time slot T processed. In block 109, one of the eight options is selected, and in block 110 it is examined whether options BestSwap or FindEven are available, and if not, the process moves on to block 111 which examines whether the locked mode is active. If the locked mode is not set, block 112 examines whether all time slots have been examined, ie. whether the process has come back to the first time slot T. Via block 113 the process moves on to the next time slot. If any of decision blocks 110 to 112 returns "yes", the process moves from loop B to decision block 114, in which is it examined whether there are still options available and whether there is still time left. If calculation can be continued, the highest-priority option is swapped in block 115 and the Output and time slot T are set according to the option.

Execution of the process continues in loop C until the reordering sequence solves at least one overlap, which is found out in block 103. The pull-push routine continues in loop C since the Output and T values are set according to the selected swap option. In other words, if the overlap is not solved by means of a swap, the Output and T values are set such that the reordering sequence follows the overlapping output, and block 103 cannot break the loop in the middle of a running reordering sequence. Overlapping signal locations are swapped using the FillerSwap option from time slot T to time slot S, as seen in the priority list and Fig. 12. The changed Output and T values are shown in the priority list below. Loop C can be ended successflilly with the FindEven option (Figs. 7a, 7b) or the BestSwap option (Fig.

9). An unsuccessful attempt is ended with the ReleaseSwap option (Fig. 15), thereby exiting loop C. Also the PPSwap option may end with a break of loop C.

Block 116 produces an error message and the process is ended in step 101 when there are no more options to be processed or when a time-out occurs (block 114).

Listed below are the 8 options available and their priority values starting from the highest priority, and their impact on the output and time slot values: Option priority Fia. output value time slot BestSwap 7 9 Output Output l T, = Tk l FindEven 6 7a,7b Output Outputk l Tk.1 HallSwap 5 11 Output Outputk,l Tk = Tk-l FillerSwap 4 12 Output Output l Tk = Sk-l BestReplace 3 13 Outputk = Outputk-1 Tk = Tk-1 ReplaceSwap 2 14 Output Outputk-1 Tk = Tk-1 ReleaseSwap 1 15 Outputk Outputk-1 Tk = Tk-1 PPSwap 0 16 Output Outputk-1 Tk = Tk-1 Subscript k- 1 represents the output and time value of a performed swap, and subscript k indicates the next value.T means the overlapping time slot and S indicates the time slot in which the replacing connection is made.

A locked pull-push sequence is activated in block 109 if there are no pp signals in the time slots of the overlapping output ports and the replacing connection, ie. if TimePP = OFF (Fig. 4) in both time slots. The locked mode also requires that there be a zero in the other time slot in the OutputReference table of the overlapping output. In the locked mode, swaps are made only between these two time slots, which means that no other time slots are handled. Therefore, the sequence will not get into more uncertain and in some cases less feasible paths. A locked sequence needs a maximum of 16 swaps to solve an overlapping connection when the DXC is ofthe size 16*16.

One of the most important rules in the pull-push routine is not shown in Fig. 10.

Namely, if a connection moved in the pull-push sequence causes a new overlap, it must not be moved again immediately after arriving in another time slot. The next connection to be moved is another one, advantageously the new one that was just selected. The moved connection also changes the states of the associated input and output. Therefore, moves are not allowed without time slot changes, ie. identical connections are not allowed. This rule forces the push sequence to try new combinations, in other words, it rearranges the connections in the time slot. The rule also applies to both signals involved in the swap.On the other hand, pp signals usually prevent the pull-push sequence from pursuing a solution in a straightforward way, because if a moved signal bound to an output hits a pp signal using that output, the swapped normal connection must not be immediately moved again because of the rule mentioned above. Therefore, the rule must be complemented with a release option, which is realized using the low-priority options BestReplace, ReplaceSwap, ReleaseSwap and PPSwap.

The PPSwap option contains a set of actions more complex than the other options.

PPSwap means that a pp signal is moved to another time slot. The target time slot of the pp signal is calculated using the same principles as in the PPCollection routine.

The whole PPCollection routine is not necessary as it is only one signal that is being processed now. Fig. 17 depicts such a revised routine. Processing of the eight pullpush options listed above begins after the start block 170 by first setting in block 172 the pp number to the number of that pp signal which is to be moved out of the overlapping time slot T. In block 173 it is indicated whether the overlapping output shown in Fig. 16 is used in time slot T. If not, the process comes to an end in block 185. which generates an error message, and the process stops in step 171. Otherwise, the process moves on to block 174 which searches for room for the pp signal, ie. examines the elements in the PPBooking table and determines whether there is a free output port for the pp signal when the time slot is not T.Those pp signals that possibly have to be moved out of time slot S are found out When a free location is found, the signal is moved and the process ends in step 171.

If no free target is found in the decision block 174, a new time slot, which is not T, is selected for examination in block 175. Then it is examined in block 176 whether the swap yielded a successful result. If not, all time slots are examined in loop C successively, until the process can move from block 176 to block 177.

If no result is obtained in the first permutation loop C after all 63 time slots have been processed, the execution moves from decision block 178 to a second permutation loop, and a new time slot, which is not T, is selected for examination in block 181. Then it is examined in block 182 whether the swap yielded a successful result. If not, all time slots are examined in loop D successively, until the process can move from block 182 to block 177. If no result is obtained in the second permutation loop D after all 63 time slots have been processed, the execution moves from decision block 183 to block 185 which produces an error message indicating that no PPSwap options were found and the process ends in step 171. If the second permutation round produces a positive result, the process moves on to block 177.

In block 177 of Fig. 17 the pp signal or pp signals are moved to the new time slot or time slots, and the corresponding PPBooking data are updated. Thereafter, the process moves on to block 180 and the process ends in step 171.

The PPSwap options found in the loops of the process shown in Fig. 17 are arranged according to the Input that causes the overlap in the original time slot T. If this input is in the targeted time slot, ie. time slot S, the pp signal then becomes the choice of the PPSwap process (Fig. 16). This Input cannot be connected to the overlapping Output. Such a situation is depicted in Fig. 16.

The principal implication of the last pull-push option, or the PPSwap process, lies in the fact that the order of the pp signals is changed without considering the connections of normal signals. Such action possibly increases the number of overlaps, le. increments the Overlap variable. because reordering the pp signals causes overlapping with already solved normal signals.

The PPTST algorithm of this invention is aimed at solving the whole configuration table. The time requirements set for the task led to the algorithm according to the invention, used to balance the processing power assigned to different tasks. The algorithm as such meets the execution time requirements without problems.

The algorithm solves the whole switching configuration, but it is also possible to have the algorithm solve one connection only as it solves only connections that require calculation. Therefore, the connection request can be in different forms, ie.

involving only one signal, several signals or the whole configuration matrix, ie.

16*63 signals.

The algorithm ignores AU-4 signals since these can be handled without special calculation as the > use the whole port, ie. all time slots in the port, and thus no other signals can then access the same space switch ports. Therefore, path protection measures on the AU-4 level can be performed without special calculation. On the AU-4 level, path protection in fact means copying the whole contents to another AU-4 signal, which can be realized by routing both protection signals simultaneously and using two output ports instead of one, as normally.

A special characteristic of the algorithm of the invention is the utilization of timers to supervise that the process will not get stuck in a solution loop if the PPTST algorithm cannot solve all possible connection cases when the signals include pp signals needing path protection. At least two time limits are required: one for supervising the calculation time for distinct connections and the other for supervising the calculation time for the whole switching matrix.

Claims (11)

Claims

1. A method for configuring the switching of digital signals connected to a digital time-space-time (TST) switch between given input and output ports in the switch, characterized in that the switching problem is described by creating time slot specific imaginary switching matrices (Fig. 4) for the input side time switch and space switch, whereby the number of matrices corresponds to the number of switched subsignals received in each input interface, and a column in the switching; matrix corresponds to a given input port (Input) and a row in the matrix corresponds to a given output port (Output), and a configuration solution is achieved by moving signals in the switching matrix so that eventually each row corresponding to a particular output in the switching matrix contains only one signal to be switched ie.

only one distinct signal is connected to a particular output port of the time switch whereby a) path-protected input signals are handled in a certain order and a free location is found for the parts of every path-protected signal in one selected time slot (52); whereafter b) point-to-point signals causing an overlapping output port reservation are marked in a table (54) controlling the switching; c) overlapping point-to-point signals are solved one overlapping signal at a time (55) by locating the signal in question in a time slot in which a free output port is found; but if no solution is found for a particular overlapping signal, the process moves to the next step, in which d) the remaining overlapping point-to-point signals are solved one overlapping signal at a time (56) by swapping the places of the signal in question and an already solved signal meeting the predetermined criteria;e) whereafter the switching configuration is completed with suitable connections in the output side time switch.

2. The method of claim 1, characterized in that before starting steps a) to e) the time slot specific switching matrix is provided with auxiliary tables, in which - an input table (InputTable) indicates for an input port the existence and code (PPNumber) of a path-protected signal and the desired output port (Port) and time slot (Time) for the signal in the input port; - a book-keeping table (PPBooking) indicates for an input port the existence and code (PPNumber) of a path-protected signal (PP); - an output reference table (OutputReference) indicates for an output port how many signals are targeted to one and the same output port in a time slot; - an output queue table (OutputQueue) indicates for an output port the input ports (Port) of the signals targeted to one and the same output port in a time slot and the code (PPNumber) of a possible path-protected signal; and - a swap record table (SwapRecord) indicates for an input port the original time slot (Time) of the signal in the corresponding input table (InputTable); - an overlap variable (Overlap) indicates the total number of the overlaps in the switching matrices of all time slots; and - a path protection pointer (TimePP) indicates whether the time slot in question containers path protection signals.

3 The method of claim 1 or 2. characterized in that the signals to be switched are multiplexed subsignals of high-speed signals, advantageously 2-Mbit/s VC-12 virtual containers in the SDH system, whereby the main signals are 155-Mbit/s STM- 1 signals, and that before switching the higher-order signals are chopped into multiple VC-12 signals which will be re-combined in the output time switch of the switch.

4. The method of any one of the preceding claims I to 3, characterized in that a path-protected signal (PP) selected at the initial stage (62) is handled in step a) and after that, all other path-protected signals are handled by turns in a loop (loop B: 63, 64, 65, 66, 79): - a free time slot is fetched (loop A: 64, 67, 68) for the signal (PP) in the bookkeeping table (PPBooking); but if no free time slot is found - a first permutation is executed (loop C: 70, 72, 73) in which one already processed signal (PP) is located in a new time slot; but if no solution is found - a second permutation is executed (loop D: 75, 76, 77) in which one or two already processed signals (PP), which have two ports of the path-protected signal in common, are located in a new time slot, whereafter - changes that have been made are recorded (PPBooking).

5. The method of any one of the preceding claims 1 to 4, characterized in that a signal selected at the initial stage ( 102) of the pull-push process is handled in step d) and after that, all other signals are handled by turns in a loop (loop A: 103, 104, 105) in which a signal meeting predetermined criteria (BestSwap, FindEven, HallSwap, FillerSwap, BestReplace, ReplaceSwap, ReleaseSwap, PPSwap) is fetchedinloops (loop B: 109, 110, 111, 112, 113; and loop C: 106, 107, 108, 114, 115).

6. The method of claim 5, characterized in that in step d) it is also performed if necessary, a forced time slot swap for the still overlapping subsignals of the pathprotected signals (PP) and the action of step d) is repeated.

7. The method of claim 6, characterized in that the forced swap is performed using random selection.

8. The method of any one of the preceding claims, characterized in that in each step a) to d) the operation is supervised by a timer and that the processing of the signal is interrupted if the predetermined maximum calculation time is exceeded.

9. A circuit arrangement realizing the method of any one of the preceding claims 1 to 8, characterized in that the auxiliary tables (InputTable, PPBooking, OutputReference, OutputQueue, SwapRecord, Overlap. TimePP) representing the switching data in each time slot of the imaginary switching matrix are implemented with an ASIC circuit/circuits and/or memory circuit/circuits.

10. The circuit arrangement of claim 9, characterized in that the switch has 16 input ports and 16 output ports.

11. The circuit arrangement of claim 9 or 10. characterized in that there are 63 imaginary matrices.