NO783069L - DATA PROCESSING EQUIPMENT. - Google Patents

Description

Data processing equipment

The present invention relates to a data processing system which comprises a data processing machine capable of processing a certain predetermined amount of data, hereinafter called workload, recording equipment for registering the amount of newly added workload that is accepted for processing by said data processing machine, detection equipment for periodically detecting the busy level of said data processing machine, as well as control equipment to regulate the workload, which control equipment is controlled by the detection equipment to periodically regulate the maximum permissible workload that can be processed by the data processing machine, as a function of the detected busy level.

Such a data processing system is already known from Norwegian patent no. (application no. 754318). The system shown there relates in particular to overload management of a telephone exchange. In this known system, the workload regulation equipment, at the end of each regulation period, compares the detected busy level of the data processing machine with a minimum and a maximum value and decreases or increases the maximum permitted workload, that is in this case the maximum permitted number of new calls which can be accepted for processing during such regulation periods, or maintain the workload at a constant value depending on whether the detected busy level is above the maximum level, below the minimum level or between these levels. The various maximum permissible workloads constitute a set of predetermined discrete values, and modifications of these are carried out in stages.

A disadvantage of this known system is that the processing of a number of new calls will be unnecessarily greatly delayed in certain cases despite the fact that a desired maximum busy level has not yet been reached in the data processing machine, where this maximum busy level is the one for which all service criteria are still met . Suppose e.g. that the number of new calls offered for processing by the telephone system increases significantly after the end of a regulation period in which the busy level of the data processing machine and the number of calls actually accepted for processing during this period were respectively well below the maximum busy level and below the permitted maximum number of new calls that must be able to be processed during this regulatory period. Such an increase in the number of new calls offered to the system can e.g. is due to a damaged multi-conductor cable between the exchanges, which will lead to an increase up to a very high number of detected line loop closures which stimulate a corresponding number of incoming calls. In such a case, the number of calls accepted for processing by the data processing machine in the subsequent regulation periods will increase significantly and quickly reach the maximum allowed value that has been determined for certain periods, while the data processing machine busy level will generally not increase. In reality, the computer after detecting such new false incoming calls and after accepting them for processing, will soon stop this processing because for such calls no prefix is received from the remote exchange to continue processing, so that little time is used per . false call and consequently a high number of such calls will be able to be processed. Although a high number of false new calls will thus be processed, the processing of a certain number of normal calls will be delayed. Due to the busy level remaining, constant or even increasing slightly, but still staying below the maximum level, as assumed, the maximum allowed number of new calls to be accepted in subsequent periods will be maintained unchanged.

The purpose of the present invention is to provide a data processing system of the above-mentioned type, but which is not affected by the above-mentioned disadvantages.

This is achieved by designing the data processing system in accordance with the patent claims set out below. In the case of a data processing system according to the present invention, the permissible maximum workload will thus, within certain limits, not only depend on the detected busy level for the data processing machine,

as in the known system, but also be dependent on the amount of workload actually served by this computing machine. E.g. in the case mentioned above and which is due to such factors as the number of processed calls increasing while the busy level remains approximately unchanged, the calculated permissible maximum workload will be greater than that calculated for the previous period. This then means that one can freely allow an even greater workload to be serviced.

Other telephone traffic congestion control systems, and particularly those using a programmed electronic traffic simulator, exhibit strategies whereby the workload of a data processing machine is modified as a function of the machine's busy level. It can e.g. appears to US Pat. nos. 3,517,123 and 3,623,007 as well as the article "Real time testing of overload control systems in a laboratory environment" by F.T. Mann, published in IEEE Transactions on Communications, September 19783, pages 1027-1031. Workload is limited by placing restrictions on the number of new calls that can be serviced during a fixed time interval and in the U.S. patent no. 3,517,123 this is specifically accomplished by cutting off the data processing machine from answering requests for service from all but specially preferred service lines. However, in the same way as in the above-mentioned Norwegian patent, the number of new calls that are actually accepted for processing during the immediately preceding fixed time interval is not taken into account.

It should also be noted that before dynamic congestion management was introduced in local traffic within telephone exchanges, similar procedures had been introduced in telephone networks, in e.g. U.S. patent no. 3,796,837 for the purpose of re-laying the line for long-distance calls in order to avoid that certain exchanges are overloaded. However, such systems are not particularly dependent on the use of data processing machine-controlled telephone exchanges, and such known techniques for rescheduling the routes in telephone networks also do not include the calculation of new workload limits based both on a previously accepted current value and on the relevant busy level for the data processing machine.

In the case of a data processing system according to the present invention, moreover, the maximum permissible workload, at least within certain limits, is directly dependent on and varies in the same direction as a desired maximum value for the occupied level.

Because the permitted maximum workload thus varies in the same direction as the desired maximum occupancy level, but in the opposite direction to the detected occupancy level, it can be ensured that this permitted maximum workload is independent of the detected occupancy level when this takes its maximum value.

A data processing system according to the present invention

is also structured in such a way that the regulation equipment for the workload periodically calculates the maximum permissible workload for each part or for a controlled period of a regulation period.

By choosing the durations of the management periods and the regulation periods differently from each other, e.g. by letting the latter be four times as long as the former, each of these durations will be given an optimal value as explained below in the description.

Other special features of a data processing system according to the present invention is that the regulation equipment for the workload at the start of each new regulation period calculates the maximum permissible workload for each of these management periods in the new regulation period by calculating a provisional value which is directly dependent on and varies in the same direction as the amount of new workload accepted for processing on average during each of the control periods of a previous control period preceding the new control period, and which also varies in the direction of the desired maximum value for the busy level, and which is directly dependent on and varies in the opposite direction to the occupancy level detected for the previous regulation period, and that the regulation equipment to regulate the workload also calculates these limits and accepts the previous value as the current maximum permissible workload when the temporary limits is being respected.

In this way, the permitted maximum workload can be calculated in a simple way, while all significant conditions are taken into account in the calculation.

Another distinctive feature of a data processing system according to the present invention is that the provisional value N(p,i+1) is calculated at the beginning of each new regulation period according to the following formula:

where

M(i) denotes the amount of new workload that is accepted on average for processing during each of the aforementioned control periods within the aforementioned preceding control period,

0(M) indicates the desired maximum value for the occupied level 0(i) indicates the detected occupied level for the previous

regulatory period,

V denotes a constant that is less than 0(i).

Using the above formula, the preliminary value for the maximum permissible value of the workload can be calculated in an efficient way because V in a telephone system can represent the time used to perform tasks that are independent of the call processing, i.e. when there is no traffic, below also called idle time.

According to a preferred embodiment of the present invention, an automatic telephone system comprises switching equipment and a data processing machine capable of handling a number of calls, recording equipment for recording the average number of new calls accepted for processing during each of several control periods in a regulation period, detection equipment for periodically detecting the busy level of the data processing machine, and regulation equipment for regulating the workload by periodically calculating at the start of each regulation period, a provisional maximum number of new calls N(p,i+1) capable of being served during of each of these management periods, according to the formula

where:

N(p,i+1) is the provisional maximum number of calls in each of the (e.g. four) consecutive management periods that make up the regulation period,

M(i) is the number of new calls that are actually accepted on average for each of the control periods in a previous period that precedes the regulation period,

0(M) is a desired maximum value for the occupied level,

0(i) is the actual occupancy level occurring for the previous regulation period,

V is the fraction of time within each regulation period that is used to perform tasks that are independent of call processing, also called idle time below.

This preliminary value of N(p,i+1) will be considered the maximum number of new calls or can be modified to a final value of N(i+1) depending on boundary conditions that are set to N(p,i+1), 0( i) and M(i).

In order to provide a clearer understanding of the present invention, reference is made to the below detailed description of an embodiment according to the present invention and to the accompanying figure showing a schematic block representation of a data processing system according to the present invention.

This data processing system more particularly forms part of an automatic telecommunications switching system and comprises a telecommunications switching equipment TSE and the relevant data processing system. The latter comprises a data processing machine which in turn comprises a memory MEM and a processor PR.

The data processing machine is able to collect call information from the switching equipment TSE and in accordance with this to control the establishment, monitoring and disconnection of call connections through the switching equipment TSE by sequentially executing higher and lower priority tasks or clock and base level<p>programs. priority tasks or clock and base level programs. The computing machine starts the execution of the higher priority tasks in each so-called clock interrupt interval of 20 milliseconds, and as soon as these tasks are executed and time is available, it starts the execution of the lower priority tasks. Normally, the execution of these lower priority tasks will also be completed before the start of the next subsequent 20 millisecond clock interrupt interval. Accordingly, for each clock interrupt interval of 20 milliseconds, this computing machine will and if the number of tasks performed by the data processing machine is normal, have some free time available. Of course, when the number of tasks performed by the data processing machine increases, the available free time will decrease accordingly. Thus, this free time will be a measure of the amount of work performed by the data processing machine and therefore also of its busy level.

Among the tasks performed by the computer, some are caused by the actual handling of calls, while another part of the tasks is independent of the call processing. This last part of the computer's tasks, which is therefore independent of the handling of the calls and which also

must be carried out even when the traffic is 0, is often called "overhead" load or in Norwegian can be called idle load. Because the time required to carry out such idle operations is fixed, it is clear that only variations in the processing time of the calls lead to variations in the computer's free time and therefore also to variations in the computer's busy level. In order to handle these variations, the computer periodically detects the busy level of the computer, calculates from the detected busy level a maximum allowed number of new calls that can be accepted by the computer for service during a fixed time interval, and limits the number of new calls allowed to be accepted to this number, i.e. to the number that the processor will really start serving. This is described in more detail below.

The computer comprises equipment for periodically detecting its own busy level, during a fixed time period T, which e.g. can be 4 seconds. This is done by, at the end of any such regulatory period, e.g. P(i), then the total idle computer time F(i) is measured, which remains unused during that period P(i), and the busy level 0(i) is calculated for that regulation period using the following expression:

The computer also includes control equipment for periodically calculating, after the detection equipment has determined the busy level 0(i) and also now with a periodicity equal to T, a temporary value N(p,i+1) for the maximum allowed number N(i+1 ) new calls that can be accepted for operation of the data processing machine during each of the control periods of the immediately following control period P(i+1). These management periods are fractions of regulation periods and have e.g. a duration equal to T/n, where e.g. n = 4 and thus T/n = 1 second.

The value of the expression N(p,i+1) is obtained by calculating the formula:

where:

N(p,i+1) and 0(i) are as defined above,

V is the above idle load, expressed as a fraction

of the period T,

M(i) is the average number of new calls that are actually accepted for service during each of the four control periods in each control period P(i),

0(M) is the desired maximum value for the computer's busy level.

It should be noted that in the above equation (2) the value V is subtracted from 0(M) and 0(i) because the number of calls to be or are served is independent of V. 0(M) - V is the maximum time that is available for such processing and 0(i) - V is the time allowed for such processing during the regulation period P(i).

The thus calculated provisional value N(p,i+1), after it has been rounded down to the nearest whole number, is taken as the new value N(i+1) for the regulation period P(i+1) if N(p ,i+1), 0(i) and M(i) have predetermined values in relation to the limit values (l-b)N(i), (l+a)N(i), N(i+1), 0(m ), 0(e), fN(i).

Here is:

a is a number less than 1 so that aN(i) is the maximum

allowed the increase of N(i+1) in relation to N(i),

b is a number less than 1 so that bN(i) is, in general, it

maximum permissible reduction of N(i+1) in relation to N(i),

f is a number less than 1 so that fN(i) is a minimum value of

the number of calls that are allowed to be accepted over a management period,

0(m) is the minimum value for the computer's busy level,

0(e) is an excess value of the occupied level.

The rules for choosing the final new value N(i+1) for the regulation period P(i+1) are as follows: - when 0(i) < 0(m) and furthermore M(i) < fN(i), then set the new value N(i+1) equal to N(i), regardless of the calculated value N(p,i+1);

- when either 0(i) ^ O (m) or M(i)^ fN(i) and furthermore

0(i) <' 0(e) then N(i+1) is either set equal to (l-b)N(i), (l+a)N(i) or N(p,i+1) depending on whether N (p,i+1) respectively is less than (l-b)N(i), greater than (l+a)N(i) or lies between these two limit values. However, N(i+1) is set equal to N(i)+1 if both N(p,i+1)^> N(i)+1 and at the same time aN(i) <^1;

- when 0(i) > 0(e), N(i+1) is set equal to (l-b-d)N(i), or (1-d)N(p,i+1) depending on whether N(p, i+1) is less than (l-b)N(i) or, respectively, lies between the values

(1-(lrb)N(i) and (l+a)N(i), where d is a number less than 1 that allows the calculation of the additional reduction dN(i) or dN(p,i+l),

where b is assumed to be the fraction that allows the calculation of a normal reduction. It should be noted that when 0(i) > 0(e), then N(p,i+1) will always be less than N(i)(1+a) because this value, according to equation (2) and because 0 (e) > 0(M), will always be less than M(i) which cannot exceed N(i).

The above rules are summarized in the following table, where y(l) to y(7) are the results of the following inequalities, where y2-y3 and y6*y7 represent the logical AND of the elements y2 and y3 and respectively of the elements y5 and y6, and where the value of yl to y7 is 1 when the corresponding inequality is true and has the value 0 when the inequality is false, x-means that the value of the elements yl to y7 is irrelevant for the choice of N(i+1).

Examples of the value of the above parameters are:

0(m) = 0.75 a = 0.1

O(M) = 0.95 b = 0.015

0(e) = 0.975 d = 0.1

V = 0.40 f = 0.75

T = 4 sec.

N

77=1sec.

With regard to these parameters, the following should be noted:

- The duration T/ n of the control period

On the one hand, this duration should not be too small, because when it is small, only a correspondingly small number of new calls can be accepted ,• and. this .can cause an increase in call processing delays when the number of calls offered for service is high. If the control period is extremely short, then even if the allowed number of calls results in a busy level for the data processing machine that is significantly less than the maximum busy level O(M), an increase of 1 in the latter number may cause the busy level to rise to a value which is greater than O(M).

On the other hand, this duration should not be too long, because when it is long, a correspondingly large number of new calls can be accepted, and this will have the effect that these calls will be accepted for service during the first 20 milliseconds of the clock interrupt- intervals in the control period, and this in turn can cause a severe peak load and correspondingly high values for the computer's busy level. This can be alleviated somewhat by limiting the number of accepted calls per clock interrupt interval of 20 milliseconds as explained e.g. in the above article by F.T. Mon.

- The duration T of the regulation period

On the one hand, this duration should not be too short, as a random lowering of the number of calls accepted for service, such as M(i), will cause the maximum allowed number of calls, such as N(i) to undergo a corresponding reduction, because the computer's busy level will not immediately reflect this reduction with lower values. In this case, this could lead to the service of calls being delayed.

On the other hand, this time duration should not be too long, because then the data processing machine will not react immediately to sudden changes in the number of calls offered for service.

It should be noted that very rapid increases in M(i) due to sudden peaks in the number of calls offered for service do not amount to anything

problem because M(i) is bounded by N(i).

- Maximum occupied level 0( M)

This is the maximum desired busy level for the data processing machine, i.e. such a large busy level that a maximum number of calls can be serviced during a predetermined time interval while the data processing machine still satisfies all service criteria. The value of 0(M) is chosen somewhat smaller in practice, e.g. 3% less than the maximum permitted occupancy level. If 0(M) is set too low, this may cause a degradation of the service level during periods of high traffic, because calls that could have been satisfactorily serviced by the data processing machine will then be unnecessarily delayed.

- Excess busy level 0( e)

This is a dangerously high busy level value that should not be exceeded for anything but very short time intervals. It is set just below the maximum permitted occupancy level.

- The minimum occupancy level 0( m)

The value of this parameter is e.g. equal to 0.4.

- Idling V

This value depends on the type of connection system and can e.g. is set equal to 0.4.

- The parameters a and b

These parameters are inserted as an aid to prevent N(i) from fluctuating strongly, e.g. caused by random fluctuations in call arrival or in the distribution of the call type, i.e. call mixing. In practice, a will be significantly larger than b, because it has been found that while it is useful to strictly limit the reduction of N(i), in most cases you will get a better service level by allowing N(i) to increase more freely. The reason for this asymmetry follows from the above discussion of the duration of the regulatory period.

- The parameter d

As explained above, this parameter is used to calculate an additional reduction of N(i).

- The parameter f

The value of this parameter is set e.g. equal to 0.75.

Referring to the figure, the above data processing system will

now described in more detail.

The memory MEM of the data processing machine comprises several storage locations SLI to SL28, while the processor PR comprises a bistable circuit BC, four clocks CL1 to CL4, a comparison circuit CR, two registers REI and RE2, and a call service and control arrangement CPA, which is connected to the telecommunications switching equipment TSE over the connection cl, to the memory MEM over the connection c2, to the registers REI and RE2 over the connections c3 and c4 respectively, to the comparison circuit CR over the connection y, to the clocks CL1 to CL4 over the connections c5 to c8, and to the 1 inputs of the bistable the circuit BS above the connection c9. The way the CPA works will be clear from the description of the way the system works.

The contents of the memory MEM are considered e.g. at the beginning of the last or fiftieth clock interrupt interval of 20 milliseconds in the last CP4 of the 4 control periods CP1 to CP4 to the ith control period P(i). The following data is then stored in this memory MEM: - above-mentioned constant parameters 0(e), O(M), O(m), a, b, d, f, V and also 1. These parameters are stored in the memory of the CPA over the connection c2 at the start of the operation of the system; - several variable parameters N(i), N(i)+1, (l+a)N(i), (l-b)N(i), aN(i), fN(i), O(i-l), M (i-l), N(p,i) and the elements yl to yl. These parameters have been stored in the memory of the CPA over the connection c2 at the end of the previous regulation period P(i)-1, and below that the following definitions must be remembered: - N(i) is the maximum allowed number of new calls that can be accepted for service for each of the four successive control periods CP1 to CP4 which constitute the control period P (i); - O(i-l) is the busy level of the data processing machine during the regulation period P(i-l); - M(i-l) is the average number of new calls that have been accepted for service during each of the four consecutive control periods CP1 to CP4 in the regulation period P(i-l); - N(p,i) is the calculated provisional number of new calls to be accepted for service during the regulation period

P(i) i

The memory MEM also stores the following other parameters:

- M which is the number of new calls which have already been accepted for service during the current control period CP4 to the current" regulation period P(i); - S which is the total number of new calls which have already been accepted for service during the current regulation period P(i); - F which is the free time that has already been measured during the ongoing regulation period P(i).

The comparison circuit CR is able to compare the values stored in the registers REI and RE2 and affects (y=l) its output y when the value stored in REI is less than that stored in RE2.

The timing pulse generator CL1 generates output pulses at 20 millisecond intervals and thus defines 20 millisecond clock intervals in the call procedure and controls the circuit CPA over the connection C5. Through the connection clO, the clock CLl synchronizes the clock CL2,

and across the connection ell it is able to reset the bistable circuit BS.

The timing pulse generator CL2 generates output pulses at 1 second intervals and defines the control period of 1 second in the circuit CPA over the connection c6. Through the connection cl2, the clock CL3 synchronizes the clock CL2.

The clock CL3 generates output pulses at 5 second intervals and defines the regulation period of 5 seconds in the circuit CPA across the connection cl.

Finally, clock CL4 is connected to the 1 output of the bistable circuit BS across connection cl3 and generates output pulses at 25 microsecond intervals when circuit CPA has triggered circuit BS to its 1 state across connection c9. These output pulses are supplied to the circuit CPA. above the connection c8.

It should be noted that each of the above-mentioned connections cl to cl3 comprises one or more wires depending on the singal need.

For the sake of overview, it will first be described how the busy level of the data processing machine is determined }at the beginning of the regulation period P(i+1) by means of detection equipment comprising the clocks CLl, CL3, CL4 and the bistable circuit BS and which is controlled by the circuit CPA which cooperates with the memory MEM .

At the beginning of each 20 millisecond clock interrupt time interval in this regulation period P(i), the circuit CPA starts the execution of a series of tasks, and when these are finished before the end of each of the respective intervals, the call processing arrangement CPA sets the bistable circuit BS to its 1 state across the compound c9. As a result, this bistable circuit BS starts at CL4 which generates output pulses with a periodicity of 25 microseconds for counting purposes. Whenever the connection c8 is activated by such a pulse, the circuit CPA responds to this by addressing storage location SL28 in the memory MEM above the connection c2 to obtain from it the free time F stored in it. This free time is then increased by 1 and stored again at storage location SL28. At the beginning of the following 20 millisecond clock interrupt interval, the bistable circuit BS is reset to its O state by the clock CL1 across the connection ell. The storage location SL28 thus accumulates for every 20 millisecond clock interrupt interval the free time F that has elapsed since the execution of the last tasks of a series of tasks. The idle time F is more specifically stored in units of 25 microseconds.

At the start of each 20-millisecond clock interrupt interval defined by clock CLl, clock CL2, which operates at 1-second intervals, is synchronized, and each time a 1-second control period has expired, clock CL2 activates its outputs c6 and cl2. It should be noted that every time the connection c6 is activated at the end of such a 1-second control period, the circuit CPA responds to this by emptying the address location SL26 because the number M stored here must always correspond to the number of new calls already accepted for processing in during the current management period.

At the start of each 1-second control period defined by clock CL2, clock CL3, which operates on a 5-second basis, is synchronized, and each time a 5-second control period has expired, e.g. period P(i), clock CL3 activates its output cl. The call process and control arrangement CPA responds to the activation of this output cl by reading the free time F stored at address location SL26 in the memory MEM over the connection c2 and calculates the busy level by means of equation (1), T being equal to 4 seconds , must be expressed in the same units as F, i.e. in units of 25 microseconds, while F has the value F(i) because the termination of P(i) has already been reached.

It will now be described how the workload regulation equipment, which includes clock CL3, registers REI and RE2, and comparison circuit CR2 works. This control circuit is controlled by the circuit CPA and cooperates with the memory MEM to calculate N(i+1) for the period P(i+D •

At the beginning of the above 50th clock interrupt interval of 20 milliseconds for the fourth 1-second control period of the i'th control period, the circuit CPA starts executing a series of higher priority tasks. One of these tasks consists in collecting data about new calls from the telecommunications switching equipment TSE over the connection cl, and each time a call is detected. check whether this new call can still be accepted for service or not. For this purpose, the circuit CPA reacts to the detection of such a new call by reading the values M and N(i) from the address locations SL26 and SL10 in the memory MEM over the connection c2, by increasing the value of M by 1, by then adding the increased the value of M and the value of N(i) to the registers REI and RE2 across each of the connections c3 and c4, and finally by examining the state of the output y of the comparison circuit CR to determine whether this output y is at its Gj or 1 value.

If this value is 0, thereby indicating that the number of accepted new calls would be greater than N(i) if the new call were also accepted, this new call is not accepted for service. This is also the case for the new calls that are further detected during the control period CP4 in the period P(i). The system can, however, be designed in such a way that priority calls are always accepted, e.g. as in US patent no. 3.517.12 3.

If the output y=l, this indicates that the number of accepted new calls is still not. greater than N(i), and the circuit CPA then stores the increased value of M in address location SL26 to memory MEM over connection c2.

It should be noted that the value of M is the number of new calls that have been counted since the start of the control period CP4 because the circuit CPA has been informed of this start of the clock CL2 over the connection c6 and has reacted to this by emptying the address location SL26.

At the moment when the circuit CPA is informed by the clock CL3 over the connection c7 that the end of the regulation period P(i) has been reached, it performs the following functions, bearing in mind that at this moment the parameters S and F have reached the values S( i) respectively F(i). - it calculates the busy level 0(i) in the manner described above and stores this in address location SL16 in the memory MEM by overwriting the previous value O(i-l); - it reads the value S(i) from address location SL27 in the memory MEM and calculates the value M(i) = S(i)/4, where M(i) is the average number of new calls accepted for service during each of the four control periods CP1 to CP4 in the period P(i); - it reads the values 0(M) and V from the address locations SL2 and SL9 in the memory MEM, and with these values and the values M(i) and 0(i) it calculates N(p,i+1) given by the above equation (2 ); - it stores the calculated values of M(i) and N(p,i+1) in address locations SL17 and SL18 by overwriting the previous values M(i-l) and N(p,i) respectively.

The circuit CPA then checks whether the calculated temporary value N(p,i+1) can be accepted or not as a new value N(i+1) by calculating the limits of N(p,i+1), 0(i) and M(i). For this purpose, the circuit CPA reads the values (l+a)N(i) and N(p,i+1) from the address locations SL12 and SL18 in the memory MEM, records these respectively in the registers REI and RE2, and then stores the state yl to the output y to equation circuit CR which compares (l+a)N(i) and N(p,i+1) (see equation 3 above) in address location SL19 by overwriting the previous values that were stored there.

In a similar way, the circuit CPA sequentially reads the pairs of values N(p,i+1), N(i)+1; aN(i), 1; N(p,i+1), (l-b)N(i); 0(e), 0(i); 0(i),

0 (m) ; and M(i), fN(i) and stores the results y2 to y7 from the corresponding comparisons (4) to (9) in address locations SL20 to SL25 by

overwrite the previous values stored there.

From the above it follows that the circuit CPA has calculated the results yl to y7 from the above-mentioned expressions (3) to (9) and stored these in the memory MEM. It then examines these results and, in accordance with them, selects the value of N(i+1) indicated by the above table, N(p,i+1) being the value previously calculated.

Finally, the circuit CPA calculates the values of N(i+1) (l+a)N(i+l), (l-b)N(i+l), aN(i+l) and fN(i+l), and stores N(i+1) and these values in the address locations SL10 to SL15 in the memory MEM.

From the above it follows that the detection equipment periodically and

with a periodicity equal to T detects the busy level 0(i) of the data processing machine, and:- that_ the regulation equipment for regulating the workload periodically, also with a periodicity of T, calculates a temporary value N(p,i+1) for the maximum permissible workload which must be able to be processed by the data processing machine. This latter preliminary value N(p,i+1) is directly dependent on and varies in the same direction as the desired maximum busy level 0(M) of the computing machine and the recorded amount M(i) of workload processed by the computing machine. This preliminary value N(p,i+1) is then accepted as the new value N(i+1) until the maximum allowable workload reaches the preliminary value, N(p,i+1),

0(i) and M(i) have predetermined values with regard to the limits Cl-b.)-N(i) , (l+a)N(i); N(i)+1, 0 (m) , 0(e) and fN(i).

Instead of measuring the busy level of the data processing machine in the above-mentioned way, it can also be measured as described in US Pat. No. 3,623,007 or in a similar way.

Claims (19)

1. Data processing system comprising a data processing machine capable of processing a predetermined amount of data, hereinafter called workload, recording equipment for recording the amount of new workload accepted for processing by the data processing machine, detection equipment for periodically detecting the busy level of the data processing machine, as well as regulation equipment for to regulate the workload, which regulation equipment is controlled by the detection equipment to periodically regulate the maximum permissible workload that can be processed by the data processing machine as a function of the detected busy level, characterized in that the regulation equipment to regulate the workload periodically calculates the maximum permissible workload (N(i+ 1)), so that this permissible workload within predetermined limit values becomes directly dependent on and varies in the same direction as the amount of new workload that is accepted for processing and becomes directly dependent on, but varies in the opposite direction to, the detected occupancy level (0 (i) ) . 2. Data processing system according to claim 1, characterized in that the maximum permitted workload (N(i+1)) within the mentioned limits is directly dependent on and varies in the same direction as a desired maximum value (0(M)) for the busy level (0 (in)). 3. Data processing system according to claim 1, characterized in that the regulation equipment for regulation of the workload periodically calculates the maximum permissible workload (N(i+1)) for each fraction (CP1-CP4) or each control period of a regulation period (P (i) ). 4. Data processing system according to one of claims 1 to 3, characterized in that the regulation equipment for regulating the workload calculates the maximum workload (N(i+1)) at the start of each new regulation period (P(i+1)) and for each management period ( CP1-CP4) in this new regulation period (P(i+1)) by calculating a preliminary value (N(p,i+1)) which is directly dependent on and varies in the same direction as the amount of new workload (N (i)) which, on average, is accepted for processing during each of the management periods of a previous regulatory period (P(i)) immediately preceding the new regulatory period (P(i+1)), and in the same direction as the desired maximum value (0(M)) of the occupied level, and which is also directly dependent on and varies in the opposite direction to the occupied level (0(i)) which has been detected for the preceding regulation period (P(i)), and that the regulation equipment for regulation of the workload also calculates the limit values and accepts d a preliminary value (N(p,i+1)) as the maximum permissible workload (N(i+1)) if the mentioned limit values do not. is exceeded. 5. Data processing system according to claim 4, characterized in that the provisional value (N(p,i+1)) is calculated at the start of a new regulation period (P(i+1)) according to the following formula: where: - M(i) is the value of new, average workload accepted for processing in each management period in the previous regulation period (P(i)); - 0(M) is the desired maximum value for the occupied level; - 0(i) is the detected occupancy level for the preceding regulation period (P (i) ); - V is a constant less than 0(i). 6. Data processing system according to claim 4, characterized in that the limit values £(l+a)N(i), (l-b)N(i), N(i); 0 (m) , 0(e); fN (i)j are the limit values of the preliminary value (N(p,i+1)), the limit values of the detected occupied level (0(i)). and the threshold value for the amount of new workload accepted for processing (M(i)). 7. Data processing system according to claim 5 or 6, characterized in that N(i) has the limits (l-b)N(i) and (l+a)N(i), N(i) being the previous maximum permitted workload calculated for each of the control periods to the previous control period (P(i)), while both a and b are positive and less than 1. 8. Data processing system according to claim 5 or 6, characterized in that 0(i) has the limits 0(m) and 0(e), with 0(m) being the minimum value of the occupied level, while 0(e) is an excess occupied level higher than 0 (M). 9. Data processing system according to claim 5 or 6, characterized in that M(i) has a lower limit equal to fN(i), where f is positive and less than 1. 10. Data processing system according to claim 7, 8 or 9, characterized in that N(i+1) = N(i), if 0(i) ^ O(m), and at the same time M(i) ^ fN(i), as N(i+1) is the maximum permitted workload. 11. Data processing system according to claim 7, 8 or 9, characterized in that N(i+1) = (l-b)N(i) simultaneously reaches N(p,i+1) < (l-b)N(i), and 0( i)^> 0(m) or M(i) )> fN(i), where N(i+1) is the maximum permitted workload. 12. Data processing system according to claim 7, 8 or 9, characterized in that N(i+1) = N(p,i+1) simultaneously reaches (l-b)N(i) < N(p,i+1) < (l +a)N(i), and 0(i) > 0(m) or M(i) < fN(i), where N(i+1) is the maximum permitted workload. 13. Data processing system according to claim 7, 8 or 9, characterized in that N(i+1) = (l+a)N(i) simultaneously reaches 0(i) < 0(e), and aN(i) > 1 or N(p,i+1)< N(i+1), and 0(i) > 0(m) or M(i)> fN(i), N(i+1) being the maximum permitted workload. 14. Data processing system according to claim 7, 8 or 9, characterized in that N(i+1) = N(i)+1 simultaneously reaches 0(i) <^ 0(e) , and N(p,i+1) < N(i)+1, and aN(i) < 1, and 0(i) > 0 (m) or M(i) > fN(i) , N(i+1) being the maximum permitted workload. 15. Data processing system according to claim 7, 8 or 9, characterized in that N(i+1) = (1-d)N(p,i+1) simultaneously reaches 0(i) > 0(e), and (l-b) N(i) < N(p,i+1) < (l+a)N(i), where N(i+1) is the maximum permitted workload, and d is positive and less than 1. 16. Data processing system according to claim 7, 8 or 9, characterized in that N(i+1) = (l-b-d)N(i), where N(i+1) is the mentioned maximum permissible workload, and d is positive and less than 1, when at the same time the following inequalities apply: 0(i) ^ 0(e) and (l-b)N(i) < N(p,i+1). 17. Data processing system according to one of claims 1-16, characterized in that the workload is data relating to the operation and/or control of an industrial process equipment. 18. Data processing system according to claim 17, characterized in that the industrial processing equipment is a telecommunications center and that the workload is data relating to the setup, establishment, operation and disconnection of a communication connection as well as necessary management and control data for such a center. 19. Data processing system according to claims 5 and 18, characterized in that - N(p,i+1) is the temporary maximum permitted number of new calls that must be serviced during each management period (CP1-CP4) in the new regulation period (P(i+1)); - M(i) is the average number of new calls accepted for service in each of the management periods of the previous regulation period (P(i) ). - V is the fraction of time in each regulation period that is used to carry out tasks that are independent of call handling.