Data Processing System With Monitoring and Regulation of Processor Free Time

1. Field of the Invention

The present invention relates to a data processing system including a computer having means for ascertaining the occupancy level of the computer and having work load regulating means controlled by the ascertaining means for regulating the computer work load as a function of the ascertained occupancy level.

2. Description of the Prior Art

Such a data processing system is known from U.S. Pat. No. 3,623,007. In this known system the ascertaining means find the occupancy level of the computer by measuring the time interval which elapses between repetitive performances of a particular program and more particularly of the program with the lowest repetition frequency. A drawback of this known system is that when this lowest repetition frequency has a relatively small value a relatively long time interval elaspes between two consecutive measurements of the computer occupancy level and consequently the work load can only be regulated at relatively infrequent moments although prompt action may be required, e.g., in case of an abnormal load.

Another drawback of this known system is that when e.g., the number of programs is modified one may have to adjust the criteria for determining the occupancy level of the computer, since the time interval elapsing between repetitive performances of the above mentioned particular program may then have changed.

An object of the present invention is therefore to provide a data processing system of the above type which does not present the last mentioned drawbacks.

According to the present invention the foregoing and ancillary objects are attained by providing that the computer under normal work load conditions and considered over a predetermined time interval (10 ms) has free time available and that the ascertaining means (FIG. 1) periodically measure said free time after time periods (100 ms) at least equal to the time interval (10 ms) and determine the occupancy level from the measured free time.

Hence, the ascertaining means determine the occupancy level of the computer with a periodicity which is independent of the repetition frequency of programs and which may be made so short that prompt action is possible if required.

In accordance with a preferred embodiment of the invention the present data processing system includes computer ascertaining means adapted to periodically measure the free time of the computer and to compare this free time with a first and a second time value, and work load regulating means which are adapted to limit the work load as a function of the measured free time. Hereby use is made of a work load register each of the positions of which defines a predetermined work load and which may be stepped between a first position defining a maximum work load and a second position defining a minimum work load depending on the result of the above comparison. Each time the measured free time is found to be at least equal to the first time value the work load register is stepped by one step towards the first position until this first position is reached. Each time the measured free time is found to be at most equal to the second time value the work load register is stepped by one step towards the second position until the second position is reached. Finally, when the measured free time is found to lie between the first and second time values the work load register is not stepped.

Another characteristic of the present data processing system is that the ascertaining means (FIG. 1) are able to compare the measured free time with a first and a second predetermined time value at the end of each of the time periods (100 ms), that the work load regulating means (FIG. 2) are adapted to vary the value of said work load between a maximum limit value and a minimum limit value and vice-versa in a stepwise manner, each step being performed at the end of a the time period, and that said work load regulating means at the end of each of the time periods increase said work load if the measured free time is found to be at least equal to said first predetermined time value, decrease the work load if the measured free time is found to be at most equal to the second predetermined time value, and do not vary the work load if said measured free time is found to lie between the second and first predetermined time values.

Still another characteristic of the present data processing system is that the work load regulating means (FIG. 2) at the end of each of the time periods (100 ms) maintain the work load constant if said measured free time is found to lie between the second and first predetermined time values.

In this way the danger of occurrence of an rapidly fluctuating computer load is considerably reduced.

The above-mentioned and other objects and features of the invention will become more apparent and the invention itself will be best understood by referring to the following description of an embodiment taken in conjunction with the accompanying drawings wherein:

FIG. 1 schematically shows a portion of a data processing system designed for use in telecommunication switching equipment TSE and those portions of the processing system, including a memory and a processor;

FIG. 2, which should be arranged below FIG. 1, shows the portion of the processor used for regulating its work load.

The data processing system shown may be of the type disclosed in U.S. Pat. No. 3,557,315 issued Jan. 19, 1971, and which includes telecommunication switching equipment TSE and a memory MEM and a processor PR for a computer.

The memory MEM includes:

a free time counter FTC to accumulate the free computer time FT counted in units of 25 microseconds during a time period of 100 milliseconds;

registers RFT.sub.1 and RFT.sub.2 storing first and second fixed time values FT.sub.1 and FT.sub.2, FT.sub.1 being larger than FT.sub.2. These time values are stored in units of 25 micro-seconds;

a call hopper CH to store data DNC about new calls;

a count register RNCP to store the number NCP of new calls processed by the computer during a time period of 100 milliseconds;

registers RNC.sub.0 to RNC.sub.3 storing fixed maximum numbers NC.sub.0 to NC.sub.3 of new calls allowed to be processed by the computer per time period of 100 milliseconds. Hereby NC.sub.0 > NC.sub.1 > NC.sub.2 > NC.sub.3.

The processor PR includes:

a bistable circuit BS;

a gating circuit GC;

AND-gating arrangements GA.sub.0 to GA.sub.4 and AND-gate G;

an inverter I;

a delay circuit DC providing a delay of 5 microseconds;

a call processing arrangement CPA to collect data DNC about new calls from the switching equipment TSE via a connection c1, to store these data DNC in the call hopper CH via connection c2 and the gating circuit GC, to extract these data DNC from the call hopper CH via the gating circuit GC, the gating arrangement GA.sub.4 and a connection c3, to store the number NCP of new calls processed in the count register RNCP via a connection c4 and the gating circuit GC, and to register the end of a series of tasks performed during a time interval of 10 milliseconds in the bistable circuit BS via a connection c5;

a comparator CR.sub.1 to compare the free computer time FT accumulated in the FTC with the first fixed time value FT.sub.1 stored in the RFT.sub.1. This comparator has an output Y1 which is activated when FT is at least equal to FT.sub.1 ;

a comparator CR.sub.2 to compare the free computer time FT with the second fixed time value FT.sub.2 stored in the register RFT.sub.2. This comparator has an output Y2 which is activated when FT is at most equal to FT.sub.2 ;

a comparator CR.sub.3 to compare one of the numbers of calls NC.sub.0 to NC.sub.3 stored in the RNC.sub.0 to RNC.sub.3 with the number NCP of new calls processed stored in the RNCP. This comparator has an output Y3 which is activated when NCP is at most equal to NCO (or NC.sub.1, or NC.sub.2, or NC.sub.3) and which is connected to an input of the gating arrangement GA.sub.4 via a connection c6;

a work load register WLR which is constituted by a five-position counter. This register WLR has an upcounting input Up, a downcounting input Dn and outputs 0 to 4. The downcounting input Dn is connected to the output Y1 of the comparator CR1 via the AND-gate G which is also controlled by the output 0 of the WLR via the inverter I, whilst the upcounting input Up is directly connected to the output Y2 of the comparator CR.sub.2. The outputs 0 to 3 of the WLR are connected to one input ofthe comparator CR.sub.3 via the AND-gating arrangements GA.sub.0 to GA.sub.3 which are controlled via the gating circuit GC by the registers RNC.sub.0 to RNC.sub.3 respectively. The output 4 of the WLR is an alarm output.

a clock CL.sub.1 generating 2 microseconds output pulses at 10 milliseconds intervals and controlling the call processing arrangement CPA via a connection c7. Via a connection c8 the clock CL.sub.1 also controls a clock CL.sub.2 and is able to reset the bistable circuit BS;

a clock CL.sub.2 generating 2 microseconds output pulses at 100 milliseconds intervals. This clock is controlled by the clock CL.sub.1 via the connection c8 as already mentioned, and is able to reset the RNCP and to enable the comparators CR.sub.1 and CR.sub.2 via a connection c9. It is also able to reset the FTC via a connection c10 and the delay circuit DC;

a clock CL.sub.3 providing 2 microseconds output pulses at 25 microseconds intervals. This clock CL.sub.3 is controlled by the 1-output of the bistable circuit BS and is able to step the free time counter FTC via a connection c11 when this 1-output is activated.

The above-mentioned gating circuit GC is controlled by the call processing arrangement CPA via a control connection cc and is able to temporarily interconnect circuits in the memory MEM and circuits in the processor PR. More particularly, the call processing arrangement CPA controls the gating circuit GC in such a manner that a connection is temporarily established between:

the output c2 of the CPA and the CH when this CPA collects data from the TSE;

the CH and the input c3 of the CPA when this CPA tries to extract data from the CH;

the output c4 of the CPA and the RNCP each time this CPA has processed data relating to a new call;

the RFT.sub.1 and the CR.sub.1, the RFT.sub.2 and the CR.sub.2 and the FTC and the CR.sub.1, CR.sub.2 at the start of each time period of 100 milliseconds;

the RNCP and the CR.sub.3, and the NCO to NC.sub.3 and the CR.sub.3 each time the CPA is trying to extract data from the CH.

The above-described system, including memory MEM and processor PR is adapted to control the establishment, supervision and release of communications through the switching equipment TSE by successively performing higher and lower priority tasks, or clock and base level programs respectively, in the way for instance described in the U.S. Pat. No. 3,557,315 to S. KOBUS et al. The computer starts the execution of these higher priority tasks every 10 milliseconds and when these tasks are finished and time is available it starts the execution of the lower priority tasks. When the computer works under normal load conditions these lower priority tasks are finished before the start of the next following 10 milliseconds time interval. Consequently, per time interval of 10 milliseconds and when working under normal work load conditions the computer normally has some free time available. Obviously, when the work load of the computer increases the available free time decreases and vice-versa. Hence, this free time is a measure of the work load of the computer and therefore of its occupancy level.

In order to be prevented from being overloaded the computer includes occupancy level ascertaining means shown in FIG. 1 to ascertain the occupancy level of the computer by measuring the free time it has available per chosen unit time period of 100 milliseconds, and work load regulating means shown in FIG. 2 to regulate part of the computer work load in function of the ascertained occupancy level. This is described hereinafter.

As mentioned above, every 10 milliseconds the computer starts a series of higher priority tasks and when these are finished and time is available it executes a series of lower priority tasks. One of the higher priority tasks consists in collecting data DNC about new calls. These data DNC are collected by the call processing arrangement CPA from the switching equipment TSE via the connection c1 and stored in the call hopper CH via the connection c2 and the gating circuit GC. One of the lower priority tasks consists in collecting the data DNC about new calls from the call hopper CH via the gating circuit GC, the AND-gating arrangement GA.sub.4 and the connection c3 and in processing these data in succession. Each time the data relating to a new call have been processed the call processing arrangement CPA increments the count register RNCP by one via the connection c4 and the gating circuit GC.

As will become clear later, the last-mentioned data processing task is the above-mentioned part of the work load which will be regulated to prevent the computer from being overloaded. This regulation is done by limiting more or less the number of data DNC regarding new calls allowed to be processed by the computer during a 100 milliseconds time period.

When all the lower priority tasks to be performed during a 10 milliseconds time interval have been finished before the end of this time interval, the call processing arrangement CPA sets the bistable circuit BS to its 1-condition via the connection c5. This bistable circuit BS starts the clock CL.sub.3 which generates output pulses with a periodicity of 25 micro-seconds. These outputs pulses via the connection c10 step the free time counter FTC which thus accumulates this free time in units of 25 microseconds. At the start of the 10 milliseconds time interval following the one under consideration the bistable circuit BS is reset to its 0-condition via the connection c8, but the FTC is maintained in its position. This counter FTC thus accumulates for every 10 milliseconds time interval the free time FT elapsed since the execution of the last lower priority task of the computer.

At the start of every 10 milliseconds time interval defined by the clock CL.sub.1 also the 100 milliseconds clock CL.sub.2 is triggered and each time a 100 milliseconds time period has elapsed the clock CL.sub.2 activates its outputs c9 and c10. Due to the output c9 being activated the operation of the comparators CR.sub.1 and CR.sub.2 is enabled and the register RNCP is reset, whilst the activated output c10 resets the free time counter FTC after the 5 microseconds time delay provided by the delay circuit DC has elapsed.

During the time period the FTC is not yet reset it stores the accummulated free time FT counter during the elapsed 100 milliseconds time interval. The enabled comparator CR.sub.1 compares this accumulated free time FT with the time value FT.sub.1 stored in the RFT.sub.1 in units of 25 microseconds, whilst the comparator CR.sub.2 compares the accumulated free time FT with the time value FT.sub.2 stored in the RFT.sub.2 in units of 25 microseconds. Hereby the time values FT, FT.sub.1 and FT.sub.2 are applied to the comparators CR.sub.1 and CR.sub.2 via the gating circuit CG.

It is clear that three cases are possible:

FT at least equal to FT.sub.1, the output Y1 of the CR.sub.1 being then activated;

FT at most equal to FT.sub.2, the output Y2 of the CR.sub.2 being then activated;

FT comprised between FT.sub.1 and FT.sub.2, none of the outputs Y.sub.1 and Y.sub.2 being then activated.

These cases will be examined hereinafter, but first the work load register WLR which may be in any of five positions 0 to 4 will be considered:

in each of the positions 0 to 3 the WLR indicates the corresponding maximum number of new calls allowed to be processed by the computer during a time period of 100 milliseconds. For instance, when the WLR is in the position 2 the maximum number of new calls allowed to be processed by the computer during a time period of 100 milliseconds is limited to the limit value NC.sub.2. Indeed, when the CPA tries to extract data from the CH this limit value NC.sub.2 is applied to one input of the comparator CR.sub.3 via the gating circuit GC and the then enabled AND-gating arrangement GA.sub.2, whilst the number NCP of calls processed by the computer is applied to the other input of the CR.sub.3 via the AND-gating arrangement GC. Consequently, when the NCP becomes larger than NC.sub.2 the output Y.sub.3 of the CR.sub.3 becomes de-activated and the gating arrangement GA.sub.4 is inhibited thus preventing data from being extracted from the CH. Obviously when the WLR is in the position 0, 1 or 3 the computer is only allowed to process a maximum number of calls equal to NC.sub.0, NC.sub.1 and NC.sub.3 respectively;

the position 4 of the WRL is an alarm position. When the WLR is brought in this position 4 alarm is given and special measures without importance for the present invention are then taken, e.g., the WLR is brought in the position 0 after the reasons for the alarm condition have been removed. This position 4 is not further considered hereinafter.

The above mentioned three cases FT at least equal to FT.sub.1, FT at most equal to FT.sub.2 and FT comprised between FT.sub.1 and FT.sub.2 are now considered in succession.

When at the end of a 100 milliseconds period the accumulated free time FT stored in the FTC is found to be at least equal to the time value FT.sub.1, this means that during this period the occupancy level of the computer was fully normal. Therefore, if the maximum number of new calls allowed to be processed by the computer and indicated by the WR previously was equal to NCO (WLR in the position 0) nothing is done, whereas if this number was previously limited to NC.sub.1 -NC.sub.3 (WLR in one of the positions 1-3) it is increased. Indeed, if the WLR is in the position 0 its output 0 is activated thereby inhibiting the AND-gate G via the inverter I and preventing the WLR from downcounting. On the contrary, if the WLR is in any of the positions 1 to 3 its output 0 is de-activated thereby activating one input of the AND-gate G via the inverter I. Since FT is supposed to be at least equal to FT.sub.1 the output Y.sub.1 of the comparator CR.sub.1 is activated thereby activating the other input of the AND-gate G. Consequently also the down-counting input Dn of the WLR is activated so that this WLR is brought from the position 1-3 in the position 0-2 respectively and that the number of new calls able to be processed by the computer is limited to NC.sub.0 -NC.sub.2 respectively.

When at the end of a 100 milliseconds period the accumulated free time stored in the FTC is found to be at most equal to the time value FT.sub.2, this means that during this period the occupancy level of the computer was abnormal. In this case the maximum number of new calls allowed to be processed by the computer and indicated by the WLR is decreased when the WLR is in any of the positions 0 to 3. Indeed, due to FT being at most equal to FT.sub.2 the output Y.sub.2 of the comparator CR.sub.2 is activated when this comparator is operated thereby also activating the upcounting input Up of the WLR. Consequently, the WLR is then brought from any of the positions 0-3 in the positions 1-4 respectively.

Finally, when the accumulated free time FT stored in the FTC is comprised between the time values FT.sub.2 and FT.sub.1, this means that inside the 100 milliseconds time period during which the FT was measured the work load of the computer was neither fully normal nor abnormal. Therefore, the maximum number of new calls allowed to be processed by the computer and indicated by the WLR is not modified. Indeed, due to FT being comprised between FT.sub.2 and FT.sub.1 none of the outputs Y.sub.1 and Y.sub.2 of the CR.sub.1 and the CR.sub.2 is activated, so that the WLR is maintained in its position.

From the above described examples follows that the ascertaining means shown in FIG. 1 periodically measure the free computer time FT after time periods at least equal to 10 milliseconds (in the examples these time periods are equal to 100 milliseconds) and determine the computer occupancy level from the measured free time. This occupancy level is either fully normal, abnormal or neither fully normal nor abnormal when the measured free time is at least equal to a first time value FT.sub.1, is at most equal to a second time value FT.sub.2, and is comprised between these first and second time values respectively. The work load regulating means shown in FIG. 2 are controlled by the above ascertaining means and regulate part of the computer work load by limiting the number of new calls allowed to be processed by the computer. Each time the ascertaining means detect a normal computer occupancy level this limit value is increased, whereas it is decreased each time the ascertaining means detect an abnormal computer occupancy level. In case the detected computer occupancy level is found to be neither fully normal nor abnormal the above limit value is maintained unchanged.

In the above described examples two different time values FT.sub.1 and FT.sub.2 are used to decide whether or not the number of new calls allowed to be processed by the computer should be modified. This is done in order to substantially prevent this number from being alternately decreased and increased when the measured free computer time alternately decreases and increases. Indeed, when the free computer time measured is found to have decreased below the time value FT.sub.1, the above number is only decreased when the measured free time has decreased below or has become equal to the time value FT.sub.2. On the other hand, when the free computer time measured is found to have increased above the time value FT.sub.2, the above number is only increased when the measured free time has increased above or has become equal to the time value FT.sub.1.

Instead of varying the work load by limiting more or less the number of new calls able to be processed by the computer one may also vary this work load by limiting more or less the number of new calls collected from the switching equipment TSE.

Also, instead of maintaining the computer work load constant when the measured free time FT is found to be comprised between the time values FT.sub.2 and FT.sub.1, one may in this case maintain the work load constant if the previously measured free time FT, i.e., the one measured at the end of the preceding time period of 100 ms, was found to be smaller than FT.sub.1 and increase this work load if the previously measured free time FT was found to be larger than FT.sub.1.

This load increase is continued after each step until the maximum limit work load value NC.sub.0 is reached or the measured free time FT is smaller than the time value FT.sub.2.

While the principles of the invention have been described above in connection with specific apparatus, it is to be clearly understood that this description is made only by way of an example and not as a limitation on the scope of the invention.