Automation and Computers #3, 1969

Z.I. Flikop = &nb= sp; = &nb= sp; = &nb= sp; = UDK 681.34

Collection of Information from Low= Level Computers by a Central Computer in the Controlling Computer Complex.

“Automation and Computer Tech= nology” #3,= 1969.

(Avtomatika I Vichislitelnay Texnik= a, Journal of the Latvian Academy of Science, published in Russian)=

Rece= ived 10/24/1968

This paper discusses a methodology= of defining a frequency of data collection by a Central computer in the Contro= lling Computer Complex. An equation allowing defining a size of a Central computer memory that is needed to save collected information is also proposed.<= /o:p>

One, even very power, computer is incapable to control a complex multistage technological process. This is why to control such processes computer compl= exes, consisting of a powerful Central computers and connected to them computers = with limited productivity, are used [1]. This approach allows consider a control complex as a two level system. A Central computer is located on the upper l= evel and controlling computers are located on the low level.

Central computer manage a whole technological process, while low level computers ha= ve limited tasks of controlling different technological stages. To be able fun= ction efficiently a Central computer periodically collects information from low l= evel computers, develops controlling commands with consideration of a global technological task and downloads them into low level computers for the execution.

Usually Central computer collects only information that either is needed for the gl= obal technological control or is needed for the readjustment parameters controll= ed by low level computers. This collection is organized in such a way, that frequency of a collection is defined depending on frequency of intersection= by controlled parameters some predefined thresholds.

This approach allows limiting a volume of information transmitted from low level computers to a Central computer. As a result, reaction of a Central compute= r to deviation of a technological process from preplanned execution can be shortened.

This vision of functioning of the Controlling Computer Complex allows us to formalize three problems related to the collection of information. Namely, = 1) which technological parameters must be controlled; 2) how often low level computers have to be interrogated; 3) what size of memory has to be dedicat= ed to each low level computer to store collected information.

In this paper we will try to find solutions for the two last problems.

Let’s start with the second problem. Each low level computer controls some specific and predefined technological parameters. We will consider = that a value of a controlled parameter changes if deviation of it from previously measured value exceeds a predefined threshold. These deviations occur at ra= ndom moments of time. Fluctuations of technological parameters are caused by multiple very often independent factors. Considering Palm’s theorem a= nd additions made by Khinchin, deviations of j-th (= j=3D1,2,..) controlled parameter represent a simple process= [2] with intensity _<=
/sub>.

If information collected from k-th (k=3D1= ,2,…) computer consists of results of deviations= of m (m=3D1,2,…) parameters, then combined flow of data from this computer will be represent= ed also as a simple process with total intensity

        =             &nb= sp;            =            _{<=
/sub>        =
            &nb=
sp;            =
            &nb=
sp;(1)}

(Since in our analysis we consider data flow from a general computer, we will omit index k).

Frequency f=3D1/t of data collection shou= ld be defined upon condition that during time t between two sequential collections, not more then n (n < m) parameters will deviate with probability _{<=
/sub> beyond pred=
efined
thresholds.}

We can also define probability <= sub><= /sub> that during= time t will deviate beyond predefined thresholds more then n paramete= rs

        =             &nb= sp;            =    _{<=
/sub>        =
            &nb=
sp;(2)}

This equation allows us define a minimal frequency of data collection _{<=
/sub> upon condit=
ion:_<=
/sub>; where _{<=
/sub> is a maximal
acceptable probability that more then n parameters
will deviate beyond predefined thresholds.}}

Now let’s consider a third problem. Information collected from low level computers is stored in the corresponding files of a Central computer. The n= umber of memory cells z dedicated to = each file has to be sufficient to store a file of a maximal expected length; however, to save memory space this number should be as small, as possible.<= span style=3D'mso-spacerun:yes'> The length of a file depends on ho= w many parameters deviate during a collection cycle t. Considering stochastic nature of controlled parameters behav= ior, number of deviated parameters is random. We can define a number of memory c= ells z by considering the following condition: the probability _{<=
/sub> that number=
of
dedicated to some file cells z will be insufficient should not exceed the
maximal excepted probability_<=
/sub>,}

where        = _{<=
/sub>        =
            &nb=
sp;            =
            &nb=
sp;           (3)}

In this case, if we can accept some_{<=
/sub>, then we can find _{<=
/sub> that satisf=
ies_<=
/sub>.}}

References

1. Golubev-Novozilov Yu.S. Multicomputer Complexes. Moscow, “Soviet Radio”, 1967.

2. Khinchin A. Ya. Queuing Theory. Moscow, Fizmatgis, 1963.

3. Gnedenko B.V. Probability Theory. Moscow, Fizmatgis, 1961.