In recent years, methods based on modeling have been widely used, as well as experimental methods based on measurements of the parameters of real-time ICS systems on solvable problems. A feature of most known approaches and methods of analyzing the CP is the formal consideration of certain processes and events occurring in the system and characterizing it at various stages of functioning at a logical level. Usually in this case, assumptions of various kinds are accepted, which are caused not by the logic of the ICS, but by the specific nature of the applied mathematical apparatus. Obviously, a different approach is required, which is devoid of this shortcoming. Such an approach can be, for example, operational analysis of computer performance. The method of operational analysis is mainly based on the use of ICS models with queues and requires from an analyst the knowledge of the principles of ICS functioning at a logical level. It is allowing to evaluate the integral performance of the system, which is dependent on a large number of parameters and approximately calculated from the results of measurements without the use of ICS means. Based on the results of approximate calculations, the values of the basic indicators of the ICS are determined and then decisions are taken on the organization of the CP. The peculiarity of this approach (i.e. engineering approach) is to consider from the engineer's point of view the structure of the ICS, the processes and events occurring in it, when it is required to decompose the system; to highlight in its complex structure the backbones and directions for which the IS is distributed; to identify the sets of IS's and parameters that characterize these streams; to establish relationships between threads and their parameters; to develop requirements for the measurement of these parameters; to conduct (by means of measurements) the collection and analysis of statistical data of IS.
The effectiveness of the application of information-computing systems (IСS) in specific conditions depends on the organization of the computing process (CP). With optimal VP organization in regards of some criterion the performance of the ICS - and, consequently, the efficiency - can be increased. Obviously, this explains the heightened interest in the organization of the CP, requiring the conduct of ICS studies operating in a particular mode (singleprogram, multiprogram with fixed or variable number of tasks, time-sharing and real time).
There are a large number of approaches and methods for analyzing the organization of the CP, which is based on the measurement, modeling and analytical study of the values of certain indicators characterizing the work of the ICS. In recent years, methods based on modeling have been widely used, as well as experimental methods based on measurements of the parameters of real-time ICS systems on solvable problems [1-3].
A feature of most known approaches and methods of analyzing the CP is the formal consideration of certain processes and events occurring in the system and characterizing it at various stages of functioning at a logical level, i.e. from the point of view of an analyst or a system programmer, when individual structural links and the distribution of information streams (IS) between the hardware components of the ICS are not taken into account. Usually in this case, assumptions of various kinds are accepted, which are caused not by the logic of the ICS, but by the specific nature of the applied mathematical apparatus. For example, by using methods based on the use of queuing networks as ICS models (queue system models), certain assumptions are made regarding the distribution of the streams entering the system of tasks and the duration of their servicing (decision time), which is caused by the requirement of analytical solvability of the model. In other words, the requirement of analytical solvability of the model imposes a number of additional restrictions [1], which is one of the reasons restraining the use of ICS models with queues.
Another reason for the limited use of these models is that the formal consideration of the CP often leads to undesirable practical consequences - for example, the adoption of incorrect decisions about changing the individual parameters of the CP based on the results of modeling, which made an unrealistic assumption about the distributions of streams entering the system of tasks and the time of their solution. Obviously, a different approach is required, which is devoid of this shortcoming. Such an approach can be, for example, operational analysis of computer performance [3]. The assumptions of this approach are directly related to the logic of the computer, and its conclusions are based on calculations of their performance through the values of operational variables (parameters) measured on a finite time interval, which is allowing to establish simple relationships between parameters and indicators of computer operation. The method of operational analysis is mainly based on the use of ICS models with queues and requires from an analyst the knowledge of the principles of ICS functioning at a logical level. It is allowing to evaluate the integral performance of the system, which is dependent on a large number of parameters and approximately calculated from the results of measurements without the use of ICS means. Based on the results of approximate calculations, the values of the basic indicators of the ICS are determined and then decisions are taken on the organization of the CP [1,3].
In a number of cases, when analyzing CP and studying the properties of ICS for specific applications, researchers and developers of universal and specialized ICS are interested in the differential IS system, which allows to assess the distribution of processes among individual structural components of ICS and, based on this, to make a decision either to improve the organization of the CP or to modernize technical resources. To obtain the differential statistics of IS, it is possible to apply an approach based on modeling and measuring the parameters of ICS information streams at the level of their hardware components (HC), i.e. at the physical level. The peculiarity of this approach (i.e. engineering approach) is to consider from the engineer's point of view the structure of the ICS, the processes and events occurring in it, when it is required to decompose the system; to highlight in its complex structure the backbones and directions for which the IS is distributed; to identify the sets of IS’s and parameters that characterize these streams; to establish relationships between threads and their parameters; to develop requirements for the measurement of these parameters; to conduct (by means of measurements) the collection and analysis of statistical data of IS. Based on the results of this analysis, an assessment is made of the distribution of IS among the HC of the system, which allows to identify “bottlenecks” in the organization of the CP or in the ICS structure and outline ways of eliminating them.
Consider the structure of the generalized information stream (GIS), which conditionally includes all information streams in the modern ICS. By GIS we mean the set of , which are the time-varying sequences of address codes information backbones example, for , instructions , microinstructions , data , conditions , control , interrupt requests
, transmitted from sources h to consumers offline formation over the . The symbol r is used to distinguish elements of a certain sequence of codes (for , the symbol denotes the address of the device ); priority loss
of the selection of sequences for processing, the reliability of transmission and magnitude information
will be characterized by the parameters
parameter
. Since the set
of parameters of the generalized information stream Φ includes subsets
of
parameters of homogeneous information streams
, then for
next expression is viable:
is the set of values of the
(
= for the subset
viable expression is:
Expressions (
of parameters of the stream Φ in the form of a matrix: The rows of the matrix are subsets
of parameters characterizing the corresponding streams and the columns are homogeneous subsets of parameters characterizing the stream Φ. The elements matrix
are parameters of the stream Φ. The elements pertaining to the row are different types of parameters of the homogeneous stream, and the elements related to the column are the same type parameters of heterogeneous streams. The fixation of the matrix
(or its elements)for a given time interval or when performing a certain work on the TDF allows to obtain the necessary data that can be used for statistical analysis of the operation of various HC systems. That includes channels and input-output devices, i.e. to analyze the work of the ICS and organize a CP on it.
The structures considered and the set of parameters of the GIF contain a set of elements by means of which any IS taking place in the ICS can be assigned or presented, i.e. the structural components included into the structure of a specific stream in a certain backbone , and from the set of the parameters of the of the GIF it is possible to select any subset
of parameters characterizing the corresponding backbone
information stream
. Thus, the expressions (
Taking into account (6) and (7), expression (
In the case where the IS is represented as a transmitted blocks of information containing a different number of bytes, the value
for the observation period T can be represented as blocks.
for the stream
is determined from expression
Let’s consider some dependencies between the indicated parameters of the IS. First of all, note that the information capacity
should be understood as the number of transmitted information codes
from the source h to the consumer l defined by the expression Usually, in the process of the ICS functioning, the transfer between its HC is not carrying out continuously, but at arbitrary time intervals : Where E is the number of time intervals; is the time difference between the end and the beginning
,
(
Typically, in the process of functioning of multiprogramm ICS’s which have a non-stationary and random input stream of tasks, there is a situation when IS is transferred between HC at arbitrary moments of time during the considered period T. In other words, the probability that the number n of information codes comes from the source to the consumer at a given time is equal to the probability of their arrival at any other time. It follows that the probability
corresponds to a Poisson distribution [5]: where
is the average value of n for a given T.
If it’s assumed that the process of transfer of information codes in streams , which are distributed between the HC of the considered ICS, is stochastic in nature, then the probability between the arrival of these codes will be less than some predetermined number corresponds to the exponential distribution for which [4,5] can be described as: where
is the average value of the time transmission of information codes
. Then the value of probability of transfer of codes at a given moment.
, It is quite possible that for some IS’s in a certain ICS there is an equable distribution of the probability of will be constant and equal to the The reliability of the transmission of information codes in the streams is determined from expression (14) where is the probability of an error in the ICS during the transmission of the stream
in the ICS is provided by automatic control and correction of the detected errors during the functioning of the system.
The considered IS parameters and some of the dependencies between them, which are represented by
expressions (
For HC, which are simultaneously the source h and the consumer of information streams, an indicator is introduced that characterizes the time
of their occupation by receive-transmit operations equal to the sum of the To analyze the usage of RAM, I/O devices and buses, usually an index is of interest, which is characterizing volume of information codes transmitted by and received in the time
Taking into account (
When researching the organization of the CP on the existing ICS, as well as in the design of new systems, it is necessary to estimate the utilization coefficient for various types of their HC constituents. Usually, this coefficient is determined [] from the ratio of the equipment load to the maximum load that this equipment can withstand, or from the ratio of the time of equipment occupancy to the total time of its operation. Denote by the utilization coefficient of HC, distinguished by the index s, on the interval T. For most HC time spent on transmitting and receiving information codes and
, In turn, and can be defined as time differences respectively between the ends and beginnings ,
of occupation: Taking into account (16)-(19), expression (15) can be described as: and consist of their time intervals (15) (16) (17) (18) (19) (20) (21) boundary:
max is the total time of HC functioning on the considered interval T, the value of has an upper
The integral indicator of the work of HC can be determined with the help of expression (21), and the differential indicators, which may be indicators characterizing the degree of occupancy (usage) of HC by the transfer and the reception of information codes on the interval T, are determined from the where and a lower expressions
, To assess the performance of the I/O channels performing the functions of the IS distributor in modern ICS and being the central link between different types of external devices (ED), RAM and the central processor, an indicator can be used that reflects their actual capacity on the interval T: is the total volume of all information transmitted from and received by it on the interval T; the occupation of by the receive-transmit operations on the given interval. For , as well as for are upper and lower boundaries; ideal case, the value of can be equal to the receive-transmit speed of the information codes the channel , from the RAM and the ED or from the ED and RAM: of i.e.
of . In the
The considered examples clearly show only some possibilities of using the proposed engineering approach in the analysis of CP and studying the properties of ICS for specific applications. However, they do not exhaust all its capabilities, which can be disclosed when considering other dependencies and indicators, introduced in accordance with the objectives of the ICS study.
For the fullest use of the capabilities of the engineering approach, it is necessary to know the structure of the ICS being studied and the peculiarities of its functioning at the HC level, which will make it possible to distinguish the lines and directions of the IS transfer among separate HC in this structure, and also to determine the types of homogeneous IS’s transmitted in these backbones.
Thus, it is possible to formally describe any IS in the IСS and obtain the required expressions for calculating the integral and differential performance of individual HC and ICS in general. The analysis of the expressions, obtained as a result of the formalization of the IS for the parameters, dependencies and indicators considered, allows to determine the basic requirements for the selection of measuring means (MM), by means of which it is possible to obtain differential and integral IS statistics in ICS. First of all, it should be noted that most of these expressions contain parameters as components, characterizing either the number of transmitted information codes, or the time taken to transmit them, or both, i.e. the parameters and are basic and are used to determine the values of other parameters, dependencies and indicators. It follows that one of the main requirements for MM is the ability to measure various values of the parameters and for multiple HC systems when transmitting or receiving . Since the transmission (reception) of between the HC during the operation of the ICS is carried out at random intervals, the measuring device must record the start and end times of transmission (reception) of these codes; calculate the value
on each time interval and accumulate its total sum at all intervals; determine the duration of each interval and accumulate the values as a sum of individual time intervals; provide the possibility of specifying different values of the measurement period T; differentiate the measured parameter values for various information codes
. In addition, the MM should ensure, during the measurement of microscopic events (the parameters of the IS considered above) at the register level, high levels of accuracy, reliability and resolution, without introducing additional interference that could affect the results of the operation of the ICS. They should be convenient and simple in practical application, correctly interact with the ICS in the process of collecting, fixating and accumulating the values of the measured IS parameters.
The autonomous multichannel hardware measurers are the most suitable for the listed requirements, the use of which is most effective for obtaining differential and integral statistics of IS in ICS. However, modern ICS of domestic and foreign production are not equipped with such measurers, so it is necessary to solve a number of issues related to their development and manufacturing.
The proposed engineering approach makes it possible to establish simple relationships between the parameters of IS and the performance of individual HC’s and ICS’s as a whole, as well as to estimate the distribution of these streams between these components, necessary for the analysis of CP’s and the study of ICS properties for specific applications.