=Paper=
{{Paper
|id=Vol-2923/paper21
|storemode=property
|title=Designing Networked High-Load Distributed Computing Web Systems for Intensive Processing of Information Data Flows
|pdfUrl=https://ceur-ws.org/Vol-2923/paper21.pdf
|volume=Vol-2923
|authors=Mykola Pasieka,Vasyl Sheketa,Iryna Halushchak,Svitlana Verbeshchuk,Mykhailo Yasinskyi
|dblpUrl=https://dblp.org/rec/conf/cpits/PasiekaSHVY21
}}
==Designing Networked High-Load Distributed Computing Web Systems for Intensive Processing of Information Data Flows==
https://ceur-ws.org/Vol-2923/paper21.pdf
Designing Networked High-Load Distributed Computing Web
Systems for Intensive Processing of Information Data Flows
Mykola Pasiekaa, Vasyl Sheketaa, Iryna Halushchakb, Svitlana Verbeshchukb,
and Mykhailo Yasinskyic
a
National Technical University of Oil and Gas, 15 Karpatska str., Ivano-Frankivsk, 76068, Ukraine
b
Vasyl Stefanyk Precarpathian National University, 57 Shevchenko str., Ivano-Frankivsk, 76000, Ukraine
c
Ukraine Academy of Printing, 19 Pid Goloskom str., Lviv, 79020, Ukraine
Abstract
The grid model and algorithm of the high-load distributed fault-tolerant web system
architecture are improved. The main difference lies in the possibility of aggregation and the
sharing of a large number of heterogeneous computing resource sets to process information
and data flows distributed between different geographical areas. Compared with the
traditional method, the proposed model and algorithm allow the use of other network
resources connected to the functional network. In the traditional method, these resources are
not available within a single computing node on an independent computing platform. Further
developed innovative methods for building high-load fault-tolerant distributed cluster
software systems, which generally greatly increase the total amount of information flow in
the node communication system. Therefore, this method is suitable for distributed fault-
tolerant software systems in which the operational loss of information data flow is
unacceptable. Further developed an algorithm for automatic load control on an independent
computer platform for information data flow to achieve effective expansion (clustering) of
the software system.
Keywords 1
Software system, algorithm, architecture, web system, network.
1. Introduction
Computing node load balancing ensures an even load of hardware and software systems on
independent computing platforms. The software system that is balancing the computational load must
automatically decide on which node to perform the computation for the new information task. Thus, the
main task of balancing is the process of effectively supporting the process of moving (migrating) part or
all of the calculations from the most loaded computing nodes to the less loaded ones, loading input
factors i of i = 1, ..., r; uj are weighing output parameters, weighing output parameters j of j = 1, ..., s.
When solving the problem of maximizing efficiency standards, there is an urgent problem, that is,
there is a share in the distribution of two linear aggregate values [46].
In addition, the problem of maximizing the efficiency of web systems is called linear particle
programming. At the same time, the possibility of converting linear particle programming into linear
programming problems is very high:
βππ=1 π£π π₯π
π0 = β πππ!, (1)
βπ π=1 π’π π¦π
βπ π£π₯
with βπ π=1 π’π π¦π β₯ 1 or all modules m = 1, 2, β¦, n ; uj ο³ο 0, j = 1, 2, ..., s; vi ο³ο 0,
π=1 π π
Cybersecurity Providing in Information and Telecommunication Systems, January 28, 2021, Kyiv, Ukraine
EMAIL: pms.mykola@gmail.com (A.1); vasylsheketa@gmail.com (A.2); pasyekanm@gmail.com (B.3); iryna.galushchak@gmail.com
(B.4); leuro@list.ru (C.5)
ORCID: 0000-0002-3058-6650 (A.1); 0000-0002-1318-4895 (A.2); 0000-0002-4824-2370 (B.3); 0000-0001-9445-6348 (B.4); 0000-0002-
4824-2370 (C.5)
Β©οΈ 2021 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
192
i = 1, 2, ..., r; yj are expression j and the output parameter of the tested computing module (node); xim is
expression i that input factor m of that computing node with i = 1, ..., r and m = 1, ..., n; yjm expression j
that input factor m of that computing node with i = 1, ..., r and m = 1, ..., n; vi β input factor weights i
with i = 1, ..., r; uj are weighing the output parameter j with j = 1, ..., s.
For further research, we use a complex fractional programming theory algorithm (a more
traditional linear programming algorithm with less resource occupancy) to modify the non-linear
problem of efficiency standard optimization. Therefore, the use of linear optimization methods can
simplify the complex problems of efficiency standard optimization into linear problems [28]. In order
to obtain the value of the efficiency standards of all computing modules (nodes), it is necessary to
solve the problem of maximizing each module involving scientific research separately. In this case,
replace the vectors xim and yim [24] with the input and output parameter configuration files of the
studied computing module, respectively. In another task of maximizing efficiency standards, they
remain the same for each computing node [18, 30, 35, 45, 50]. In this task, restrictions are imposed to
ensure that qualitative efficiency values are found in the region between zero and one (e0 ο[0,1]). The
basic properties of the model are formed by the efficiency limit criterion. Its objective function
proportionally attempts to increase the observed output parameters of the calculation module to the
limit of the efficiency standard. In the mathematical decoupling of a certain function, after applying
the duality theorem to it, an equivalent form or so-called hug form is formed. The dual principle of
linear programming is used to prove that for each directly optimized linear model considered, the
efficiency criterion has a corresponding dual optimized linear model, and when one model decision
includes all expressions, the other model decision also includes all expressions. formula. Therefore,
the initial optimization model will be regarded as a direct linear programming problem. Thus, using
the dual method in our scientific research, we are guaranteed to obtain minimization of the weighted
sum of input parameters in one normalized output parameter. As a result of these manipulations, we
obtain a mathematical formulation of the linear programming model using the dual method, which is
performed as follows:
π
min π0 = β π‘π π₯π, (2)
π=1
with g0 is the value of the efficiency criterion of the studied computing node; xi is an expression i of
that computing node input factor with i = 1, ..., r; ti is variable weighting factors.
The linear combination determines the potential reference group, which is used to measure the
efficiency standards of the computing nodes involved in the research. Therefore, according to the
optimization task of minimizing the objective function (2), the proposed method selects such a
reference group in which the efficiency standard of the computing node involved does not seem to be
effective enough [10, 37, 40].
Therefore, the mathematical problem can be explained as for the computing node to be studied, the
minimum efficiency standard for determining the input parameter G0 is compared with the weighted
probability of the comparison unit. The weighted combination of any input parameter and output
parameter does not subtract the output parameter, any The weighted integral combination of the input
factors is G0 times larger than the input factors.
2. Parametric Network Model with Time Index to Calculate Load
The basic principles of the operation and function of the cloud computing network model on an
independent computing platform are very different from the traditional serial and parallel models. The
main feature of the cloud computing network model is the ability to aggregate and collectively use a
large information set of heterogeneous computing key data streams distributed in geographic
locations. This method has significant advantages. For example, when the developed software system
needs information resources that are not available in a computing node, it can get it to other
computing nodes connected to the cloud network. [5, 8, 9, 11, 23, 27].
193
Suppose that m computing resources are available in the network cloud environment, and there is a
task flow t distribution system, which evenly distributes information tasks jο t among the available
resources. In the framework of combining this developed software system with cloud technology,
each userβs information task can be divided into certain computational actions k ο j.
When setting the task, determine the processing time dj to obtain the corresponding node. Each
information task of user j and all corresponding actions k ο j are in the cloud grid and at a certain
point rj. Therefore, there is a cloud network running in online mode with custom rj. A large number of
these tasks have unknown values beforehand. Once a custom task to be processed arrives on the cloud
network, plan to search for and allocate the necessary resources to run it. Assume that the final
allocation result of the calculation task S for each action k ο j is to process Ck(S) within a certain
period. Itβs required. Therefore, the processing speed of the userβs computing task in the cloud
network j will not be lower than the processing speed in the specific period determined by the
expression:
πΆπ (π) = max πΆπ (π), (3)
πβπ
with Ck(S) is the maximum processing time of the userβs task; k is action; j is mission.
Let us define pj as the processing time of the userβs task k ο j. So, the processing time of a userβs
task in a cloud network can be calculated as follows:
ππ = πΆπ (π) β πππ(πΆπ (π) β ππ ), (4)
πβπ
with pj is user task processing time; Ck(S) is execution time; Cj(S) is maximum execution time; pk is
decision implementation time.
The received point value allows estimation of the properties of the cloud network computing
environment. To analyze the quality of cloud services provided by the network computing
environment, you can use the value of the maximum delay in the user task:
πΏπππ₯ = max(πΆπ (π) β ππ ), (5)
πβπ‘
with Lmax is a maximum delay of the userβs task; Cj(S) is maximum processing time; dj is a time of
uncompleted user tasks.
Thus, to optimize distributed cloud resources it is necessary to minimize the value of Lmax, and equal to
the previous value of the parameter you can also use the value of Tj, which will determine for us the delay
time of user tasks that are calculated:
π β π‘ β πΆπ > ππ , (6)
with t is information computing resources are available; dj is a time of uncompleted user tasks; Cj is time
to complete one task; k is action; j is mission.
Therefore, the indicator provides information about the number of outstanding calculation requests of
users who have entered the cloud network for processing. The computing resource consumption of the
RCk cloud network is a specific subtask of computing, which is defined as the product of the
corresponding solution time multiplied by the number of resources used in the network:
π
πΆπ = ππ Γ ππ , (7)
with RCk is total consumption of network computing resources; pk is decision implementation time;
mk is the number of resources used.
Using the total use value of cloud computing resources used, you can calculate the total cost of
available necessary information resources U:
π
πΆ(π)
π= , (8)
π Γ (max πΆπ (π ) β min(πΆπ (π) β ππ )
πβπ‘ πβπ‘
with U is the amount of use of available information resources; m is resource quantity; pj is processing
time of the jth problem; RC(S) β time of total consumption; Cj(S) is the time of the jth task.
194
Thus using the formed standard that characterizes the optimal use of computing resources in a cloud
distributed network through a certain value. The following situations usually occur in the process of data
processing of information requests in a cloud distributed network: the software system crashes when
processing highly loaded user tasks [6, 7, 26, 43, 44, 47].
Based on the mathematics outlined above, we conclude that the userβs task of processing an
information request must be performed several times before it can be completed. Since users and
administrators may have different (or even conflicting) requirements for cloud-based web systems, it is
difficult to find a common metric that everyone is satisfied with. From the perspective of a distributed
software system developed by users using cloud technology, the following metrics can be distinguished
between average query response time (Average Response Time, ART) and average query latency
(Average Wait Time, AWT):
1
π΄π
π = β(πΆπ (π)) , (9)
|π‘|
πβπ‘
th
with Cj(S) is the time of the j task; pj is the time of realization of the jth task; t is available resources.
1
π΄ππ = β(πΆπ (π) β ππ ) , (10)
|π‘|
πβπ‘
with Cj(S) is the time of execution of the jth task; t is available resources.
The ART parameter value represents the average response time to user information requests, that is, the
speed of processing user tasks. In addition, the value of the parameter AWT is very important for
developers of action algorithms in small information tasks and their query processing. Therefore, the best
and simple way to measure the efficiency of fair use of information and hardware resources is to calculate
the deviation of the expected weighted average time for query processing:
1 πΆπ (π) β π π 2
π΄πππ· = β(πΆπ (π) β π π )2 β (β ) , (11)
|π‘| β |π‘|
πβπ‘ πβπ‘
with AWTD is the deviation of the average waiting time for processing the userβs information request; t is
available resources; Cj(S) is the time of execution of the jth task; pj is the time of realization of the jth task.
The AWTD must be minimized to get the best results in the cloud network. In the modern cloud
network web system, the ability to complete the processing of a given number of tasks is more important
than the acceleration of the distributed high-productivity web system obtained by using this processing
method. It should be noted that, compared with the userβs information task executed in a parallel system
of the traditional architecture, the userβs information task executed in a cloud network environment may
have a quite complex architecture. For example, the information flow of a user task has a more complex
logical structure than the corresponding task package [25]. The use of cloud grid needs to change
concepts such as errors in the web service development program system designed based on the grid
model. Once there is a situation where the user information task cannot be successfully executed and
completed, an error message will be generated.
Using a cloud mesh requires changing such notions as errors of a developed program system of web
services which is designed on a mesh model, generates error messages as soon as there comes a situation
when it is impossible to successfully execute and finish a userβs information task. For example, if suitable
resources cannot be found to perform information calculation or information calculation cannot be
completed, the developed software system using cloud technology may malfunction.
Using the concept of fault tolerance in software systems using cloud technology, we can define it as
the main possibility of increasing the time delay of software and hardware errors until the work of
processing user requests cannot be completed. An indicator of completed user request processing in a
software system that uses cloud technology βworkload completionβ, which is formed by the ratio of
completed user tasks to the total number of requests set by the cloud network scheduler:
β π β π‘ β (π πππππππ‘ππ)
ππΆ = , (12)
|π‘|
195
with WC is the indicator of completed user requests processing; t is available resources; j is the task.
This indicator allows defining the main limitations of the cloud network software system, the
maximization of which may be the main goal. However, from the perspective of using free
information and hardware resources, it also has some key limitations, because user tasks with a small
number of computing operations have a significant impact on the change of this value [16].
Task completion calculates the number of completed operations relative to the total number of
operations performed. These have been implemented in the cloud software system for user task
assignments.
Through this indicator, you can define the main limitations of the cloud network software system,
and maximizing it can be the main goal. However, from the perspective of using free information and
hardware resources, it also has some key limitations, because the tasks of users with fewer
computational actions have a greater impact on the change of the value.
Task completion calculates the number of completed operations relative to the total number of
operations performed. These have been implemented in the cloud software system for user task
assignment:
β π β π‘ β π β π β (π πππππππ‘ππ)
ππΆ = , (13)
β π β π‘ |π|
with TC is the index of completed actions for processing user tasks; t is available resources; k is the
action; j is the task.
It is also worth considering the concept of completing unlocked operations from user tasks to handle
the completion of enabled tasks, i.e., they can only be executed when all corresponding dependencies of a
given sequence of operations will be executed by a software system using cloud technology:
β π β π‘ β π β π β (π πππππππ‘ππ)
πΈππΆ = , (14)
β π β π‘ β π β π β (π πππππππ)
with ETC is the integral indicator of completed unlocked actions relative to user tasks; t is available
resources; k is the action; j is the task.
Therefore, when developing a program system that is very different from traditional serial and parallel
models, we have considered the main principles of cloud network model operation. The main difference
between them is that they can gather and share a large number of heterogeneous information resources
distributed among geographically independent computing independent platforms. This architecture
method of using cloud technology to develop program systems has brought considerable advantages and
financial benefits. For example, when a program web system needs a large number of information
resources that cannot be accessed within a computing node, it can combine them Connect to another
connected node to the cloud network.
3. Models and Algorithms for Optimization of Distributed High-Performance
Software Web System Architecture
Any web system or cloud service that serves many users with a priori high load. However, high-load
distributed software web systems cannot effectively apply models, methods, and algorithms. These
models, methods, and algorithms are used as the basic methods for developing ordinary websites [4, 42].
If there is no proper way to use cloud technology to optimize the system development architecture of the
program, then a large number of users and corresponding calculations will lead to complexity in timely
maintenance [1, 17, 21, 29]. Now, all models, methods, and algorithms used to develop distributed fault-
tolerant web system architecture have been developed and used through recognized information
technology [19]. The method of sending information packets should use an algorithm to calculate the
checksum of the control protocol transmission of the packet, which allows you to use the checksum of the
header of this packet to display the checksum of the control protocol transmission of the data of the
packet [2, 12, 20, 22, 36].
196
Consider the checksum recalculation algorithm, which will use the output packet checksum and the
input packet checksum to calculate the checksum. Therefore, if the clipboard is divided into two
information parts, the checksum of the entire buffer can be expressed linearly by the checksum of its
parts. The typical length of the packet header protocol control transmission is usually several times
shorter than the base of the entire information packet. Therefore, the calculation amount of the entire
software web system resources is reduced by n times, and the redirection cost of client requests from one
node to another can be significantly reduced. In the next technical stage, user requests will be passed to
the computing node for processing. The architecture of computing nodes used to process user information
requests should be organized in a hierarchical structure. At the bottom of this hierarchy are important
computing modules used to process user requests. The failure of these modules will temporarily slow
down the overall operating speed of the web system, but will not paralyze the entire system. If the lower
computing module does not work, the cloud network web system allows the corresponding computing
module of other nodes to be used. However, in the calculation process of this organization, it is obvious
that the communication level between nodes when processing user requests should be perfect.
The organizational characteristics of the architecture of computing nodes for processing high user
requests require that these modules have a clear hierarchical structure. The essence is that each computing
module of a node must be isolated from each other as much as possible, and exchange information at the
message level of the public system while relying on the information. Therefore, the design of the software
web system should be developed according to the following principles: βThe fault is a specific part of the
web service work and should be controlled at the system message exchange level.β The architecture of
the fault hierarchy is that any web system is composed of computing modules, and when the descendant
modules refuse, the next one has a series of operations. The organizational features of the computational
node architecture for processing high user requests require that these modules have a clear hierarchical
structure. The essence is that each nodeβs computing module must be as isolated as possible from each
other and exchange information at the level of the public system message, relying on the information.
Therefore, the design of the web system software should be developed according to the following rules:
βThe fault is a specific part of the website operation and should be controlled at the system message
exchange level.β The architecture of the fault hierarchy is that each network system consists of computing
modules, and when the child modules refuse, the next one performs a series of operations (Fig. 1).
Figure 1: Model of failure hierarchy processing in software web systems
The selection of the computing module used to execute user requests is not only based on its load but
also based on information about the number of failures in other modules of the cloud node. This
197
algorithm is necessary in order not to create a packet for processing the instantaneous data flow from a
very busy node module to a less busy node module. As a result of these manipulations, a complex cloud
network for exchanging computing information arrays between corresponding modules of computing
nodes is formed. However, the main purpose of building a high-load distributed software web system is
to create such a software system without using cloud technology, which will not fail during industrial use
and will be able to work and process various types for as long as possible information. Challenges and
own mistakes [41]. After the information message flow has been redirected, the computing module will
not stop working.
Therefore, we can emphasize the strategies that can be implemented in each computing module [15].
In the first strategy, we will understand that the load on the computing module is large. In this case, the
module will execute the entire list of user tasks that enter or remain in it for processing, perform a master
restart, and generate an information message to the module higher in the hierarchy, and prepare to work
[48]. It is assumed that the cloud network module of the computing node and the main load balancer
constantly exchange information messages about readiness, and the module will receive a new data
stream for processing in each subsequent data stream iteration of load distribution on the node. The
snapshot/restore method of data replication is usually used in high-load distributed web systems.
Therefore, a hybrid system model with a centrally balanced node for load balancing and cloud network
communication between computing nodes allows you to simply add more nodes by applying the cluster
structure. The high-level communication between computing nodes brings problems to the data cloning
method because, in the event of a failure of one of the nodes, it is necessary to recalculate the overall
performance of the entire web system. Obviously, in this case, we can divide the problems related to
computing node failures into two condition groups. The first group of software and hardware technical
issues. If an error occurs in the software of the developed web system, because the redistribution of the
calculation amount occurs at the level of the calculation module, if the calculation module at the top of
the hierarchy is designed correctly, the system will automatically and quickly restore its performance and
not get the level connection error [51, 53, 54].
Then it will warn other computing nodes about the error and redirect the computing information data
flow to other modules. Therefore, in this case, there is no duplication of specific computing nodes on an
independent platform, but only the launch of a standard instance. In order to significantly reduce the risk
of losing a large number of user requests, the balanced computing node must include a buffer that will act
as a "retry" when a node is physically unavailable. Therefore, the use of replication methods can provide
the creation of typical computing nodes and at the same time restore the performance of the web system.
However, when using cloud technology to develop program systems, there is a whole class of distributed
algorithms, in which certain structures of information data streams (arrays, records) depend on the total
number of interactive processes at the same time. The processing is performed by different computing
nodes on independent platforms. An example of the application of this algorithm is a protocol for making
coordinated decisions [34]. When the topology of the distributed and highly advantageous computing
web system is modified, the structure of the data stream and its processing algorithm will also change.
Therefore, there is an urgent need to modify the behavior scheme of the simulation model used, that is,
to modify the algorithm and structure of the data flow describing the behavior of the corresponding
object.
4. The Realization of the Data Stream Processing Method to Restore the
Computing Node
In the development of a high-load distributed web system, the architect consists of multiple computing
nodes, and they are an instance of a single computing system, thus forming a βcluster.β However, the
modified data stream processing algorithm not only needs to be separated at the physical level but also
needs to be separated in the program. Traditionally, the software of the web system is divided into logical
modules according to its functional purpose [33]. By designing a software web system, architects need to
design so that the computing modules in the structure have as little logic as possible because the main
198
task of redistributing the load is to support system balance. An equally important requirement for the
software architecture of web systems is to avoid strong connections. The strong connectivity between
computing modules is that one of the modules is in contact with many other modules on the independent
platform.
In other words, if a failure occurs, a large number of modules will send a message to the service
module to notify other nodes that they need to be replaced temporarily. In order to effectively overcome
possible calculation problems, you can use the following methods to restore the functions of the
developed software web system using cloud technology: The clone_args method stores the values of the
parameters that have been sent to the failed calculation module [3, 14, 38, 39, 49].
Since in this case, the module did not respond in a timely manner, it is classified as a failure, so the
parent method of the module must obtain appropriate information about the parameters that have been
sent to the child method. However, the direct function of the clone_args method does not stop there,
especially the indirect attributes they involve are:
ο· Calculate the current integral processing time value of the module or the faulty module method.
ο· Fixed time for processing failures of computing nodes in the cloud web system.
ο· Calculate the start-up frequency of the module, which allows you to determine the demand for
this module. Therefore, when distributing the load between nodes, you need to allocate the
necessary software and hardware resources more efficiently.
The reporting method allows the service module to send statistical data about the information flow
of processing all child nodes at certain intervals. Moreover, its use helps to automatically predict the
degradation dynamics of the developed software web system. Therefore, the reporting method can not
only be used to notify the statistical indicators in the plan but also can be used when one or more sub-
computing modules fail.
The restart method provides the use of cloud technology for the developed web system program by
restarting the computing module in the event of a failure. The main purpose of restarting is to avoid
disruptive incoming data flow.
The callback method allows information messages about the successful or failed restart of the
compute node to be sent to the parent module. This method is passed to the calculation module as a
parameter. After starting or restarting the software web system that provides the βcallback method,β it
should be checked whether the module is running stably. After analyzing the characteristics of errors
that may occur in the calculation module of the developed program web system, we will notice
several main reasons: errors in the program code, and high load in the problematic data flow and
nodes.
The feedback method uses cloud technology to provide communication between the computing
nodes of the developed program web system and is independent of each other, which is why it can be
surrounded by the microservices of many web systems, and these microservices must exchange
information flows with them. For this, you should of course notify the service computing module,
which is at the top of the corresponding hierarchy. This is why each computing module adopts a
feedback method that knows how to contact it and can redistribute the data flow from modules that
have failed or are heavier than working nodes. The revised calculation model of the recovery method
(the failed module) is shown in Fig. 2 [32].
This modified algorithm only implements a service module for processing data streams. Compared
with traditional computing modules, the number of operations provided by the service modules is much
smaller, and positive results can be obtained, which directly affects their work. In this case, the service
module acts as βa stable part of the software system in a computing pointβ on a node on an independent
computing platform.
Therefore, when developing a program web system, it is important to use methods to provide the
smallest loopholes in this computing module.
These calculation methods will continuously notify the results of the work for a period, or
unexpectedly notify the results of the work when leaving the work state. The service module has working
information about all the software web systems developed. The less time the information flow about the
199
work results of the developed program web system is updated, the more accurately and effectively the
necessary computing nodes can be defined to transfer its load.
Figure 2: Load redistribution algorithm when a computing node fails in a software web system
Packing, state transfer, and data streaming. In the web system of the developed program using cloud
technology, once a consensus is reached between the computing nodes that authorize and accept part of
the computing work, it is necessary to prepare a package of raw data for the new node. And only after
this, a connection can be established between the user module that failed to send the user request to
another working module in another computing node. In addition, the service module will also send a
message stating that the computing module is damagedββtemporarily unavailable to process new
requests for data streams,β and then perform the standard restart process. Due to the corresponding restart
method, this process is implemented in each computing module and is also connected to the service
module on the computing node through a feedback method. Once the software system receives the
message that the computing module has failed, clone_args will copy all the input data streams sent for
processing and a set of indirect data about the current state of the system. In this case, there is almost no
time to make management decisions, so the service module may rather try to acquire a new computing
node while copying the entire information call thread through the clone_args method and directing it to
the service module.
When analyzing the sequence of the corresponding actions, it is obvious that it is necessary to form a
list of processing calls, which are now waiting for other parent computing modules or their methods.
However, this list is a kind of queue for processing user requests processed by the new work calculation
module from the created queue. Amazon Simple Queue Service works according to the same principle-a
service that can quickly receive user data flow queues. The service node combines this data and links the
corresponding modules and their status messages for storage and processing. Once the compute node is
available for information. In order to ensure the reliability of data streaming, it is necessary to divide the
specific set of these data into appropriate fragments and add a header with a sequence number to each
fragment. The fragments of the information data stream obtained in this way form a segment. In the next
step of data processing, each segment enters a packet, and then reaches the recipientβs computing node
using the transmission information protocol. After the packet is delivered to the recipientβs computing
node, the correctness of the received information flow in this segment is checked by means of checksum
200
recalculation, and it is automatically determined that the previous information flow segment has also been
successfully received Fig. 3.
Figure 3: Packing, state transition, and data information flow model
At this stage, the recipientβs computing node sends a request for information about new data packets or
repeated transmissions of previous data packets to the senderβs node. This operation algorithm ensures
that all previous data packets in the data stream sequence have been successfully received. In the
improved model developed, the computing modules are isolated from each other and do not have
information about the overall state of the web system when establishing connections between them and
receiving messages about their status. For the interaction of computing modules, we use the
asynchronous exchange of information messages, where each corresponding module has its information
message turn. Therefore, the computing module that sent the message will wait for the next notification
about its delivery, otherwise, the receiver will ignore the message.
5. Conclusions
The system analysis performed considers the basic principles of distributed web system design and
the technology most commonly used by the architects of the program system in the design. We also
concluded that redundancyβan indispensable attribute in most web system development, redundancy
is highly loadable relative to user requests. The main criterion for web system development is
scalability. When the operation requires a large number of computing resources, its scalability will be
used, which will greatly reduce the performance of the system and require it to increase the overall
capacity. In addition, methods for estimating the reliability and fault tolerance of the developed
program web system are studied because any programming system should be objectively monitored
and predicted during work. Consider the basic aspects of the operation of distributed fail-safe
software web systems because they have management problems and the ability to recover from
failures during highly advantageous operations. By analyzing the technology used in the development
of the software web system in the research, we improved the method and comprehensively analyzed
the balance of the computing load based on this. This is the main task of designing the software
201
distributed fail-safe web system. In order to solve the problem of effective working of the developed
software system, we used the algebraic method of optimizing the calculation of load balance on the
node and its distributed network model, thereby creating a hybrid mechanism for information flow
transmission. It can be considered to provide theoretical and practical foundations for the effective
functions of the fault hierarchy mechanism. These foundations can provide effective calculation of the
overall computing load of the program modules and nodes. By analyzing the high-load distributed
fault-tolerant software web system, it can be determined that its basic criterion is the efficiency of
processing an information data flow, which is usually calculated as the partial division of the sum of
all output parameters and the sum. All input data streams. For each specific computing module or
node, determine its efficiency value. The comparison of computational node efficiency is carried out
using linear programming methods and using different basic models and their variants at the same
time. It is recommended to determine the number of computing modules or nodes involved by
constructing efficiency boundary criteria and all other conditions (its invalid criteria). The improved
function of the network model is very different from the usual serial and parallel models. The main
difference lies in the ability to gather and share important heterogeneous computing resource sets to
develop geographically distributed information flows. In many cases, the development of software
web systems brings considerable advantages and requires other resources that are not available in a
computing node, and these resources can be obtained from other nodes connected to the functional
grid. The research improves the algorithm for transferring computing load without using redundant
resources, and proposes an improved interaction model between computing modules to handle the
information flow within a single node, and proposes software developed as a whole web system. In
addition, the technology to determine the cause of the failure of the computing module and the node is
considered, and two basic problems related to the failure of the node are emphasized namely, software
and hardware problems. Modified the computer process control model used for the development of
information data flow, in which modules are isolated from each other and have no common state, but
information communication can be established between them, and notifications about their status can
be established. For the interaction of computing modules, asynchronous exchange of information
messages is used, in which each module organizes its message wheel. The module sends a reference
message, waits for confirmation of the corresponding delivery message, but does not wait for the
receiver of the message to ignore it. An important aspect of this research is that the free computing
resources of the developed program web system can be found within the scope of its functions.
6. References
[1] S. Aleti, BjΓΆrnander, L. Grunske, I. Meedeniya, ArcheOpterix: An extendable tool for
architecture optimization of AADL models, in: Proceedings of the 2009 ICSE Workshop on
Model-Based Methodologies for Pervasive and Embedded Software, MOMPES, 2009, pp. 73β
77. doi:10.1109/MOMPES.2009.5069138.
[2] V. Andrunyk, A. Vasevych, L. Chyrun, N. Chernovol, N. Antonyu, A. Gozhyj, M.
Korobchynskyi: Development of information system for aggregation and ranking of news taking
into account the user needs, 2020. CEUR-WS.org, online CEUR-WS.ORG/Vol-
2604/paper74.pdf.
[3] S. Babichev, V. Lytvynenko, J. Skvor, M. Korobchynskyi, M. Voronenk, Information
technology of gene expression profiles processing for purpose of gene regulatory networks
reconstruction, in: IEEE 2nd International Conference on Data Stream Mining and Processing,
DSMP, 2018, pp. 336-β341. doi:10.1109/DSMP.2018.847845.
[4] B. Boehm, D. Rosenberg, N. Siegel, Critical Quality Factors for Rapid, Scalable, Agile
Development, in: 2019 IEEE 19th International Conference on Software Quality, Reliability and
Security Companion (QRS-C), 2019, pp. 514β515. doi: 10.1109/QRS-C.2019.00101.
[5] C. Chen, M. Shoga, B. Boehm, Exploring the Dependency Relationships between Software
Qualities, in: 2019 IEEE 19th International Conference on Software Quality, Reliability and
Security Companion (QRS-C), 2019, pp. 105β108. doi:10.1109/QRS-C.2019.00032.
202
[6] D. Ageyev, A. Mohsin, T. Radivilova, L. Kirichenko, Infocommunication Networks Design with
Self-Similar Traffic, in: 2019 IEEE 15th International Conference on the Experience of
Designing and Application of CAD Systems (CADSM), 2019, pp. 24β27.
doi:10.1109/CADSM.2019.8779314.
[7] D. Ageyev, A. Mohsin, T. Radivilova, L. Kirichenko, Infocommunication Networks Design with
Self-Similar Traffic, in: 2019 IEEE 15th International Conference on the Experience of
Designing and Application of CAD Systems (CADSM), 2019, pp. 24β27.
doi:10.1109/CADSM.2019.8779314.
[8] D. Ageyev, O. Bondarenko, T. Radivilova, W. Alfroukh, Classification of existing virtualization
methods used in telecommunication networks, in: 2018 IEEE 9th International Conference on
Dependable Systems, Services and Technologies (DESSERT), 2018, pp. 83β86.
doi:10.1109/DESSERT.2018.8409104.
[9] I. Dronyuk, I. Moiseienko, J. Gregus, Analysis of Creative Industries Activities in EuropΠ΅an
Union Countries, Procedia Computer Science 160 (2019) 479β484. (2019).
doi:10.1016/j.procs.2019.11.061.
[10] E. Awad, M. W. Caminada, G. Pigozzi, M. Podlaszewski, I. Rahwan, Pareto optimality and
strategy-proofness in group argument evaluation, Journal of Logic and Computation 27 (2017)
2581β2609.
[11] P. Galkin, R. Umiarov, O. Grigorieva, D. Ageyev, Approaches for Safety-Critical Embedded
Systems and Telecommunication Systems Design for Avionics Based on FPGA, in: 2019 IEEE
International Scientific-Practical Conference Problems of Infocommunications, Science and
Technology (PIC S&T), 2019, pp. 391β396 (2019). doi:10.1109/PICST47496.2019.9061421.
[12] A. Hassan, Y. Jamalludin, Analysis of success factors of technology transfer process of the
information and communication technology, in: 2016 International Conference on Advances in
Electrical, Electronic and Systems Engineering (ICAEES), 2016, pp. 382β387.
doi:10.1109/ICAEES.2016.7888074.
[13] M. Kabir, M. Rashed, Multi-level client server network and its performance analysis, LAP
Lambert Academic Publishing, SaarbrΓΌcken, 2012.
[14] M. L. Abbott, M. T. Fisher, The art of scalability: scalable web architecture, processes, and
organizations for the modern enterprise, 2nd ed., Addison-Wesley Professional, Boston, 2015.
[15] M. Pasyeka, V. Sheketa, N. Pasieka, S. Chupakhina, I. Dronyuk, System Analysis of Caching
Requests on Network Computing Nodes, in: 2019 3rd International Conference on Advanced
Information and Communications Technologies (AICT). 2019, pp. 216β222.
doi:10.1109/AIACT.2019.8847909.
[16] M. O. Medykovskyy, M. S. Pasyeka, N. M. Pasyeka, O. B. Turchyn, Scientific research of life
cycle perfomance of information technology, in: 2017 12th International Scientific and Technical
Conference on Computer Sciences and Information Technologies (CSIT), 2017, pp. 425β428
(2017). doi:10.1109/STC-CSIT.2017.8098821.
[17] M. Nazarkevych, A. Marchuk, L. Vysochan, Y. Voznyi, H. Nazarkevych, A. Kuza: Ateb-Gabor
Filtering Simulation for Biometric Protection Systems, in: Conference on Computer Science and
Information Technologies, 2020. CEUR-WS.org, online CEUR-WS.ORG/Vol-2746/paper2.pdf
[18] O. Mishchuk, R. Tkachenko, I. Izonin, Missing Data Imputation through SGTM Neural-Like
Structure for Environmental Monitoring Tasks, in: Advances in Intelligent Systems and
Computing, 2020, pp. 142β151. doi:10.1007/978-3-030-16621-2_13.
[19] M. A. J. Idrissi, H. Ramchoun, Youssef Ghanou and Mohamed Ettaouil, Genetic algorithm for
neural network architecture optimization, in: 3 International Conference on Logistics Operations
Management (GOL), 2016, pp. 1β4.
[20] H. Mykhailyshyn, N. Pasyeka, V. Sheketa, M. Pasyeka, O. Kondur, M. Varvaruk, Designing
Network Computing Systems for Intensive Processing of Information Flows of Data, in: T.
Radivilova, D. Ageyev, N. Kryvinska, (Eds.), Data-Centric Business and Applications: ICT
Systems-Theory, Radio-Electronics, Information Technologies and Cybersecurity , volume 5,
Springer International Publishing, Cham, 2021, pp. 391β422. doi:10.1007/978-3-030-43070-
2_18.
[21] N. Pasieka, V. Sheketa, Y. Romanyshyn, M. Pasieka, U. Domska and A. Struk, Models, Methods
and Algorithms of Web System Architecture Optimization, in: IEEE International Scientific-
203
Practical Conference Problems of Infocommunications, Science and Technology (PIC S&T),
Kyiv, Ukraine, 2019, pp. 147β153. doi:10.1109/PICST47496.2019.9061539.
[22] N. Pasyeka, H. Mykhailyshyn and M. Pasyeka, Development algorithmic model for optimization
of distributed fault-tolerant web-systems, in: IEEE International Scientific-Practical Conference
Problems of Infocommunications, Science and Technology (PIC S&Tβ2018), 9-12 October,
Kharkiv, 2018, pp. 663β669.
[23] M. Nazarkevych, M. Logoyda, O. Troyan, Y. Vozniy, Z. Shpak, The Ateb-Gabor Filter for
Fingerprinting, in: International Conference on Computer Science and Information Technology,
Springer, Cham, 2019, pp. 247β255.
[24] M. Nazarkevych, M. Logoyda, S. Dmytruk, Y. Voznyi, O. Smotr: Identification of biometric
images using latent elements, 2019. CEUR-WS.org, online CEUR-WS.ORG/Vol-
2488/paper8.pdf
[25] M. Nazarkevych, N. Lotoshynska, I. Klyujnyk, Y. Voznyi, S. Forostyna, I. Maslanych,
Complexity Evaluation of the Ateb-Gabor Filtration Algorithm in Biometric Security Systems,
in: 2019 IEEE 2nd Ukraine Conference on Electrical and Computer Engineering (UKRCON),
2019, pp. 961β964.
[26] M. Nazarkevych, N. Lotoshynska, V. Brytkovskyi, S. Dmytruk, V. Dordiak, I. Pikh, Biometric
identification system with ateb-gabor filtering, in: 11th International Scientific and Practical
Conference on Electronics and Information Technologies (ELIT), 2019, pp. 15β18.
doi:10.1109/ELIT.2019.8892282.
[27] O. Riznyk, Yu. Kynash, O. Povshuk and V. Kovalyk, Recovery schemes for distributed
computing based on bib-schemes, in: First International Conference on Data Stream Mining &
Processing (DSMP), 2016, pp.134β137.
[28] P. Ravi, V. K. Sundar, A. Chattopadhyay, S. Bhasin, A. Easwaran, Authentication Protocol for
Secure Automotive Systems: Benchmarking Post-Quantum Cryptography, in: IEEE International
Symposium on Circuits and Systems (ISCAS), Sevilla, 2020, pp. 1β5. doi:
10.1109/ISCAS45731.2020.9180847.
[29] N. Pasieka, V. Sheketa, Y. Romanyshyn, M. Pasieka, U. Domska, A. Struk, Models, methods
and algorithms of web system architecture optimization, in: IEEE International Scientific-
Practical Conference: Problems of Infocommunications Science and Technology, PIC S&T,
2019, 147β152. doi:10.1109/PICST47496.2019.9061539.
[30] M. Pasyeka, V. Sheketa, N. Pasieka, S. Chupakhina, I. Dronyuk, System analysis of caching
requests on network computing nodes, in: 3rd International Conference on Advanced
Information and Communications Technologies, AICT, 2019, pp. 216β222.
doi:10.1109/AIACT.2019.8847909.
[31] M. Pasyeka, T. Sviridova, I. Kozak, Mathematical model of adaptive knowledge testing, in: 5th
International Conference on Perspective Technologies and Methods in MEMS Design,
MEMSTECH, 2009, pp. 96β97.
[32] M. Pasyeka, V. Sheketa, N. Pasieka, S. Chupakhina, I. Dronyuk, System analysis of caching
requests on network computing nodes, in: 3rd International Conference on Advanced
Information and Communications Technologies, AICT, 2019, pp. 216β222.
doi:10.1109/AIACT.2019.8847909.
[33] Q. Linling, W. Qingfeng, Research on Automatic Test of WEB System Based on Loadrunner,
in: 13th International Conference on Computer Science & Education, Sri Lanka, 2018, pp. 1β4.
doi:10.1109/ICCSE.2018.8468852.
[34] R. AllΓ©aume, I. P. Degiovanni, A. Mink, T. E. Chapuran, N. Lutkenhaus, M. Peev, C. J.
Chunnilall, V. Martin, M. Lucamarini, M. Ward, A. Shields, Worldwide standardization activity
for quantum key distribution, in: 2014 IEEE Globecom Workshops (GC Wkshps), 2014, pp.
656β661. doi:10.1109/GLOCOMW.2014.7063507.
[35] R. Paul, J. R. Drake, H. Liang, Global Virtual Team Performance: The Effect of Coordination
Effectiveness, Trust, and Team Cohesion, IEEE Transactions on Professional Communication 59
(2016) 186β202. doi:10.1109/TPC.2016.2583319.
[36] R. Privman, S. R. Hiltz, Y. Wang, In-Group (Us) versus Out-Group (Them) Dynamics and
Effectiveness in Partially Distributed Teams, IEEE Transactions on Professional Communication
56 (2013) 33β49. doi:10.1109/TPC.2012.2237253.
204
[37] O. Riznyk, O. Povshuk, Y. Kynash, M. Nazarkevich, I. Yurchak, Synthesis of non-equidistant
location of sensors in sensor network, in: 14th International Conference on Perspective
Technologies and Methods in MEMS Design, MEMSTECH, 2018, pp. 204β208.
doi:10.1109/MEMSTECH.2018.8365734.
[38] L. Sikora, N. Lysa, B. Fedyna, B. Durnyak, R. Martsyshyn, Y. Miyushkovych, Technologies of
Development Laser Based System for Measuring the Concentration of Contaminants for
Ecological Monitoring, in: 2018 IEEE 13th International Scientific and Technical Conference on
Computer Sciences and Information Technologies (CSIT), 2018, pp. 93β96. doi:10.1109/STC-
CSIT.2018.8526602.
[39] L. Sikora, R. Martsyshyn, Y. Miyushkovych, N. Lysa, B. Yakymchuk, Problems of data
perception by operators of energy-active objects under stress, in: The Experience of Designing
and Application of CAD Systems in Microelectronics, 2015, pp. 475β477.
doi:10.1109/CADSM.2015.7230909.
[40] T. Aslam, T. Rana, M. Batool, A. Naheed, A. Andaleeb, Quality Based Software Architectural
Decision Making, in: 2019 International Conference on Communication Technologies
(ComTech), 2019, pp. 114β119. doi:10.1109/COMTECH.2019.8737836.
[41] R. Tkachenko, I. Izonin, N. Kryvinska, I. Dronyuk, K. Zub, An Approach towards Increasing
Prediction Accuracy for the Recovery of Missing IoT Data based on the GRNN-SGTM
Ensemble, Sensors 20 (2020). doi:10.3390/s20092625.
[42] R. Tkachenko, I. Izonin, P. Vitynskyi, N. Lotoshynska, O. Pavlyuk, Development of the non-
iterative supervised learning predictor based on the ito decomposition and sgtm neural-like
structure for managing medical insurance costs, Data 3 (2018). doi:10.3390/data3040046.
[43] U. Banerjee, A. Pathak, A. P. Chandrakasan, An Energy-Efficient Configurable Lattice
Cryptography Processor for the Quantum-Secure Internet of Things, in: IEEE International
Solid- State Circuits Conference - (ISSCC), San Francisco, CA, USA, 2019, pp. 46β48.
doi:10.1109/ISSCC.2019.8662528.
[44] V. B. Dang, F. Farahmand, M. Andrzejczak, K. Gaj, Implementing and Benchmarking Three
Lattice-Based Post-Quantum Cryptography Algorithms Using Software/Hardware Codesign, in:
International Conference on Field-Programmable Technology (ICFPT), Tianjin, China, 2019, pp.
206β214. doi: 10.1109/ICFPT47387.2019.00032.
[45] V. DrΔgoi, T. Richmond, D. Bucerzan and A. Legay, Survey on cryptanalysis of code-based
cryptography: From theoretical to physical attacks, in: 7th International Conference on
Computers Communications and Control (ICCCC), Oradea, 2018, pp. 215β223.
doi:10.1109/ICCCC.2018.8390461.
[46] W. Liu, J. Yang, Y. Song, X. Yu and S. Zhao, Research on Software Quality Evaluation Method
Based on Process Evaluation and Test Results, in: 6-th International Conference on Dependable
Systems and Their Applications (DSA), Harbin, China, 2020, pp. 480β483. doi:
10.1109/DSA.2019.00077.
[47] Y. Romanyshyn, V. Sheketa, L. Poteriailo, V. Pikh, N. Pasieka, Y. Kalambet, Social-
communication web technologies in the higher education as means of knowledge transfer, in:
IEEE 2019 14th International Scientific and Technical Conference on Computer Sciences and
Information Technologies (CSIT), Lviv, Ukraine, 2019, pp. 35β39.
[48] Y. Gao-Wei, H. Zhanju, A Novel Atmosphere Clouds Model Optimization Algorithm, in:
International Conference on Computing, Measurement, Control and Sensor Network, 2012
pp. 217β220.
[49] Q. Linling, W. Qingfeng, Research on Automatic Test of WEB System Based on Loadrunner, in:
13th International Conference on Computer Science & Education, Sri Lanka, 2018, pp. 1β4,
doi:10.1109/ICCSE.2018.8468852.
[50] M. Zharikova, V. Sherstjuk, Academic integrity support system for educational institution, in:
IEEE 1st Ukraine Conference on Electrical and Computer Engineering, UKRCON, 2017, pp.
1212β1215. doi:10.1109/UKRCON.2017.8100445.
[51] M. Zharikova, V. Sherstjuk, N. Baranovskiy, The plausible wildfire model in geoinformation
decision support system for wildfire response, in: Paper presented at the International
Multidisciplinary Scientific GeoConference Surveying Geology and Mining Ecology
Management, SGEM, 2015, pp. 575β584.
205