CC BY
7
Simulations of Computer, Telecommunications, Control and Social Systems
Моделирование вычислительных, телекоммуникационных, управляющих и социально-экономических систем
Research article
DOI: https://doi.org/10.18721/JCSTCS.16203 UDC 519.876.5
A SET OF SIMULATION MODELS OF WORKFLOW ELEMENTS FOR TRANSACTIONAL SERVICES OF INFOCOMMUNICATION SYSTEMS
N.M. Redrugina' H
1 Bonch-Bruevich Saint Petersburg State University of Telecommunications, St. Petersburg, Russian Federation
H redrugina.nm@sut.ru
Abstract. This article describes the development of a set of simulation models for workflow elements related to transactional services in infocommunication systems. The article presents modeling methods used to calculate, analyze and predict various indicators of the quality of infocommunication services, such as delays, service delivery time and productivity. The necessity of using this method is due to the reduction of labor costs for modeling complex systems and the evaluation of the rationality of the developed mathematical models. The article presents the results of the study in the form of a library of components for simulation modeling, which can be used to improve the quality of transactional services in the development of modern high-load infocommunication systems. The approach described in the article is an effective tool for research and optimization of the quality of transactional services. It can be applied in various areas where analysis and improvement of the performance of data processing systems are required, including for the development of integrated solutions for big data processing, cloud computing or the development of digital counterparts of infocommunication infrastructure.
Keywords: simulation modeling, transactions, work-flow, infocommunication systems, fork-join, parallel processing
Citation: Redrugina N.M. A set of simulation models of workflow elements for transactional services of infocommunication systems. Computing, Telecommunications and Control, 2023, Vol. 16, No. 2, Pp. 29-39. DOI: 10.18721/JCSTCS.16203
© Redrugina N.M., 2023. Published by Peter the Great St. Petersburg Polytechnic University
Научная статья
DOI: https://doi.org/10.18721/JCSTCS.16203 УДК 519.876.5
КОМПЛЕКС ИМИТАЦИОННЫХ МОДЕЛЕЙ ЭЛЕМЕНТОВ РАБОЧИХ ПРОЦЕССОВ ДЛЯ ТРАНЗАКЦИОННЫХ УСЛУГ ИНФОКОММУНИКАЦИОННЫХ СИСТЕМ
Н.М. Редругина' и
1 Санкт-Петербургский государственный университет телекоммуникаций им. проф. М.А. Бонч-Бруевича, Санкт-Петербург, Российская Федерация
и redrugina.nm@sut.ru
Аннотация. Данная статья описывает разработку комплекса имитационных моделей элементов рабочих процессов, связанных с транзакционными услугами в инфокоммуни-кационных системах. В статье представлены методы моделирования, используемые для расчета, анализа и прогнозирования различных показателей качества инфокоммуника-ционных услуг, таких как задержки, время предоставления услуги и производительность. Необходимость применения данного метода обусловлена снижением трудозатрат на моделирование сложных систем и оценкой рациональности разработанных математических моделей. В статье представлены результаты исследования в виде библиотеки компонент для имитационного моделирования, которые могут быть использованы для повышения качества обслуживания транзакционных услуг при разработке современных высоконагру-женных инфокоммуникационных систем. Описанный в статье подход является эффективным инструментом для исследования и оптимизации качества предоставления тран-закционных услуг, и может быть применен в различных областях, где требуется анализ и улучшение производительности систем обработки данных, в том числе для разработки комплексных решений обработки больших данных, облачных вычислений или при разработке цифровых двойников инфокоммуникационной инфраструктуры.
Ключевые слова: имитационное моделирование, транзакции, work-flow, инфокоммуника-ционные системы, fork-join, параллельная обработка
Для цитирования: Redrugina N.M. A set of simulation models of workflow elements for transactional services of infocommunication systems // Computing, Telecommunications and Control. 2023. Т. 16, № 2. С. 29-39. DOI: 10.18721/JCSTCS.16203
Introduction
The purpose of the study is to develop models for calculating and determining the qualitative characteristics of infocommunication services in general and their elements. The works [1—2] describe the reasons why it is irrational to collect statistics and analyze the behavior of the system under the influence of various external and internal factors on the functioning system, which led to the need for their modeling. In [3—6], a comparative analysis of modeling methods and their differentiation, as well as the synthesis of results, is carried out.
In the case of high labor costs for mathematical modeling of systems with complex architecture, when introducing complex distributions into the model (for example, long-tailed distributions: Weibull or Pare-to), or if it is necessary to calculate error correlation, it is advisable to use simulation models, a comparative analysis of software tools for the development of which is presented in [7—8].
The goal set within the framework of this work is to develop a family of simulation models for calculating the qualitative characteristics of infocommunication loosely coupled transaction services, including the development of a library of components of the workflow for servicing tasks within transactional services.
© Редругина Н.М., 2023. Издатель: Санкт-Петербургский политехнический университет Петра Великого
To achieve this goal, the following tasks are proposed:
1. Study of the main technologies and methods of implementing transaction servicing in loosely coupled services. Including using the queuing theory apparatus.
2. Development of a workflow component Library: Creating a library that contains ready-made components used in task maintenance workflows within transactional services.
3. Implementation of simulation models: Development of software code for the implementation of simulation models based on a specific architecture and components from the library.
4. Evaluation of simulation results with a combination of components in the workflow.
The research carried out within the framework of this work will allow develop simulation models of architectures of varying complexity for infocommunication networks and systems, to assess and predict the characteristics of the service being developed. This will shorten the development and implementation time, verify and verify the results obtained by using a mathematical apparatus, as well as obtain qualitative characteristics of systems that are complex enough for mathematical modeling. Thus, the approach to software development based on Model-Driven Architecture models allows you to use models as a set of recommendations for structuring specifications used in software design.
The authors of works on simulation modeling of infocommunication services [9-10], simulate specific services in order to obtain their characteristics, which does not allow using the proposed models for the analysis of other systems. Separately, it should be noted the work of the author [11] in the framework of which an analysis of a particular service was presented with a detailed description of the approach to its modeling, however, the lack of a description of the implementation of structural elements does not allow to accurately reproduce the study.
1. Approach to the development of simulation models of systems.
The research methods used in this work include empirical studies using mathematical and simulation models to analyze qualitative characteristics. As well as analytical methods for formalizing processes, applying statistical methods to extract information from data. Designing a library of workflow components, which includes defining the functionality of each component, their interaction and interfaces for information exchange.
To create a simulation model of an infocommunication service, the following steps must be performed [12-13]:
1) Definition of the purpose of modeling. This will help determine the required level of detail, quantitative characteristics and other parameters of the model.
2) Collecting data that will be used as input characteristics of the system. This can be data on the flow of applications processed in the system, the characteristics of the network infrastructure, the number of users, and others.
3) Definition of the model structure — based on the data and modeling goals, the structure of the model, the processes occurring in the system, and the parameters of each block are determined.
4) Creating a program for the implementation of the model, using special software tools, such as AnyLogic.
5) Run the simulation and analyze the results. As a result of the analysis, it is possible to determine the efficiency of the system, optimize processes and suggest possible improvements.
There are many different types of simulation models, each of which has its own characteristics and applications. Within the framework of this work, it is advisable to use discrete event modeling methods. These models are quite flexible in the implementation of the necessary functionality and can be used to model a large number of different processes. Examples of the implementation of a transactional service [14—15] and a session service [16—18]. In this paper, simulation modeling of infocommunication services and their elements is carried out in the Anylogic development environment [19].
The main element of the model is a queuing system (QS), which can be part of a network of queuing systems implementing a complex service of applications, data blocks, users depending on the purpose of the system being modelled.
4
The quality of the system functioning is determined by many factors such as:
— Response time: This is the mathematical expectation and variance of the time it takes the system to process a request and send a response. For most transactional systems, the response time should be minimal.
— Lead time: this is the time it takes the system to perform the operation. For transactional systems, the execution time should also be minimal.
— Downtime: this is the time when the system does not work and does not process requests. The less downtime, the better for users.
— System reliability: this is the probability that the system will be able to process the request.
— Throughput: This is the number of requests the system can handle per unit of time. For transactional systems, throughput should be high enough to meet the needs of users.
— The system's node utilisation factor is a measure of the extent to which a particular node's resources are utilised within a distributed system.
For the latter, the expression for the mathematical calculation of the load factor is as follows:
XXt X.
P, == —. (1)
V, ^ , X V,
To obtain this parameter in the system, the following algorithm is required at the "Queue" input:
count= count+delay.size(); // counts the total number of delays in the node
ro=count/(v*number); // determines the average node load
This paper considers modern information systems that include transactional systems designed to automate business processes by entering, storing and processing information. They are an important component of infocommunication systems that support workflows [20] and e-commerce applications. Transactional systems play an important role in the operation of such systems by enabling interaction between various components, including web services, servers, local databases and other business transactional participants.
In this paper, the definition of workflows is presented as rule-based software that coordinates and controls the execution of the activities (transactions) that constitute a business process. Workflow can include issues of concepts (Cloud-Fog-Edge), distributed systems, multi-phase maintenance, etc. [21—22].
Simulation models of workflow scenarios
The study of transactional systems using simulation modelling provides accurate estimates of the probabilistic and temporal characteristics in various task processing scenarios in transactional services.
Since transactional services are an implementation of a workflow for servicing a user task, there are an infinite number of implementation options. It can be argued that there is no universal model of workflow definitions. In the following, the developed approaches to the simulation of the main elements of the workflow are presented.
A sequential chain of local transactions is a series of connected mass service systems, where the output from one QS is the input to the other. An example implementation of a workflow simulation model of this type will be shown in Figure 3.
The branching model of system workflow scenarios, allows to consider the processes of a task along the route of local nodes (e.g. Nodel ^ Node 2, Node 1 ^ Node 3, Node 1 ^ Node 4). An example of a service with this approach could be a payment system with different payment methods, the choice of which implements several scenarios. The user information notification system works in the same way with the following scenarios (through SMS, push notifications and mail).
This model includes the ability to choose the path of a request to nodes. The probabilities of user transitions between nodes in the system are given by an indecomposable stochastic matrix R = | p 11, in the case of a system of three nodes:
P11 P12 P13
R = P21 P22 P23
P31 P32 P33
The route selection blocks "SelectOutputOut" ^ "Probability", or "Variable" ^ "Initial value" with reference to the database include the following condition SELECT probability of change FROM "DB name" WHERE route = "pij" // select the correct transition, from i to node j
In the scenarios of this model, the intensity of requests to the Semo nodes will be calculated as follows:
^ =AXPo, XPfl■
(2)
i=1
A script for aggregating requests from different sources to merge, process and generate either a merged request by work result or split up and redistribute between the next units in the workflow chain.
When modelling such a structure, it is essential to identify the requests within the workflow, to set the route and conditions for the distribution between nodes in the system.
The result of modelling systems with scenario partitioning and reallocation of requests to nodes in a mass service network is presented in [17].
3) Parallel service (Fork-Join [23]) involves replicating a request to M nodes, with two possible objectives, the first of which involves simulating the process of simultaneously polling two loosely coupled services in order to get a response from all nodes.
The approximate distribution function for the whole network with replication factor equal to m, is calculated by the expression:
M
M
Ffj (t) -1 - Z ci [1 - Fs,(t)] n F(t) « S (ci - ci+1 )nF(t),
j=1 i =1 j=1
(3)
i=1
where c. — is the replication factor, equal to cM+1 = 0, c1 = 1, c2 = 1 — p1 /4 and
Pi
'1 - Za -v 8
1 r i-2 8
3 < i < M.
(4)
The second option involves the need to divide the task into smaller equal parts (data block) and service them simultaneously by parallel nodes [24]. It should be noted that then the maintenance time of the transaction needs to be changed:
t =
serv.
t
M
(5)
A simulation model of this type of system has been implemented as follows:
It is worth noting that this simulation model automates the process of changing the number of replicas by introducing the functions described below:
This model has resulted in a software module for characterising parallel transaction processing systems [25].
Modelling results
The result of the simulation of a transaction service as a sequential chain of local transactions branching into parallel service Fork-Join M = 2, is the simulation model shown in Fig. 3.
4
Fig. 1. Branching, aggregation and redistribution of flows a) scheme b) model
Fig. 2. Fork-Join M node simulation model
As a result of the simulation, the time characteristics were obtained as a function of the input load intensity of the workflow and the performance of its fragments (service intensity p.). Also using the convolution equation, the results of analytical modelling were obtained. The M/M/1 systems are independent systems, so the relative error of analytical modelling is insignificant. However, modeling systems with other distribution laws that are not included in the results of Burke's theorem is time consuming and may have a high percentage of inaccuracy. It is for such systems that it is rational to use simulation methods.
Fig. 3. Transaction chain simulation model with FJ
Description of the functional units in the FJ model
Table 1
Block Destination Prerequisite
Function "number_of_nodes" Changes the number of active nodes by the value defined in the "M_nodes" parameter if (M_nodes == 1) {Nodes_1();} else if (M nodes == 2) {Nodes 2 ();} etc.
Function "Nodes_i" where i=1...M Determines the probabilities of replica processing by specific nodes For replica factor M=2 M 1=1/M nodes; M 2=1/M nodes; M_3=0; etc.
ID=agent.number; replica_counter=0; for (i=0;i <Buffer_replic. size();i++) { if ((Buffer_replic.get(i)). number==ID) {replica_counter++; } if (replica_counter==M_nodes){ vhod konteiner.inject(); i=0; break;}}
Queue "Buffer_replic" Waits for all M replicas to arrive before combining them into an application and ending service
Pickup "Receiving_replic" Collects incoming replicas in the "Buffer_replic" block into a single application Requirement: agent.number==ID;
Dropoff "Transfer_replic" Transmits application from "Receiving_replic" to system output Requirement: agent.v_replication==1
Fx (t ) = F (t )* F2 (t )* Ffj (t). (6)
As can be seen from Fig. 4, when the input intensity is increased for the whole system, the task service time, or more precisely, the waiting time for service by an application in each element, increases. The increase occurs uniformly until the first node is overloaded at A = 1 request/sec, at this point node 1 is identified as the bottleneck of the system, thereby slowing down the rate of requests to subsequent nodes. Further modelling suggests that for a given critical input load A , to gradually increase the service
KjJV.Y}!..
intensity at all nodes of the systems, which, as can be seen in Fig. 5, significantly reduces the time to implement the problem.
0 JS.,- —
0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 0,92 0,94 0,96 0,98 1 Input load intensity, request/sec.
Fig. 4. Dependence of residence time on A
Fig. 6. Density (a) and distribution function (b) of time spent in systems
The moment when the load factor of all nodes p. ^ 1 at the input intensity A = 0.99 the system tends to overload and has an unacceptable service time. The density and probability of time in the system and in each element are shown in Fig. 6.
1000 _ _ _
1111111111
0 I II III IV
The solution
Fig. 7. Dependence of the time in the system on the chosen solution
The question of the rationality of improving the qualitative performance of all elements of the system at the same time has emerged. Simulation modelling allows the time characteristics of the system to be obtained with a minimum of effort by conducting a number of different experiments. This allows you to significantly reduce the cost of resources to provide the required quality of service. For example, Fig. 7 shows the results of a simulation of a transaction chain with the input characteristics given in Table 2.
Table 2
Quantitative characteristics of node performance decisions
The solution
0 I II III IV
X request /sec. 1,2 1,2 1,2 1,2 1,2
^ node 1 request /sec 1,1 1,2 1,3 1,3 1,2
^ node 2 request sec. 1,2 1,2 1,2 1,3 1,3
^ node 3 request /sec 1,3 1,3 1,3 1,3 1,4
The zero solution is the initially set system parameters, and as can be seen from the diagram, with an increase in the performance of a highly loaded node by 1—2 divisions, where division = 0.1), the quality of service of the entire service will not increase, as it will overload the subsequent node. The most rational method in this experiment is to spend the same 3 divisions as in solution IV, but on the first and second node, thereby reducing the load on the nodes and not spending unnecessary resources on the stable node. This approach allows us to calculate the characteristics of the system when implementing an algorithm based on the equivalent microservices workflow focused on the quality-of-service delivery [26].
Conclusion
As part of this work, simulation models of the functional blocks required to create holistic infocommu-nication services were developed by assembling them into the request service workflow scenarios presented in the following elements:
1) a sequential chain of local transactions;
2) branching workflow scenarios;
3) scenarios for merging. splitting and redistributing query flows;
4) Fork-Join parallel service.
The model blocks presented are suitable for the layout of different variants of complex services. The use of simulation modelling was found to provide accurate estimates of the probabilistic and temporal charac-
teristics of task servicing in transactional service scenarios. More complex task processing scenarios can be considered in further studies, such as:
1) multiphase mass service networks;
2) distributed systems in workflow scenarios (Edge-Cloud computing concept);
3) systems for priority processing of requests;
4) transactional systems for handling requests initiated by session services.
This will extend the scope of simulation modelling and provide more accurate results for more complex task processing systems in transactional services.
REFERENCES
1. Khomonenko A.D., Basyrov A.G. i dr. Modeli i metody issledovaniya informatsionnykh system [Models and methods of information systems research], 2019.
2. Paramonov I.Yu. i dr. Metody i modeli issledovaniya slozhnykh sistem i obrabotki bolshikh dannykh [Methods and models of complex systems research and big data processing], 2020.
3. Shanthikumar J.G., Sargent R.G. A unifying view of hybrid simulation/analytic models and modeling, Operations research, 1983. T. 31, no. 6. Pp. 1030—1052.
4. Nyhuis P. et al. Applying simulation and analytical models for logistic performance prediction, CIRP annals, 2005. T. 54, no. 1. Pp. 417-422.
5. Litvinov V.L., Zhuravkova M.A. Issledovaniye sredstv modelirovaniya slozhnykh sistem metodami sis-temnoy dinamiki [Research of modeling tools of complex systems by methods of system dynamics], Regional Informatics (RI-2022) Conference materials. 2022, Pp. 561-562.
6. Gaydamako V.V. Comparative analysis of analytical modeling methods for evaluating the performance of cloud applications, VestnikKyrgyzsko-Rossiyskogo Slavyanskogo universiteta, 2020. T. 20, no. 12. Pp. 145-151.
7. Kuznetsova T.A., Meshcheryakov V.A. Comparison of software tools for simulation of queuing systems. Materials of the international scientific and practical conference of students and postgraduates. Omsk: National Development Strategy Fund, 2015, Pp. 187-192.
8. Apanasenko A.V., Berg D.B. Sravnitelnyy analiz metodov integratsii imitatsionnykh modeley s vneshnimi informatsionnymi sistemami [Comparative analysis of methods of integration of simulation models with external information systems], Vesenniye dni nauki VShEM [Spring Days of HSE Science]. 2019. Pp. 184-186.
9. Dorosinskiy L.G., Aksenov K.A., Popov M.V. Multiservices communications networks simulating dynamic modeling and their engineering design, Nauchno-technicheskie vedomosti SPbGPU. Informatika, telekommunikatsii i upravleniye. 2009, no. 1 (72). Pp. 153-159.
10. Abdalov A.V., Grishako B.G., Loginov I.V. An analysis of the efficiency of the process of servicing the flow of requests for creating IT-services used a simulation model, Software & Systems, 2022. T. 35, no. 1. Pp. 75-82. DOI: 10.15827/0236-235X.137.075-082
11. Bykov A.Yu., Panfilov F.A., Sumarokova O.O. Simulation with usage of the Java class library, developed for "cloud" services, Inzhenernyy zhurnal: nauka i innovatsii. 2013, no. 2 (14). P. 10.
12. Kolesov Yu.B., Senichenkov Yu.B. Modelirovaniye sistem. Obyektno-oriyentirovannyy podkhod [Modeling of systems. Object-oriented approach], BKhV-Peterburg, 2012.
13. Verzun N.A., Kolbanev M.O., Romanova A.A. Imitatsionnoye modelirovaniye informatsionnogo vzai-modeystviya v kibertekhnicheskoy sisteme [Simulation modeling of information interaction in a cybertechnical system], Conference Information Security of Russian regions (IBRD-2021), 2021. Pp. 211-212.
14. Shatalova Yu.G. Issledovaniye modeli obrabotki zaprosov k baze dannykh v srede AnyLogic [Investigation of the database query processing model in AnyLogic environment], Vostochno-yevropeyskiy zhurnal peredovykh tekhnologiy [Eastern European Journal of Advanced Technologies], 2012. T. 4, no. 2 (58). Pp. 46-48.
15. Nazarenko D.N., Sorokin A.A. Diskretno-sobytiynaya model obrabotki zaprosov serverom sistemy distantsionnogo obucheniya v srede imitatsionnogo modelirovaniya AnyLogic [Discrete-event model of request processing by the server of the distance learning system in the AnyLogic simulation environment], Aktualnyye problemy sovremennoy nauki — novomu pokoleniyu [Actual problems of modern science for a new generation]. 2014. Pp. 171-175.
16. Svidetelstvo o gosudarstvennoy registratsii programmy dlya EVM 2015612377 Diskretno-sobytiynaya model Call-tsentra v srede imitatsionnogo modelirovaniya Anylogic [Certificate of state registration of a computer program 2015612377 Discrete event model of a Call center in the Anylogic simulation environment]. Sorokin A.A., Zhukov I.A., Melnikova K.V. 2015.
17. Zarubin A.A., Redrugina N.M. Modelirovaniye igrovykh mnogopolzovatelskikh servisov [Modeling of multi-user gaming services], Vestnik svyazi [Bulletin of Communications], 2020. No. 8. Pp. 11-16.
18. Redrugina N.M. Models and methods for calculating delays in the provision of services by the user on the service platforms of session infocommunication services. T-Comm, 2023. Vol. 17, no. 4, pp. 32-38. (in Russian). DOI: 10.36724/2072-8735-2023-17-4-32-38
19. Yefromeyeva Ye.V. Simulation modeling: fundamentals of practical application in AnyLogic environment: tutorial / Ye.V. Yefromeyeva, N.M. Yefromeyev. Saratov: Vuzovskoye obrazovaniye, 2020.
20. Leff, Avraham & Rayfield, James. WSO: Developer-Oriented Transactional Orchestration of Web-Services. 2017. Pp. 714-720. DOI: 10.1109/ICWS.2017.86
21. Vorobyev S.P., Gorobets V.V. Issledovaniye modeli tranzaktsionnoy sistemy s replikatsiyey fragmentov bazy dannykh, postroyennoy po printsipam oblachnoy sredy [Investigation of a transactional system model with replication of database fragments built on the principles of a cloud environment], Inzhenernyy vestnik Dona [Engineering Bulletin of the Don]. 2012. T. 22, no. 4-1. P. 49.
22. Wang S., Ding Z., Jiang C. "Elastic Scheduling for Microservice Applications in Clouds", in IEEE Transactions on Parallel and Distributed Systems, vol. 32, no. 1, pp. 98-115, 1 Jan. 2021, DOI: 10.1109/ TPDS.2020.3011979
23. Lokhvitskogo V.A., Khabarova R.S., Dudkina A.S. Approximation of the residence time for a Fork-Join queuing system based on relationship invariants, Intellektualnyye tekhnologii na transporte [Intelligent technologies in transport], 2020, no. 2, Pp. 46-50.
24. Khalid M., Yousaf M.M. A comparative analysis of big data frameworks: An adoption perspective. Applied Sciences, 2021, vol. 11, no. 22, article number: 11033. DOI: 10.3390/app112211033
25. Certificate of state registration of the computer program No. 2023618304 Russian Federation. A software module for calculating the characteristics of parallel transaction processing systems developed in the Anylogic environment / A.A. Zarubin, N.M. Redrugina; the applicant and the patent holder of SPbSUT. No. 2023616789; declared 07.04.24; publ. 21.04.23.
26. Song Z., Tilevich E. Equivalence-Enhanced Microservice Workflow Orchestration to Efficiently Increase Reliability. 2019, Pp. 426-433. DOI: 10.1109/ICWS.2019.00076
INFORMATION ABOUT AUTHOR / СВЕДЕНИЯ ОБ АВТОРЕ
Nataliya M. Redrugina Редругина Наталия Михайловна
E-mail: redrugina.nm@sut.ru
Submitted: 16.04.2023; Approved: 25.06.2023; Accepted: 06.07.2023. Поступила: 16.04.2023; Одобрена: 25.06.2023; Принята: 06.07.2023.
