Authors

  • Mantu Singh
    Software Architect, Reged Acton, USA

DOI:

https://doi.org/10.37547/tajet/Volume07Issue04-21

Keywords:

cloud computing microservices service mesh processing patterns scalability

Abstract

This study examines a decentralized approach to implementing a service mesh for microservice-based systems designed for scalable data processing. Unlike traditional solutions dominated by the pipes-and-filters pattern and a centralized control plane, this approach utilizes the concept of Eblocks—unified modules that incorporate service discovery, authentication, monitoring, and load management components. This allows for the formation of various patterns (manager-worker, divide-and-conquer, hybrid models) directly at the microservice level without strict dependence on centralized logic. It is demonstrated that such an architecture accelerates data processing through automatic scaling and parallel execution, simplifies configuration, and provides flexible security and observability mechanisms. The proposed results, supported by findings from other researchers, indicate a significant increase in system throughput when handling documents requiring pipeline, parallel, and distributed processing. The presented information is of interest to researchers and professionals in distributed systems, cloud computing, and microservice architecture, aiming for a deeper understanding and implementation of innovative service mesh architectures to enhance the scalability, reliability, and efficiency of modern IT applications.


background image

The American Journal of Engineering and Technology

157

https://www.theamericanjournals.com/index.php/tajet

TYPE

Original Research

PAGE NO.

157-165

DOI

10.37547/tajet/Volume07Issue04-21



OPEN ACCESS

SUBMITED

17 February 2025

ACCEPTED

25 March 2025

PUBLISHED

30 April 2025

VOLUME

Vol.07 Issue 04 2025

CITATION

Mantu Singh. (2025). Implementing Service Mesh Architecture for Scalable
Applications. The American Journal of Engineering and Technology, 7(04),
157

165. https://doi.org/10.37547/tajet/Volume07Issue04-21

COPYRIGHT

© 2025 Original content from this work may be used under the terms
of the creative commons attributes 4.0 License.

Implementing Service
Mesh Architecture for
Scalable Applications

Mantu Singh

Software Architect, Reged
Acton, USA

Abstract:

This study examines a decentralized approach

to implementing a service mesh for microservice-based
systems designed for scalable data processing. Unlike
traditional solutions dominated by the pipes-and-filters
pattern and a centralized control plane, this approach
utilizes the concept of Eblocks

unified modules that

incorporate

service

discovery,

authentication,

monitoring, and load management components. This
allows for the formation of various patterns (manager-
worker, divide-and-conquer, hybrid models) directly at
the microservice level without strict dependence on
centralized logic. It is demonstrated that such an
architecture accelerates data processing through
automatic scaling and parallel execution, simplifies
configuration, and provides flexible security and
observability mechanisms. The proposed results,
supported by findings from other researchers, indicate
a significant increase in system throughput when
handling documents requiring pipeline, parallel, and
distributed processing. The presented information is of
interest to researchers and professionals in distributed
systems,

cloud

computing,

and

microservice

architecture, aiming for a deeper understanding and
implementation

of

innovative

service

mesh

architectures to enhance the scalability, reliability, and
efficiency of modern IT applications.

Keywords:

cloud computing, microservices, service

mesh, processing patterns, scalability, decentralized
architecture, Eblocks.

Introduction:

Cloud computing provides flexible and

dynamic access to computational resources with
minimal management costs. The shift from monolithic
applications to microservices enhances autonomy,
elasticity, and accelerates the deployment cycle of new
features.

Microservice

architecture

promotes


background image

The American Journal of Engineering and Technology

158

https://www.theamericanjournals.com/index.php/tajet

The American Journal of Engineering and Technology

modularity and simplifies the maintenance of large-
scale systems; however, it presents engineers with
several challenges, including the complexity of
managing interactions among independent services,
ensuring observability, securing communication, and
enabling automatic scaling.

To address these challenges, service mesh platforms
have emerged, handling traffic routing, load balancing,
encryption, monitoring, and service management
without modifying business logic. However, most
modern service meshes are primarily designed around
the pipeline processing pattern and lack built-in
support for a broader range of processing patterns,
such as manager-worker or divide-and-conquer. This
limitation restricts developers and architects in
designing complex, high-performance systems that
require flexible management of parallel data streams
and distributed task execution.

A comprehensive review of contemporary research
follows. Nicolas-Plata A., Gonzalez-Compean J. L., and
Sosa-Sosa V. J. [1], along with Alboqmi R. and Gamble
R. F. [4], propose an innovative approach based on
service mesh integration to support processing
patterns within microservice applications. Parallel to
this, the evolutionary transformation of software
systems from monolithic to microservice architectures
is explored in the works of Akerele J. I. et al. [3],
Newman S. [9], and Decimavilla-Alarcón D. C. and
Marcillo-Franco P. F. [8]. These studies focus on
enhancing scalability and flexibility through cloud
technologies and containerization, allowing systems to
adapt to specific requirements, such as healthcare or
IoT environments.

Another research direction involves developing tools
for evaluating scalability and selecting optimal design
patterns. Wrona Z. et al. [5] expand the capabilities of a
cloud infrastructure simulator to model the dynamics
of scalable systems, while Dhait S. et al. [6] conduct a
comparative analysis of creational patterns in software
development.

Cybersecurity in cloud platforms is also a significant
topic in modern literature, as examined in the study by
Molnar V. and Sabodashko D. [2]. Their work aims to
compare security standards based on NIST guidelines to
identify vulnerabilities in leading cloud platforms.

Additionally, scalability is explored in the context of
serverless computing trends in the study by Li Y. et al.
[7]. Their research summarizes the current state of the

technology, identifies key challenges, and outlines
future prospects for serverless approaches in cloud
services.

A notable gap in the existing research is the absence of
a unified mechanism for combining multiple processing
patterns, such as the sequential integration of
manager-worker and divide-and-conquer models,
within a service mesh. Creating such hybrid patterns
still requires either manually configuring multiple tools
or abandoning the advantages of a transparent service
mesh in favor of low-level orchestration. This lack of a
universal method for integrating diverse microservice
processing patterns into a service mesh model defines
the research gap that this study aims to address.

The objective of this study is to examine the
implementation features of service mesh architecture
for scalable applications.

The scientific novelty lies in the systematic analysis and
comparative evaluation of existing research on service
mesh architectures for scalable applications, enabling
the identification of problem areas and future research
directions.

The proposed hypothesis suggests that implementing a
comprehensive service mesh, where each microservice
is equipped with built-in discovery, authentication, and
monitoring functions, will:

Reduce deployment and configuration time for

distributed processing patterns.

Significantly increase architectural flexibility by

supporting additional processing patterns.

Decrease reliance on traditional proxy-based

approaches

and

centralized

controllers

while

maintaining performance.

The methodological framework of this study includes
an analysis of existing research on microservice
architecture and service meshes.

1. Theoretical and technological foundations of service
mesh

Service mesh initially emerged as a methodology for
abstracting network functions and transferring them to
a separate layer, independent of the business logic of
microservices. Traditionally, service mesh architecture
consists of two planes:


background image

The American Journal of Engineering and Technology

159

https://www.theamericanjournals.com/index.php/tajet

The American Journal of Engineering and Technology

Data Plane: Comprises proxies (sidecars) that

attach to each service. These sidecar proxies intercept
traffic, route requests between services, and provide
authentication, encryption, and tracing mechanisms.

Control Plane: Manages the configuration of

sidecar proxies, monitors network state, balances load,
and enforces security policies, such as mandatory
traffic encryption and authorization. In some
implementations, such as Istio, the control plane
operates as a centralized component where
microservices register to receive routing policies [1].

Below, Figure 1 illustrates the functions of a service mesh.

Fig.1. Functions of the service mesh [1].

Despite the clear advantages of service meshes, their
implementation requires consideration of several
factors,

including

compatibility

with

cloud

orchestrators

(e.g.,

Kubernetes),

configuration

complexity, and the resource-intensive nature of
proxies running alongside each microservice.

The primary limitations of existing implementations
stem from the fact that widely used service mesh
platforms predominantly support the Pipes & Filters
pattern, as it naturally aligns with the concept of
sequential data flow through sidecar proxies. However,
more complex patterns such as Manager

Worker or

Divide & Conquer remain poorly automated in standard
service meshes [7].

The root cause of this limitation lies in the fact that
distributed load management (for instance, launching

multiple instances of worker microservices) is
traditionally

delegated

to

external

container

orchestration systems such as Kubernetes or Docker
Swarm. A service mesh primarily detects new services
and, at best, routes traffic to them in a round-robin
manner [3]. However, it often lacks a flexible
mechanism for defining roles, distributing tasks, and
automatically shutting down redundant worker
services when load conditions change.

Additional compatibility issues arise in practice when
certain microservices require specific network policies,
such as gRPC streaming or WebSocket connections,
which may not be fully supported by the service mesh.
Another challenge is the increased overhead associated
with proxy operation, particularly in environments with
a high density of microservices [7].

Service Discovery: automatic detection of addresses and metadata of
new services.

Load Balancing: Distributing requests across different instances of
the same service.

Monitoring and Tracing: collecting metrics and logs for further
performance analysis.

Security (Security): authentication of microservices in front of each
other, traffic encryption, access rights control.


background image

The American Journal of Engineering and Technology

160

https://www.theamericanjournals.com/index.php/tajet

The American Journal of Engineering and Technology

As a result, developers frequently face a dilemma:
either limit themselves to the convenience of simple
patterns and built-in service mesh mechanisms or
manually implement complex scenarios such as
Manager

Worker or parallel data partitioning,

bypassing

standard

functionality.

This

issue

underscores the need for new decentralized service

mesh solutions that offer greater flexibility in
integrating various microservice processing patterns [2,
6].

Table 1 below presents a comparative analysis of
microservice processing patterns.

Table 1. Comparative characteristics of microservice processing patterns [1, 5, 9]

Pattern

Key features

Typical applications

Support in most

service meshes

Pipes

&

filters

Sequential data transmission between
"filters";

facilitates

pipeline

processing

Simple ETL processes,
log

analysis,

data

transformation

Full

(native

to

service meshes)

Manager–
worker

A central "manager" distributes tasks
to a pool of "workers"; scales by
increasing the number of workers

High-load

processes

requiring

distributed

parallel processing

Partial

(mainly

through orchestrator
functionality)

Divide &
conquer

Task is divided into smaller parts,
processed in parallel, and results are
aggregated

Large-scale
computational

tasks

(ML models, big data
analysis)

Weak

(usually

requires

manual

configuration

and

coding)

As shown in Table 1, service meshes fully support the
classic pipeline-based approach (Pipes & Filters).
However, Manager

Worker and Divide & Conquer

patterns remain on the periphery of adoption and often
require extensive customization at the Kubernetes
manifest level and the integration of additional tools.

2. Proposed architectural approach for integrating
multiple patterns

This section explores a decentralized service mesh
model designed to support flexible data processing
scenarios that require the simultaneous use of multiple
patterns, such as pipes and filters, manager

worker,

and divide and conquer. The primary concept is based
on Eblocks, which encapsulate the business logic of a
microservice along with key service mesh mechanisms,
including authentication, monitoring, service discovery,
and, when necessary, automated load management [4,
9].

Traditional service meshes, such as Istio and Linkerd,
rely on a control plane and a data plane. However, this

architecture centralizes most routing logic, making it
difficult to dynamically expand functionality, such as
quickly adding new processing patterns. In contrast, the
Eblocks-based approach employs the following
principles:

1.

Decentralized storage of service logic. Each

Eblock consists of:

Processing Microservice (PM): The core

business application or function, such as an encryption
service or risk analysis module.

Workload Manager (WM): A local (within the

Eblock) implementation of load distribution algorithms
that activates predefined roles, including manager,
worker, divider, conquer, and combine.

Discovery: A self-registration and peer-to-peer

lookup mechanism, implemented using a distributed
hash table or similar technology.

Authentication:

Built-in

authentication


background image

The American Journal of Engineering and Technology

161

https://www.theamericanjournals.com/index.php/tajet

The American Journal of Engineering and Technology

mechanisms for issuing and verifying security tokens
used for inter-Eblock communication.

Monitoring: A local metrics collection service

accessible via REST API or client libraries [7, 8].

2.

Intercomponent communication interfaces.

These are based on REST, gRPC, WebSocket protocols,
and message buses to ensure reliable routing.

3.

Security and authentication mechanisms.

These include in-transit and at-rest data encryption,
digital certificate management, authentication, and
tokenization.

4.

Anomaly

detection

and

dynamic

risk

assessment. Implementation of intrusion detection
systems, monitoring algorithms, and dynamic threat
evaluation for real-time cybersecurity responses.

5.

Adaptive load balancing and scaling. Local and

global resource allocation algorithms that analyze
current workloads and forecast peak demands.

To implement a specific pattern, groups of Eblocks can
assume various roles, such as Filter, Manager, Worker,
Divide, and Conquer. For example, in a manager

worker pattern, one Eblock acts as the Manager,
handling task segmentation and scheduling, while
multiple Eblocks in the Worker role process the
workload. Figure 2 illustrates how the Eblocks approach
integrates multiple patterns.

Fig. 2. The scheme of the Eblocks approach in the process of integrating several patterns [1, 4].

For the Eblocks model to function effectively, a
comprehensive orchestration strategy is required. This
typically relies on Kubernetes or a similar orchestrator
but does not duplicate its functions. To provide a

clearer understanding, Table 2 compares key elements
of the traditional service mesh model (using Istio as an
example) and the proposed Eblocks architecture.


background image

The American Journal of Engineering and Technology

162

https://www.theamericanjournals.com/index.php/tajet

The American Journal of Engineering and Technology

Table 2. Comparison of the classical service mesh model and Eblocks [1, 4]

Characteristic

Classical service mesh (e.g., Istio)

Eblocks

Interaction
management

Centralized control

plane:

Istio

Control Plane (Pilot, Mixer, Citadel,
etc.)

Decentralized model: Each Eblock
contains

Discovery,

Auth,

and

Monitoring components

Pattern
implementation

Default support for pipes and filters;
other

patterns

require

complex

configuration or external logic

Support for multiple roles (manager,
worker, filter, etc.) with local task
distribution

Load balancing
mechanism

Sidecar proxy (Envoy) surrounds the
microservice, handling routing and
load balancing

Workload Manager inside each
Eblock with pseudo-random or
custom algorithms

Monitoring and
tracing

Centralized Mixer (pre-Istio v2) or
Telemetry input in later versions;
requires consistent configuration

Local Monitoring in Eblock with
REST API, ensuring uniform metrics
within the container

Security

and

authentication

mTLS between proxies, certification
through Citadel, granular RBAC
policy

Built-in

Auth

component

for

distributed

token

verification,

eliminating unnecessary nodes

Extensibility
and
customization

High, but requires deep knowledge of
Istio CRD and Kubernetes integration

Flexible pattern combinations since
roles are defined at the Eblock level

Overhead

Additional

sidecar

proxies,

centralized configuration storage,
control plane load

Higher resource consumption within
each Eblock than a standalone
service, but no central bottleneck

The classical service mesh model offers high
configurability for pipelines but complicates load
management and parallel execution. In contrast, the
Eblocks approach inherently supports these patterns
through a local Workload Manager, reducing reliance
on a centralized service mesh.

A key distinction of Eblocks is the reduced need for
constant interaction with a central controller. The
combination of service discovery and authentication
within each block removes some network constraints
but imposes additional requirements on the self-
sufficiency of each Eblock. Nevertheless, the proposed
model provides broader capabilities for integrating new
pattern combinations.

3. Practical implementation and performance

evaluation

An analysis was conducted on the methodology for
validating the functionality and efficiency of the
approach based on decentralized Eblock components,
as proposed by Nicolas-Plata A., Gonzalez-Compean J.
L., and Sosa-Sosa V. J. [1]. This method enables the
transformation of traditional application sets into
modern microservice-based systems by integrating
hybrid data processing patterns, optimizing pipeline
analysis, encryption, and result storage. The core idea
is to create a system capable of dynamically combining
the manager

worker and pipeline (pipes and filters)

patterns while also incorporating the divide-and-
conquer strategy for processing large-scale data
workloads.


background image

The American Journal of Engineering and Technology

163

https://www.theamericanjournals.com/index.php/tajet

The American Journal of Engineering and Technology

The primary objective of the system is to ensure
efficient pipeline processing and detailed analysis of
large text document datasets, followed by mandatory
encryption and storage in a centralized repository. To
achieve this, a set of specialized microservices was
developed, each performing specific functions within
the overall architecture. The system includes a
semantic analysis module for extracting keywords and
thematic attributes from documents, a risk assessment
component for evaluating data sensitivity levels,
symmetric encryption and decryption services to
ensure data security, a database for storing metadata
and processing results, a user interface for end-user
access, and a routing and request parameterization
service, which, in the experimental phase, operates as
an independent microservice despite its potential
integration into a unified Eblock.

Key system requirements include the ability to create
multiple instances of worker modules for parallel
document processing, minimizing manual configuration
through a declarative architectural approach, and
ensuring transparency in monitoring and security
processes via built-in service authorization. A
comparative analysis of approaches before and after
integrating

decentralized

Eblock

components

demonstrates

significant

improvements.

The

traditional method, where microservices were
manually interconnected via REST interfaces and
required additional scaling scripts, was outperformed
by the new approach, in which each component is
packaged into an Eblock with preconfigured Discovery,
Authentication, Monitoring, and Workload Manager
modules. This allows for seamless activation of
manager

worker roles or the implementation of

sequential pipeline schemes at the configuration level.

To validate the developed system, a modern hardware
and software infrastructure was utilized. The
containerization

environment

was

based

on

Kubernetes version 1.23 (or later), deployed on two
worker nodes and a single master node. The server
hardware included Intel Xeon processors with 12

16

logical cores and RAM ranging from 64 GB to 128 GB,
providing the required computational capacity. The
network environment was structured within an internal
10 Gbps network, and user interface access was
secured through port 443 (HTTPS). Data storage relied
on the distributed SkyCDS system, integrated with a
local file system for handling temporary data, ensuring
high reliability and fast data access.

The system deployment process consists of multiple

sequential stages. First, Docker images were built for
each microservice, embedding the necessary Eblock
components such as Discovery, Authentication,
Monitoring, and, when required, Workload Manager
for Manager or Worker roles. The next step involved
generating YAML configurations where a DevOps
engineer defined functional roles and interactions
between system components. A dedicated generator
processed these descriptions to create Kubernetes
manifests (Deployments, Services, ConfigMaps),
specifying container launch parameters and internal
component initialization. The final stage involved the
step-by-step deployment and initialization of Eblock
components while considering dependencies: the
Discovery module identified other instances through a
distributed hash table, Authentication components
exchanged tokens, and Monitoring aggregated metrics
for subsequent analysis.

System performance was assessed using a set of

metrics characterizing the overall architecture’s

efficiency and reliability. Key measurements included:

Response Time (RT): The time interval from

sending an HTTP request to receiving a response.

Pattern Response Time (PRT): The total

response time of all microservices involved, including
additional coupling time (CT) required for result
aggregation.

Throughput (TPS or docs/sec): The number of

documents processed per unit of time.

Infrastructure overhead: Evaluated in terms of

memory consumption, CPU load, and dynamic scaling
time for worker nodes.

Load tests were conducted using a dataset of 2,745 text
documents, with an average size of approximately 4 KB.
Some documents required computationally intensive
operations, such as semantic analysis and encryption
using long keys dependent on risk level. The experiment
involved three data processing scenarios:

1.

Linear pipeline processing: Each document

sequentially passed through semantic analysis, risk
assessment, encryption, and was finally stored in a
database.

2.

Manager

worker decryption: One Eblock acted

as the Manager, dynamically launching between one
and four worker nodes for parallel processing, while
other components functioned in filtering mode.


background image

The American Journal of Engineering and Technology

164

https://www.theamericanjournals.com/index.php/tajet

The American Journal of Engineering and Technology

3

.

Hybrid processing: A combination of sequential

pipeline processing and dynamic scaling at the
decryption stage using a worker Eblock pool.

Table 3 presents a comparison of the average
throughput (docs/sec) across different pattern
integration scenarios.

Table 3. Throughput (TPS) in different scenarios of pattern integration [1]

Scenario

Number of worker

nodes

Average TPS

(docs/sec)

Increase from

baseline (A)

A (Pipeline)

12.4

B (Manager–Worker)

2

15.8

+27%

B (Manager–Worker)

4

17.5

+41%

C (Hybrid)

2 (Decipher)

16.3

+31%

C (Hybrid)

4 (Decipher)

18.1

+46%

(Data averaged over multiple test runs, margin of error within ±2%.)

The analysis demonstrated a significant reduction in
overall system response time when transitioning from
a traditional linear pipeline to a hybrid solution
incorporating the manager

worker pattern. On

average, response times decreased by 23

27%,

primarily

due

to

efficient

parallelization

of

computational processes, particularly in decrypting
large files. Additionally, dynamic scaling of worker
nodes increased system throughput by approximately
35% compared to the baseline configuration.

The results indicate the strong potential of the
proposed approach for building scalable and resilient
microservice architectures. The combination of
multiple patterns and automated worker node
deployment accelerates data processing by up to 46%
compared to a linear pipeline without parallelization.

CONCLUSION

The study examined the limitations of existing service
meshes, which are primarily designed around the
pipeline processing pattern. It has been shown that this
approach is not always suitable for scenarios requiring
dynamic scaling and the combination of multiple
patterns, such as manager

worker and divide-and-

conquer. A decentralized service mesh implementation
has been proposed, where key tasks such as service
discovery, authentication, monitoring, and load
distribution are handled by the microservices
themselves in the form of Eblock structures.

Experimental results confirm that the described
architecture reduces configuration overhead, simplifies
the addition of new patterns, and significantly
accelerates data processing for large workloads.
Furthermore,

the

system

has

demonstrated

adaptability to varying load conditions through the
automatic deployment of additional worker instances.
It is important to note that decentralization imposes
specific resource requirements but effectively balances
the load, eliminating the bottleneck associated with a
centralized controller.

Future research directions include the development of
tools for the automatic selection of the optimal pattern
or their combination based on workload profiles, as
well as the expansion of the service mesh with
intelligent scaling and load forecasting capabilities.
Additionally, further integration of the Eblocks
mechanism with standard orchestration systems such
as Kubernetes and Docker Swarm is planned, along with
an in-depth evaluation of efficiency in real-world
industrial use cases.

REFERENCES

Nicolas-Plata A., Gonzalez-Compean J. L., Sosa-Sosa V.
J. A service mesh approach to integrate processing
patterns into microservices applications //Cluster
Computing.

2024.

Vol. 27 (6).

pp. 7417-7438.

Molnar V., Sabodashko D. Comparative analysis of
cybersecurity in leading cloud platforms based on the


background image

The American Journal of Engineering and Technology

165

https://www.theamericanjournals.com/index.php/tajet

The American Journal of Engineering and Technology

NIST framework //Social Development and Security.

2024.

Vol. 14 (6).

pp. 68-80.

Akerele J. I. et al. Improving healthcare application
scalability through microservices architecture in the
cloud //International Journal of Scientific Research
Updates.

2024.

Vol. 8 (2).

pp. 100-109.

Alboqmi R., Gamble R. F. Enhancing Microservice
Security Through Vulnerability-Driven Trust in the
Service Mesh Architecture //Sensors.

2025.

Vol. 25

(3).

pp. 914.

Wrona Z. et al. Scalability of Extended Green Cloud
Simulator //2024 International Conference on
INnovations in Intelligent SysTems and Applications
(INISTA).

IEEE. - 2024.

pp. 1-6.

Dhait S. et al. Analysis Of The Best Creational Design
Patterns In Software Development //2024 8th
International

Conference

on

Computing,

Communication, Control and Automation (ICCUBEA).

IEEE. - 2024.

pp. 1-5.

Li Y. et al. Serverless computing: state-of-the-art,
challenges and opportunities //IEEE Transactions on
Services Computing.

2022.

Vol. 16 (2).

pp. 1522-

1539.

Decimavilla-Alarcón D. C., Marcillo-Franco P. F.
Arquitectura

de

microservicios

basada

en

contenedores para despliegue ágil de aplicaciones IoT
en la nube //Revista Científica Episteme & Praxis.

2025.

Vol. 3 (1).

pp. 35-49.

Newman S. Monolith to microservices: evolutionary
patterns to transform your monolith.

O'Reilly Media.

- 2019.

pp. 5-20.

References

Nicolas-Plata A., Gonzalez-Compean J. L., Sosa-Sosa V. J. A service mesh approach to integrate processing patterns into microservices applications //Cluster Computing. – 2024. – Vol. 27 (6). – pp. 7417-7438.

Molnar V., Sabodashko D. Comparative analysis of cybersecurity in leading cloud platforms based on the NIST framework //Social Development and Security. – 2024. – Vol. 14 (6). – pp. 68-80.

Akerele J. I. et al. Improving healthcare application scalability through microservices architecture in the cloud //International Journal of Scientific Research Updates. – 2024. – Vol. 8 (2). – pp. 100-109.

Alboqmi R., Gamble R. F. Enhancing Microservice Security Through Vulnerability-Driven Trust in the Service Mesh Architecture //Sensors. – 2025. – Vol. 25 (3). – pp. 914.

Wrona Z. et al. Scalability of Extended Green Cloud Simulator //2024 International Conference on INnovations in Intelligent SysTems and Applications (INISTA). – IEEE. - 2024. – pp. 1-6.

Dhait S. et al. Analysis Of The Best Creational Design Patterns In Software Development //2024 8th International Conference on Computing, Communication, Control and Automation (ICCUBEA). – IEEE. - 2024. – pp. 1-5.

Li Y. et al. Serverless computing: state-of-the-art, challenges and opportunities //IEEE Transactions on Services Computing. – 2022. – Vol. 16 (2). – pp. 1522-1539.

Decimavilla-Alarcón D. C., Marcillo-Franco P. F. Arquitectura de microservicios basada en contenedores para despliegue ágil de aplicaciones IoT en la nube //Revista Científica Episteme & Praxis. – 2025. – Vol. 3 (1). – pp. 35-49.

Newman S. Monolith to microservices: evolutionary patterns to transform your monolith. – O'Reilly Media. - 2019. – pp. 5-20.