2024 10th International Conference on Web Research (ICWR)

A Client-Centric Consistency Model for Distributed

Data Stores using Colored Petri Nets
Ahmad Taghinezhad-Niar
Faculty of Electrical and Computer Engineering,
University of Tabriz, Tabriz, Iran
Email:[email protected]
2024 10th International Conference on Web Research (ICWR) | 979-8-3503-9498-6/24/$31.00 ©2024 IEEE | DOI: 10.1109/ICWR61162.2024.10533365

Abstract— The aim of increased reliability and performance requirements, each DS and application can choose a level of
in system architectures has resulted in the development of data consistency ranging from weak to strong.
replication systems. However, preserving consistency during
concurrent requests across various replicated data servers Implementing consistency models in large-scale
presents a significant challenge, especially when write distributed databases presents considerable challenges. Two
operations are involved. This issue highlights the necessity of primary models of consistency in distributed systems are
establishing a precise balance between system performance and Data-centric and Client-Centric (CC) consistencies. Data
consistency, particularly in the context of large-scale distributed consistency ensures that each read operation returns the result
databases. Large-scale distributed databases must navigate a of the last write operation. CC consistency, focusing on a
trade-off between performance and consistency. Client-centric single client perspective, offers a simpler approach. It
(CC) Consistency addresses this problem by guaranteeing encompasses four consistency models: Monotonic-reads,
consistency for single-client access to distributed data stores Read-Your-Writes, Monotonic-Writes, and Write-follow-
(DDS). CC consistency encompasses four distinct types of Reads [1]. CC consistency examines Read and Write
consistency assurances: 1) monotonic read, 2) monotonic write, operations from the client side of the application [3]. Strong
3) read your writes, and 4) write follow reads. In this paper, we consistency such as sequential consistency is provided among
present a formal model encapsulating CC consistency and its old data management systems which limit the system to
associated consistency models within distributed systems. Our
establish better performance, However, many new
model uses high-level ML functions to represent CC consistency
applications use weak consistency to provide less limitation
and its models. Integrity is verified CPN tools, ensuring
consistency. We demonstrate its effectiveness in discerning
on the system to achieve high performances, therefore, it is
supported CC consistencies with relevant inputs. This formal easy to achieve more performance in the requirements of weak
approach aids understanding and assists in optimizing consistency than strong consistency requirements. DS
distributed system performance while maintaining consistency. management needs strict ordering and synchronization
constraints on read-write operations in replicas. Hence, it
Keywords—Component; Distributed Systems; Replication; leads to lower performance but higher costs [1].
Client-Centric Consistency; Formal Modeling; Colored Petri Net.
Many distributed systems, including storage and
I. INTRODUCTION computation-based servers on the cloud, require consistency
among their replicated data system. Weak consistency has
The distribution of data across multiple remote servers different levels, and eventual consistency is the most basic
improves performance, availability, and reliability. In today's one, which means that all replicas eventually synchronize with
digital landscape, numerous business services have moved the latest data items. PNUTS and Dynamo are examples of
online, permeating our daily lives through applications like distributed systems that use a weaker consistency than strong
online shopping, banking, and messaging. These services consistency but stronger than eventual consistency [4].
require constant availability and millisecond response times.
There are different levels of weak consistency, with
Data Store (DS) replication serves a variety of functions at eventual consistency being the most basic one. This means
different scales and sizes. Replicated databases provide that all replicas will eventually synchronize with the latest data
resilience; the failure of one replica does not disrupt items. For example, PNUTS is a distributed system that has a
operations because the system seamlessly switches to a consistency level weaker than strong consistency but stronger
working replica. This redundancy is especially useful in large- than eventual consistency. Different distributed systems and
scale and geographically dispersed systems, where replicated applications have varying consistency and performance
servers (RSs) can efficiently distribute access loads. However, expectations [3].
the aim of performance in shared data stores necessitates a
tradeoff between consistency and performance, according to In [5], the authors proposed a model for implementing
the CAP theorem [1]. monotonic read consistency. Their work focused on creating
an adaptable model that only used one client-centric
No single replication solution fits all environments and consistency model which is monotonic-read consistency.
applications. Designing a replication approach requires However, in this paper, we extend the problem by proposing
careful consideration of various factors [2]. Replicating Data a model that can adapt to all four simultaneous client-centric
Stores (DS) can be a challenging task due to the absence of a consistency models, including Monotonic-Read consistency,
global clock. One of the primary difficulties is determining the Read-your-writes consistency, Monotonic-Write consistency,
order of write and read operations, which is essential for and Write-Follow-Read consistency. Therefore, our proposed
creating a consistently replicated database. As a result, model is suitable for applications that require strong
consistency issues may arise while using replicated data, and consistency [1] and provide a model-checking of the paper.
there is no one-size-fits-all solution. Depending on specific

309
979-8-3503-9498-6/24/$31.00 ©2024 IEEE
Authorized licensed use limited to: Universidad Federal de Pernambuco. Downloaded on June 04,2024 at 20:39:36 UTC from IEEE Xplore. Restrictions apply.
 In [6] a new consensus mechanism known as Proof of • Developing a model that avoids state-space explosion
Replication, and Resilient Data Availability to ensure data to enable efficient verification
consistency, integrity, and availability is presented. The
authors address privacy concerns, by introducing a privacy- • Presenting a system that avoids deadlock using
preserving data replication mechanism called Private Data model-checking proof
Replication. The study includes a thorough analysis of the
performance and security comparing it with other data II. BACKGROUNDS
replication solutions and discussing its potential applications In distributed systems, executing users' operations in a
and future research directions. However, the client-centric different order can lead to varying data item copies on
consistencies mentioned in this study remained as problem replicas, which can result in an inconsistent state of data items
that is going to be addressed in this paper. on distributed system (DS) replicas. To address this issue, four
client-centric consistencies come with different requirements
In [7], the authors have formalized the essence of client-
for data item copies [1].
centric isolation using Temporal Logic of Actions (TLA+) and
proposed it use for automated validation. They have shown the A. Monotonic Read
practical applicability of their work by reproducing examples One such consistency model is Monotonic Read
and demonstrating how to model check runtime traces and consistency, which ensures that if a client performs a read
algorithms with this formalization. However, TLA+ has operation on a specific data item (let's say X), any subsequent
complexity and a lack of graphical representation. Therefore, read operations by that same client on X should see either the
in this paper, we employed Colored Petri net (CPN) for same value or a newer value. The importance of monotonic
modeling client-centric distributed data-stores, enhancing read consistency lies in the fact that it guarantees that the client
system understanding and visualizing its behavior. will never see out-of-order data [1]. For instance, consider the
The CPN model is an effective tool for modeling complex case where a client connects to a mail server to check for a
systems and Distributed Systems like [8], [9]. For instance, new email. Later, if they connect to a different replica server
Kaid et al. [10] used CPN to represent Flexible Manufacturing (RS) that has not yet been updated with the new mail server,
Systems and their properties and behaviors such as conflict, they will not find the latest email as before. Monotonic read
sequencing, and synchronization. Similarly, Zhao et al. [11] consistency ensures that such a scenario does not occur, thus
demonstrated how Petri nets can simulate various aspects of providing a strong consistency guarantee to clients.
batch production processes, highlighting their potential B. Read-Your-Writes Consistency
benefits for scheduling industrial processes. Therefore, the
CPN model is appropriate for providing a strong and adaptable In distributed systems, executing users' operations in a
framework for modeling complex systems like distributed different order can lead to varying data item copies on
systems. replicas, which can result in an inconsistent state of data items
on distributed system (DS) replicas. To address this issue,
To provide consistency in distributed data stores, Jiang et different client-centric consistencies have been proposed that
al.'s [12] investigated MongoDB's tunable consistency come with varying requirements for data item copies [1].
models, which emphasizes the models' adaptability in
providing casual consistency guarantees. Crooks et al. [13] C. Monotonic Write Consistency
provided a client-centric perspective on transactional data One such consistency is Monotonic Write Consistency,
stores, specifically addressing weak notions of isolation and where a client must ensure that a write operation on a data item
consistency. However, they left gaps in their study by X must contain the effect of old writes of that data item. This
disregarding the four client-centric consistencies as a CPN means that if a client connects to replica 1 of a DS and runs an
model. update, then wants to update a newer version of that data item
but connects to replica 2, the previous update operations of
To this end, we present a formal model to represent CC
data item X must have been run on replica 2 likewise replica
consistency in distributed systems with four simultaneous
1.
consistency models that proceed with it. Our model represents
CC consistency and its models using high-level ML functions. D. Write follow reads Consistency
Our model’s consistency is guaranteed by using CPN tool's Another consistency is Write-follow-reads Consistency,
proposed automated verification functions. We demonstrate which guarantees that a write operation on a data item X will
the effectiveness of our methodology by providing relevant be performed on the data item that resulted from previous read
inputs for identifying supported CC consistencies. This operations of the client. It ensures that the client reads the
structured approach improves the performance of distributed same or a newer data item. For example, it can be used to
DS while maintaining CC consistency. guarantee that a client can only post a comment after reading
In Section 2, we provide background information on the the article [1].
problem addressed in this paper. We then describe related By adopting these consistencies, distributed systems can
works in Section 3 and propose a CPN system in Section 4 provide users with a more reliable and consistent experience,
that provides all four client-centric consistencies. In this ensuring that data items are accurately and consistently
section, we also present the formal verification of the replicated across replicas.
proposed model. Finally, we conclude our paper in Section 5,
highlighting the contributions of our work, which include: III. RELATED WORKS
• Modeling a system that guarantees all four client- Petri nets are a formal method introduced by Adam Petri.
centric consistencies Petri nets provide a mathematical and graphical approach to
describe, analyze, and verify concurrent systems. The
graphical interface of Petri nets consists of four fundamental

310

Authorized licensed use limited to: Universidad Federal de Pernambuco. Downloaded on June 04,2024 at 20:39:36 UTC from IEEE Xplore. Restrictions apply.
concepts: Transition, Place, Arc, and Token. Places, depicted perspective on transactional data stores, specifically
by an elliptical shape, represent the current state of a system. addressing weak notions of isolation and consistency. They
Tokens, which are inputs and outputs of places, represent the proposed a state-based formalization of isolation guarantees,
state of places. Transitions, depicted by a rectangular shape, which allows application developers to understand the
represent system actions or events. When a transition fires, it observable states of storage systems. This formalization also
changes the state of the system. Transitions can have gives implementors of storage systems maximum freedom in
conditions that enable them when certain conditions are meeting isolation levels. By focusing on client-centric
satisfied for changing the state of the system. The fourth perspectives, the study highlights the importance of
fundamental concept of Petri nets is Arcs, which are understanding the needs and requirements of the client when
connectors between places and transitions that carry tokens designing and implementing transactional data stores.
among them [14]. However, they did not address the four client-centric
consistencies and did not model it as a CPN model. We will
In [5], a model with monotonic read Client-Centric (CC) address these gaps in this paper.
consistency is proposed, which is a model of a system with a
monotonic read. However, the classical Petri net model can be Jiang et al. [12] investigated MongoDB's tunable
extended through the use of the colored Petri net. The colored consistency models and implementation. The study
Petri net adds an artificial intelligent programming language particularly emphasized MongoDB's adaptability in providing
called ML as an important upgrade to the classical one. In casual consistency guarantees. The authors explored various
colored Petri nets, Colorsets take the place of tokens. Color aspects such as causal consistency, transactions with
sets are programming data types that enable us to program snapshots, and ongoing improvements for future releases.
them. The strongest tool that supports colored Petri nets is This paper discusses achieving optimal consistency in big data
CPN-Tools, which is used for modeling concurrent systems stores, including cluster-wide logical clocks and causal
and validation of state space [14]. consistency in MongoDB [12]. However, our study proposes
a general client-centric consistency in DDS based on a CPN
In the context of client-central consistencies in large-scale model.
data stores, Zhu et al. [3] have analyzed CC consistencies to
obtain their implementation details using mathematics. Their According to [16], CPN-Tools, a modeling language and
work was based on a single user and single operation toolset based on Colored Petri Nets, has proven to be an
consistency. However, our paper introduces an effective means of resource monitoring in federated clouds. In
implementation of client-centric consistency that involves their study, the authors propose a model for scheduling
concurrent and multiple client operations. We use CPN-Tools applications in a federated cloud environment. The potential
to present a novel model of CC consistency that supports all of CPN-Tools in modeling complex distributed systems, such
of the CC consistency models upon each other. Additionally, as scheduling workflow tasks on multi-clouds with different
an analysis of the modeled system state space is presented. objectives, has been demonstrated to ensure the adherence to
service level agreements. Overall, this research highlights the
CPN have gained popularity as a powerful modeling tool importance of employing advanced modeling techniques,
in various fields. One recent application of CPN is the such as CPN-Tools, in facilitating the management and
simulation modeling approach proposed by Drakaki et al. optimization of cloud computing resources [16], [17].
[15]. This approach aims to evaluate the impact of inventory
management decisions on the efficiency of a multi-stage serial To this end, the CPN model, specifically the colored Petri
supply chain under normal operating conditions and in the net model, has proven to be a highly effective formal method
case of interruptions. The authors constructed a CPN-based for modeling and analyzing concurrent systems across a
model of the inventory management system and analyzed two diverse range of domains. As such, we have utilized CPN-
types of setups, traditional and information-sharing, using tools in our proposed work to model a system with client-
hierarchical timed CPN. centric consistency and ensure that the four key client-centric
consistencies (monotonic read, read your writes, monotonic
Kaid et al. [10] utilized CPN as a modeling tool to depict write, and write follows reads) are guaranteed in large-scale
the properties and behaviors of Flexible Manufacturing data stores.
Systems, including conflict, sequencing, and synchronization.
This allows for the effective representation of complex system IV. PROPOSED MODEL
interactions. The state-space analysis and model-checking are
not taken into account in their proposed model. however, we In this section, we will discuss the behavior of the
present a state-space analysis for our proposed model. presented model in the given scenario. The order in which
users execute their operations on the data items in DDS can
Zhao et al. [11] examined the effectiveness of Petri nets result in different copies of the data items on replicas. In some
and iterative greedy algorithms for batch production cases, this can even lead to inconsistent situations of data
scheduling. The paper investigates the use of Petri nets to items on DDS replicas. Therefore, there are four client-centric
simulate many aspects of batch production processes, consistencies, each with different expectations of data item
including formulas, recipes, and work orders, as well as copies [1]. It is important to note that client-centric
iterated greedy algorithms as a search approach for consistency comes with eventual consistency.
optimization problems. The integration of these two systems
for heuristic scheduling is also discussed, as well as their A. Guaranteed Client-Centric Consistency Model
potential advantages for industrial production process In a DDS environment, users execute atomic operations on
scheduling. data items, categorized as either 'Write' or 'Read' actions.
These data items are distributed across multiple replicas
Crooks et al. [13] studied the challenges of offloading data within the large-scale DDS infrastructure. To denote these
to the cloud and the necessary background on correctness for
transactional data stores. They provided a client-centric

311

Authorized licensed use limited to: Universidad Federal de Pernambuco. Downloaded on June 04,2024 at 20:39:36 UTC from IEEE Xplore. Restrictions apply.
operations and their corresponding effects, we adopt a colset VALUE=record vari:VAR* vali:INT*tstamp:TSTAMP;
notation where: colset VALUES =list VALUE;
colset CLIENT=record
• A 'Read' operation on data item X is represented id:ID*values:VALUES*server:SERVER;
in. (1), indicating the retrieval of value V from replica 1: colset CLIENTS=list CLIENT;
colset OPR=with READ|WRITE;
colset CHANGELOG=record vari:VAR* vali:INT*client:INT;
= (1) colset CHANGELOGS=list CHANGELOG;
colset OPERATION=record opr:OPR* cvar:VAR* cval:INT*
• A 'Write' operation on data item X is denoted in clientid:INT;add:bool;
(2), signifying the update of the value of X on replica 1 colset OPERATIONS=list OPERATION;
to V. colset LOG=record tstamp:TSTAMP*clog:CHANGELOGS;
colset LOGS=list LOG;
colset DATABAS=record server:SERVER*logs:LOGS;
= (2) colset DATABASE=list DATABAS;
The CC model presented here uses several color sets to
In the presented model shown in Fig.1, we assume that represent various aspects of the system. For example, the
DDS is composed of distributed replicated DSs, where each "colset OPR" represents an enumerated type that indicates the
store saves data items and data logs. The data log contains type of operation (read or write) performed on a data item. The
updates of data items with a global timestamp. Each client has "colset OPERATION" is a record that contains fields for the
a history of data items that it has read. The RS ID is the data type of operation, the name of the data item, the value related
store that a user is connected to. The list of operations to the operation, and the index number of the client executing
represented by client operations contains 'Read' and 'Write' this operation. The "colset VALUES" represents a list of type
operations on the data items performed by each client. VALUE, which is defined as a record with fields for the name
Operation and client log composition are considered sessions, of the data item, the value of the data item, and the timestamp
which represent an order of operations on a specific data item at which this value for the data item was seen. The "colset
to be run in DDS by a client. DDS runs client operations while CLIENT" represents a record with fields for the client ID, a
respecting client-centric consistency. list of values of data items along with their timestamps, and an
B. COLOR SETS index of RS related to visited values for data items. The "colset
CHANGELOG" is a record that denotes the name of the data
The CC model is a fundamental concept in distributed item, the related value of the data item, and the ID of the
systems that ensures that multiple processes can access shared related client. The "colset LOG" is a record that represents the
resources without interference. In this context, color sets are timestamp and a list of data items, their values, and related
used to represent the different data structures and operations clients' IDs. Finally, the "closet DATABASE" represents a
that constitute the CC model. The following describes the record with two fields: the ID of RS and a list of related logs
color sets for a presented CC model.

Fig. 1. Proposed model of Client Centric Consistency

312

Authorized licensed use limited to: Universidad Federal de Pernambuco. Downloaded on June 04,2024 at 20:39:36 UTC from IEEE Xplore. Restrictions apply.
for this RS. Overall, these color sets are used to represent the V. STATE-SPACE ANALYSIS OF PROPOSED MODEL
various data structures and operations used in the CC model, This section provides a concise overview of the model's
ensuring that multiple processes can access shared resources state space. Furthermore, a new function for detecting
without interference. deadlock in the paradigm above is proposed to demonstrate its
C. Places and Fundamental Functions of the proposed CC correctness. Each state of the system reveals system properties
Model over time, and DeadMarking is a state that occurs when the
model is unable to fire any transitions. In other words, dead-
The following part presents a comprehensive description
marking nodes in the state space graph are terminal nodes
of the proposed model's behavior in a given scenario. The
without any outgoing arcs. Table II presents the state space for
order of execution of users' operations plays a crucial role in
the given case. By extensively extracting the state space, we
determining the data item copies on replicas. Fig.2 depicts the
can perform various model-checking scenarios in the model.
schema of the presented scenario for three replicated DDS. To
We have devised a function called "IsDeadlockFree" shown
illustrate the model's functionality, we have provided a
in Algorithm 1 to examine the state space for any deadlock. It
scenario in Table I, which aims to simplify the subject matter.
evaluate all the items in the deadmarking list to ascertain
The first operation of the scenario is depicted in Fig.2, whether the place of operations has any marks while the
where a client sends a write operation to server 1 along with system is disabled or not. The ML code of IsDeadLockFree is
its timestamp and operation details. Table I showcases the presented at the following, and its output yields false.
details of the given scenario, which involves a replicated DDS
In the context of state-space analysis, Fig.3 illustrates the
comprising three servers. Each replica logs the operations
reachability graph generated using CPN tools. The dead
performed by clients. There are three clients, each performing
marking has been computed, and a path has been identified
operations on RSs, which are connected to the DDS. Table I(a)
between the initial scenario (first node) and the first dead
displays the partial order of operations performed on each RS.
marking. The behavior of the model can be examined to detect
On the other hand, Table I(b) illustrates the data item's value
any potential inconsistencies. Additionally, Table II presents
in each replica over time. The letter ‘R’ represents the read
relevant findings that can inform further analysis of the model.
operation, while the letter ‘W’ denotes the write operation. In
In addition, in this figure, the output of running
the case of write operations, the value following the ‘=’ sign
IsDeadlockFree is provided.
represents the updated value of the data item. Conversely, for
read operations, the value represents the value returned from
the RS. It is essential to understand that client operations are
related to an RS, which is represented by the letter ‘S’. For
example, = 5 in Table I(a) indicates C1's initial
operation, which assigns a value of 5 to the data item X. The
letter ‘TS’ represents a timestamp, which is used as a time
marker.
In the presented study, a colored Petri net model is utilized
to analyze a given scenario. Table I(a) depicts the scenario,
where Client 1 writes a value of ‘5’ on data item X in RS1.
Later, when the client is directed to RS2 and attempts to
update the data item by adding ‘2’, the system behavior differs Fig. 2. Presented replicated DS schema
based on whether or not the DDS supports Monotonic Write. TABLE I. EXAMPLE OF MONOTONIC READ AND READ YOUR WRITE
Specifically, if the DDS does not support Monotonic Write, it CONSISTENCY
reads the value of data item X from RS2, which is ‘0’, and Client Time
then updates X by adding ‘2’ to it. As a result, the updated
value of X becomes ‘2’. However, in the presented model that C1: = =
= +
supports Monotonic Write, the result would be ‘7’. C2: = = =
Furthermore, assuming that Client 1's third operation C3: = =
involves reading from RS1, the system behavior also differs = −
based on whether or not the DDS supports Read-Your- a) Client operations
Writes consistency. If the DDS does not support Read-Your-
Timestamp
Writes consistency, it would return ‘5’ as the last value of data Server
0 1 2 3 4 5
item X from RS1. In contrast, the presented model supports RS1 X:0 X:5 X:6
Read-Your-Write consistency, and thus the result would be Eventual
RS2 X:0 X:7 Consistency
X:6
‘7'. This is because Client 1 has already written the value of '7' RS3 X:0 X:6 X:6
for data item X on RS2. When it sees that RS1 has an older b) Status of RSs by timestamp
value of X, it reads the data item X from its local log. Lastly,
assume that Client 3 reads X from RS2, which is ‘7’. TABLE II. STATE-SPACE REPORT OF THE PROPOSED MODEL.
Afterward, the client is directed to RS3, which contains ‘0’ for
data item X. The client updates the data item X by subtracting Report of the State Space Graph
Nodes Number 18648
1 from its value. In the presented system, which follows Arcs Number 21684
Write-Follows-Read consistency, the result of this operation Generation Time of the Graph 32 Second
would be ‘6’. However, in the system that does not follow this Generation status of the Graph Full
consistency, the result would be -1. Count of Dead Markings 4512
Nodes of Dead Markings 18648,18647,18646,18645,...

313

Authorized licensed use limited to: Universidad Federal de Pernambuco. Downloaded on June 04,2024 at 20:39:36 UTC from IEEE Xplore. Restrictions apply.
 Algorithm 1: IsDeadlockFree algorithm [5] A. Taghinezhad and S. Pashazadeh, “Modelling and analysis of the
monotonic read consistent distributed system using coloured Petri net,”
fun IsDeadLockFree ():bool= 2016 8th Int. Conf. Inf. Knowl. Technol. IKT 2016, pp. 85–90, 2016,
let val LD=ListDeadMarkings () doi: 10.1109/IKT.2016.7777791.
val length= List.length (LD)
[6] K. Jannes, “Secure and resilient data replication for the client-centric
val i= ref 0
decentralized web,” in Proceedings of the 23rd International
val n= ref ~1
Middleware Conference Doctoral Symposium, in Middleware ’22
val opr= ref [] Doctoral Symposium. New York, NY, USA: Association for
val rs= ref true Computing Machinery, 2022, pp. 1–4. doi: 10.1145/3569950.3569961.
in while !i< length do(
n:= List.nth (LD,!i); [7] T. Soethout, T. van der Storm, and J. J. Vinju, “Automated Validation
opr := List.hd (Mark.CSModel'OPERATIONS 1 (!n)); of State-Based Client-Centric Isolation with TLA+,” in Software
if !opr<>[] then rs:= true Engineering and Formal Methods. SEFM 2020 Collocated Workshops,
else rs:= !rs; L. Cleophas and M. Massink, Eds., Cham: Springer International
Publishing, 2021, pp. 43–57.
i:= !i+1 );
if !rs=false then false [8] A. Taghinezhad-Niar, S. Pashazadeh, and J. Taheri, “QoS-aware online
else true scheduling of multiple workflows under task execution time
end uncertainty in clouds,” Cluster Comput., vol. 8, 2022, doi:
10.1007/s10586-022-03600-8.
[9] A. Taghinezhad-Niar, S. Pashazadeh, and J. Taheri, “Workflow
scheduling of scientific workflows under simultaneous deadline and
budget constraints,” Cluster Comput., vol. 24, no. 4, pp. 3449–3467,
Dec. 2021, doi: 10.1007/s10586-021-03314-3.
[10] H. Kaid, A. Al-Ahmari, Z. Li, and W. Ameen, “An Improved Synthesis
Method Based on ILPP and Colored Petri Net for Liveness Enforcing
Controller of Flexible Manufacturing Systems,” IEEE Access, vol. 10,
no. June, pp. 1–1, 2022, doi: 10.1109/access.2022.3186287.
[11] Z. Zhao, S. Liu, M. Zhou, D. You, and X. Guo, “Heuristic Scheduling
of Batch Production Processes Based on Petri Nets and Iterated Greedy
Algorithms,” IEEE Trans. Autom. Sci. Eng., vol. 19, no. 1, pp. 251–
261, 2022, doi: 10.1109/TASE.2020.3027532.
[12] X. Jiang, H. Wei, and Y. Huang, “Tunable causal consistency:
Specification and implementation,” in 2022 IEEE 28th International
Conference on Parallel and Distributed Systems (ICPADS), 2023, pp.
169–176.
[13] N. Crooks, “A client-centric approach to transactional datastores,” in
Proceedings of the 2021 International Conference on Management of
Data, 2021, pp. 3–5. doi: 10.1145/3448016.3461471.
[14] R. J. Machado, K. B. Lassen, S. Oliveira, M. Couto, and P. Pinto,
“Requirements validation: Execution of UML models with CPN tools,”
Int. J. Softw. Tools Technol. Transf., vol. 9, no. 3–4, pp. 353–369, 2007,
Fig. 3. Reachability graph of a client-centric consistency at CPN-tools doi: 10.1007/s10009-007-0035-0.
[15] M. Drakaki and P. Tzionas, “A colored petri net-based modeling
VI. CONCLUSION method for supply chain inventory management,” Simulation, vol. 98,
no. 3, pp. 257–271, 2022, doi: 10.1177/00375497211038755.
Concurrent systems have long been a focus of research. In [16] A. Taghinezhad-Niar, T. Javadzadeh, and L. Farzinvash, “Modeling of
this study, we present a novel method of creating a client- resource monitoring in federated cloud using Colored Petri Net,” in
centric consistency system that uses the formalism of colored 2017 IEEE 4th International Conference on Knowledge-Based
Petri nets for modeling. Our studies highlight the high efficacy Engineering and Innovation (KBEI), IEEE, Dec. 2017, pp. 0577–0582.
doi: 10.1109/KBEI.2017.8324866.
of using colored Petri nets to mitigate design defects in
[17] A. Taghinezhad-niar, J. Taheri, and S. Member, “Reliability, Rental-
concurrent systems, hence improving their reliability and Cost and Energy-Aware Multi-Workflow Scheduling on Multi-Cloud
resilience. Our research extends beyond definition, Systems,” IEEE Trans. Cloud Comput., pp. 1–12, 2023, doi:
proceeding into the thorough examination of the state space 10.1109/TCC.2022.3223869.
and detecting probable deadlocks within the model. This
analysis not only ensures the accuracy of our implementation
but also provides crucial insights into the system's operational
dynamics. In the future, we aim to enhance our client-centric
consistency system by supporting strong consistency, such as
sequential consistencies in DDS. This future model will
facilitate the system to handle a larger number of clients and
more complex operations.
REFRENCESS
[1] A. S. Van Steen, Maarten and Tanenbaum, Distributed Systems, 4th ed.
Maarten van Steen Leiden, 2023.
[2] B. Kemme, R. J. Peris, and M. Patino-Martinez, DataBase Replication.
Morgan & Claypool Publishers, 2010.
[3] Y. Zhu and J. Wang, “Client-centric consistency formalization and
verification for system with large-scale distributed data storage,”
Futur. Gener. Comput. Syst., vol. 26, no. 8, pp. 1180–1188, 2010, doi:
10.1016/j.future.2010.06.006.
[4] B. F. Cooper et al., “PNUTS,” Proc. VLDB Endow., vol. 1, no. 2, pp.
1277–1288, Aug. 2008, doi: 10.14778/1454159.1454167.

314

Authorized licensed use limited to: Universidad Federal de Pernambuco. Downloaded on June 04,2024 at 20:39:36 UTC from IEEE Xplore. Restrictions apply.