SQL vs. NoSQL Databases: A Systematic Review of Performance, Scalability, and Consistency Trade-offs – with Distributed Systems and Cloud Environment Perspectives

SQL vs. NoSQL Databases: A Systematic Review of Performance, Scalability, and Consistency Trade-offs – with Distributed Systems and Cloud Environment Perspectives

with Distributed Systems and Cloud Environment Perspectives

Shenamae J. Torillo

Department of Computer Engineering University of Southern Mindanao Kabacan, North Cotabato

Abstract – Choosing a proper database management system has been elevated to an essential architectural decision in an era when software engineering encompasses cloud computing, IoT integration, real-time analytical capabilities, and AI-driven data processing flows. This paper presents a systematic literature review on SQL versus NoSQL databases, focusing on aspects such as performance, scalability, and consistency trade-offs. A comprehensive literature search process, using the Kitchenham (2004) approach, was carried out on around

340 records gathered from six sources. The final set of articles included 18 scholarly works, spanning from 2011 to 2025, categorized into four thematic clusters. NoSQL databases were discovered to be superior to SQL databases in write-intensive, distributed, and cloud-native environments, whereas SQL databases excel in query efficiency and consistency enforcement. Empirical evidence supports the CAP theorem trade-off by demonstrating that higher consistency leads to decreased throughput and latency in NoSQL databases. There is no one-size-fits-all solution; the best choice is dictated solely by the workload type. The four research questions addressed in this study provide valuable insights, with the most significant gap being the lack of investigations into hybrid SQL/NoSQL configurations and their failure consistency modes.

Keywords – SQL; NoSQL; relational databases; document databases; performance benchmarking; scalability; consistency models; CAP theorem; ACID; BASE; distributed systems; cloud computing; systematic literature review

1. INTRODUCTION

   The rise in popularity of cloud computing, real time analysis, Internet of Things (IoT) sensor data, and artificial intelligence applications has significantly transformed the requirements of database systems. The ingestion and processing of massive volumes of data in record time have made database engineers rethink which model best fits their application scenario [1][3]. It is not an easy task; the migration of database models is often expensive, costing millions of dollars in a comprehensive system overhaul, and making the wrong choice is costly and leaves lasting architectural impacts [5][11].

   Two types of DBMS can be considered in contemporary system development. First, SQL databases like MySQL and PostgreSQL rely on the relational model with strict table schemas and have ACID (atomicity, consistency, isolation, and durability) properties. The latter means that transactions should always be consistent and reliable. These types of databases are usually employed in the financial, ERP, and other regulation-compliance industries [1][7]. Second, NoSQL databases like MongoDB (document-oriented), Cassandra (column-family), and Redis (key-value) were invented to overcome the drawbacks of the relational model when manipulating large-scale datasets without specific schemas. These DBMS focus more on scalability and availability by using distributed storage technologies, which makes them useful in cloud computing, social networks, IoT, and big data analysis [12][18].

   	CRUD speed)	processing
   NoSQL	Scalability	Appropriateness
   databases	(sharding	of database for
   (MongoDB,	capabilities,	large-scale,
   Cassandra,	horizontal	native cloud
   Redis)	scalability,	apps
   	replication
   	performance)
   Deployment	Consistency	Appropriateness
   environment	trade-offs	for transactional
   (on-premises,	(ACID vs.	and
   cloud, cluster)	BASE, CAP	compliance-
   	theorem,	focused
   	tunable	applications
   	consistency)

   While both methods have been broadly employed, choosing the right database is still a difficult issue. Research on benchmarking delivers contradictory, methodologically influenced results [2][14], and most of the existing reviews focus only on one aspect or have questionable criteria for the selection of papers. In order to fill these gaps, this paper analyzes 18 papers employing the SLR model by Kitchenham from 2004 and analyzes trade-offs between performance, scalability, and consistency while estimating gaps in literature. Graph databases like Neo4j and cloud-based managed database systems like Amazon DynamoDB are not considered due to their specific applications.

2. METHODOLOGY

   1. Research questions

      The following research questions have been formulated to organize the structure of the present paper:

      1. RQ1: What differences arise in the performance of SQL and NoSQL databases in write-heavy, read-heavy, and hybrid CRUD (create, retrieve, update, delete) environments?

      2. RQ2: Which database would be the optimal selection concerning scalability on distributed and cloud infrastructures?

      3. RQ3:RQ3: How does the configuration of consistency models affect the efficiency and integrity of NoSQL databases according to the rules of the CAP theorem?

      4. RQ4: What are the research gaps in the comparative analysis between SQL and NoSQL?

   2. Conceptual framework

      The conceptual framework is outlined in TABLE I below. The independent variable sets of interest here are the SQL databases (MySQL, PostgreSQL) and NoSQL databases (MongoDB, Cassandra, Redis), which will be compared using three factors: performance, scalability, and consistency. The deployment environment will moderate the analysis process. v

      Independent variables	Comparative dimensions	Dependent outcome
      SQL databases	Performance	Appropriateness
      (MySQL,	(writing	of database for
      PostgreSQL)	performance,	write-heavy,
      	reading latency,	distributed

      TABLE I. Conceptual Framework for SQL vs. NoSQL Comparative Review

      Note: The criteria for comparison are consistent with the four research questions. lt is the dependent variables that form the basis for recommendations in Section V.

      In this study, the methodology employed is the systematic literature review technique as proposed by Kitchenham (2004) [19]. This technique is widely used in the software engineering and computer science domains. It entails a four-step process: searching, selection, evaluation, and synthesis.

   3. Search protocol

      A systematic search was conducted from January to March 2025 using IEEE Xplore, arXiv, MDPI Open Access, ACM Digital Library, SpringerLink, and ResearchGate with key terms including "performance benchmarking of SQL vs NoSQL databases," "scalability and distributed computing with NoSQL databases," "models of consistency in databases CAP theorem," "performance evaluation of relational databases CRUD operations," "comparative study of MongoDB and MySQL," "Cassandra consistency benchmarking," and "cloud database performance evaluation." About 340 documents were obtained at the start.

   4. lnclusion and Exclusion Criteria

      Studies were selected if they met the following criteria: (1) were peer-reviewed and published in reputable journals, conferences, or preprints; (2) studied the performance, scalability, or consistency of SQL and/or NoSQL databases; (3) were published between 2010 and 2025; (4) were written in English; and (5) had clear methodology sections. Studies that solely dealt with mobile databases, database security without performance studies, and those lacking reproducibility were excluded.

   5. Quality Assessment

      Five quality criteria were applied to evaluate the quality of the paper prior to its inclusion in the systematic review for all studies that successfully passed the full-text screening phase. The five criteria, along with their purpose and weight during synthesis, are displayed in TABLE II below. Papers meeting all criteria were considered high quality and assigned high weight for analysis purposes.

      Quality Criteria	Purpose	Application Method	Weight
      Relevance	Requires that the study assesses SQL or NoSQL efficiency, scalability, or consistency.	Requires comparison between at least one SQL database management system and one NoSQL database management system or assessment of consistency or scalability separately.	High
      Methodological Transparency	Requires reproducibility and interpretability of experimental findings.	Requires reporting of hardware details, dataset size, and workload parameters.	High
      Publication authenticity	Enhances the legitimacy of the academic study and peer review process	Favorable journals: IEEE, ACM, MDPI, Springer; arXiv is allowed if frequently cited	Medium
      Technology novelty	Remains relevant in the context of modern implementation	Published between 2010-2025; seminal papers published	Medium

      TABLE II. Quality Assessment Criteria

      Citation effect

      Signals community recognition in follow-up papers

      Studies cited by at least two later papers have greater synthesis weight

      Low

   6. Screening based on F.PRISMA guidelines

      The process of screening the studies is shown in TABLE III below, following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [20].

      TABLE III. PRISMA-Aligned Study Selection Flow

      Stage	Action	Record count
      Identification	Database searches: IEEE Xplore, arXiv, MDPI, ACM, Springer, ResearchGate	-340 records identified
      Deduplication	Cross-repository duplicate removal	-280 records retained
      Title/abstract screening	Applied inclusion/exclusion criteria	-60 records selected
      Full-text eligibility	Full texts assessed for methodological quality and relevance	-30 full texts reviewed
      Excluded (full-text)	Excluded: mobile-only, security-only, non-DBMS, insufficient methodology	-12 studies excluded
      Final included	Met all criteria; organized into 4 thematic groups	18 studies included

   7. Thematic Classification and Data Extraction

   Four thematic classifications were conducted with respect to all 18 studies that have been analyzed, which are listed below:

   (1) Comparison of SQL and NoSQL architecture and design concepts [1][2][3][4][5]; (2) Performance and CRUD

   benchmarking [6][7][8][9][10][11][14]; (3) Scalability and distributed systems [12][13][15]; and (4) Consistency models and CAP theorem analysis.

   TABLE IV. Extracted Data Matrix – 18 Included Studies

   #	Auth or	Focus area	Contributi on	Key finding	Qualit y
   1	[1]	SQL/NoSQ	SLR of	NoSQL	High
   		L survey	architectur	improves
   			es	scalability
   				; SQL
   				ensures
   				ACID

   				deployme nts
   1 2	[12]	Scalability architectur e	Architectur e review	NoSQL designed for horizontal scaling	High
   1 3	[13]	SQL partitioning	OLTP partitioning model	Partitionin g improves distributed SQL performan ce	High
   1 4	[14]	Value-length benchmark	Data-size impact study	Performan ce varies substantial ly with payload size	Mediu m
   1 5	[15]	Cluster evaluation	Low-power cluster study	NoSQL better under distributed load; energy varies	High
   1 6	[16]	Consistenc y models	CAP theorem taxonomy	Most NoSQL systems favor AP; eventual consistenc y default	High
   1 7	[17]	Cassandra consistency	Performanc e vs. consistency	Strong consistenc y measurabl y reduces throughpu t	High
   1 8	[18]	NoSQL theory	Big data classificati on	NoSQL adopts BASE; prioritizes availabilit y over ACID	Mediu m

   				consistenc y
   2	[2]	Performanc e benchmark	Early empirical benchmark	NoSQL scales better; consistenc y reduced	High
   3	[3]	Big data comparison	Analytical comparison	Workload type determine s best database	Mediu m
   4	[4]	Performanc e evaluation	Experiment al benchmark ing	NoSQL faster on writes; SQL on structured reads	High
   5	[5]	NoSQL decision survey	Decision framework	No universal best DB; applicatio n-driven selection	High
   6	[6]	CRUD benchmark ing	CRUD performanc e study	NoSQL better in distributed writes	High
   7	[7]	SQL optimizatio n	Query optimizatio n analysis	Optimizati on significant ly improves SQL performan ce	High
   8	[8]	Document DB study	Performanc e comparison	MongoDB better for unstructur ed data	High
   9	[9]	Experiment al benchmark	Real-world insert testing	MongoDB faster in insert-heavy workloads	Mediu m
   1 0	[10]	CRUD replication	Replication performanc e study	Async replication improves scalability ; risks stale reads	High
   1 1	[11]	Cloud DB evaluation	Cloud benchmark ing	NoSQL outperfor ms SQL in cloud	High

   Quality Score depends on quality of the studies included (High=3; Medium=2). No studies with quality scores lower than 2 were taken into consideration.

3. LITERATURE REVIEW

   A. SQL vs. NoSQL: Basics and Comparison of Architectures

   The basic differentiation of SQL from NoSQL database management systems lies in their differences regarding the data model and capability to provide consistency. Comparative analysis of performance showed the superiority of NoSQL databases concerning scalability because of their distributed architecture, whereas the advantages of SQL databases in performing transactions through ACID properties become evident [1]. Thus, the first study that empirically proved the contrast mentioned before indicated the superior scalability of NoSQL database systems under load [2].

   It has been established that workload features, not paradigm features, play a determining role in deciding on the appropriate database choice [3]. SQL has proved its advantage over NoSQL when it comes to high query complexity and transactions; meanwhile, NoSQL performs better with data ingestion and variable schema [3]. This hypothesis has been tested and proved empirically since it was observed that the performance of NoSQL databases significantly exceeded that of MySQL, especially concerning write tasks, while the latter worked better with complex read operations [4]. Even further research has concluded that an evaluation framework accounting for the heterogeneity of NoSQL paradigm could help make better database decisions [5].

   Collectively, all five papers prove that there is no universal paradigm that is always the best choice. Selection of databases must be based on how suitable the features of the application are for the database system design [5].

   In the cluster of performance benchmarking research, there are seven studies. In general, write performance is greater when using NoSQL, while read performance is greater in the case of SQL usage. At the same time, there are some important exceptions to this trend. Thus, according to the benchmark results comparing MySQL and CouchDB, the latter showed superiority in the case of distributed writes, and MySQL was more effective in the case of structured reads [6]. Moreover, another study demonstrated that query optimization through indexing, query modification, and connection pooling would greatly improve the write performance of MySQL which implies that the advantages of NoSQL may be overestimated during unoptimized SQL-based testing [7].

   The superior efficiency of MongoDB was proven with respect to unstructured and semi-structured data because of lower join overhead costs, the latter being regarded as the major cause for better performance of MongoDB [8]. Moreover, it was proven that MongoDB is more efficient compared to MySQL in insert-based data loads because of higher throughput depending on database size [9]. Replication

   turned out to be another factor influencing performance, since asynchronous replication improved write throughput for NoSQL while inducing stale reads, an effect ignored by all papers [10].

   The findings reveal that NoSQL significantly outperforms SQL in cloud-native scenarios thanks to native scalability [11]. The study conducted using the most scientific methodology demonstrates that data payload size matters when evaluating the performance of databases, and equal small payloads may provide misleading results [14].

   C. Scalability and Distributed Systems

   There are well-known architecture design techniques for scaling horizontally in NoSQL databases: sharding, consistent hashing, and replication across multiple nodes [12]. This has been experimentally proved in a low-power cluster environment, yet again showing the superiority of NoSQL over SQL in distributed systems, emphasizing the energy efficiency difference between them too [15].

   A strong counterpoint that intelligent partitioning due to skewed distribution improves the performance of SQL in OLTP databases, thus proving that SQL scalability issues are only due to its immature design and not its limitations per se [13].

   D. Consistency Models and CAP Theorem Evaluation

   Well-known NoSQL databases have been thoroughly evaluated in regards to CAP theorem [20], finding out that most of them have been developed with availability and partition tolerance as primary principles that ensure eventual or customized consistency as opposed to full consistency [16]. The reason behind such a choice was clear – AP-based databases traded consistency window duration in favor of availability in the face of unavoidable network partitions in globally distributed systems [16].

   Contextualization within History has been successfully performed by determining the emergence of the BASE architecture (Basically available, soft state, Eventual consistency) as a direct counterpart of the ACID one in NoSQL environment [18]. The trade-offs associated with the CAP theorem have been quantified by tweaking consistency settings in Cassandra and evaluating throughput/latency numbers obtained [17].

4. COMPARISON OF THE RESULTS

   On the basis of all the thematic groups, the analysis of the results concerning all three dimensions of the research

   is provided in this part, together with the assessment of the quality of benchmarking methodology.

   1. Matrix of comparative results

      The matrix of comparative results on the basis of the literature review is provided in TABLE V.

      TABLE V. Comparative Summary – SQL vs. NoSQL Databases

      Dimension	SQL databases	NoSQL databases	Key refs
      Data model	Relational tables; fixed schema; strong normalization	Document, key-value, column-family; flexible schema- optional	[1][3][18]
      ACID compliance	Full ACID guarantee; strong transactional consistency	Partial; BASE model – basically available, soft state, eventually consistent	[1][16][18]
      Write performance	Moderate; degrades under high-volume distributed writes	Faster for bulk inserts and distributed writes; scales with volume	[4][6][9]
      Read performance	Excellent for complex multi-table joins and aggregation	Fast for key/document lookups; weaker on joins	[4][8][14]
      Scalability model	Primarily vertical; partitioning can improve distribution	Horizontal via sharding, replication, consistent hashing	[12][13][15]
      Consistency model	Strong linearizable consistency (ACID)	Tunable or eventual; typically optimized toward availability	[16][17][18]
      CAP position	Typically optimized toward consistency + availability (CA)	Typically optimized toward availability + partition tolerance	[16][18]

      		(AP)
      Cloud performance	Adequate for structured transactional cloud workloads	Superior in distributed, cloud-native, high-throughput deployments	[11][15]
      Best use cases	Banking, ERP, healthcare, e- commerce, compliance	Social media, IoT, real-time analytics, cloud microservices	[1][3][5]

      Note: Positioning of CAP refers to criteria for architectural optimization instead of classifications

   2. Performance Synthesis (RQJ)

      Respond to RQ1: NoSQL databases outperform SQL databases in terms of writing speed and writing distributions, while SQL databases excel at structurally complex queries. This statement may be supported by the following benchmarking results [4][6][8][9][10][11] under the following reservations about benchmarking methods in certain benchmarks [14] as well as SQL database optimization practices [7]. The extent of advantage of NoSQL databases in terms of write performance is very much dependent on the database size.

   3. Scalability synthesis (RQ2)

      Respond to RQ2: NoSQL databases can take advantage in terms of architecture because of the possibility of achieving scalability by using methods such as sharding, hashing, and replication across various nodes [12][15], based on empirical evidence from tests conducted on distributed clusters which include limited-edge computing devices [15]. Conversely, SQL databases can match their distributed performance through careful partitioning [13], making the scalability difference between the two approaches smaller.

   4. Consistency synthesis (RQ3)

      Respond to RQ3: According to the CAP theorem, there exists a fundamental trade-off. Unlike SQL databases that have an apparent consistency model called ACID which, when applied to distributed NoSQL databases such as Cassandra, has clear drawbacks about efficiency and latency [17], NoSQL databases adopt the AP solution to ensure availability through eventual consistency rather than strong consistency [16]. When strong consistency is required, the SQL model should be used instead.

   5. Evaluation of benchmarking research methodology

      TABLE VI gives a methodological assessment of eight benchmarking papers based on five parameters, demonstrating some drawbacks of their methodology that make it difficult to generalize their findings. The best methodology used in benchmarking research can be observed in [14]. The missing replication set-up information [4][6][7][8][9] is one of the most common drawbacks of benchmarking research.

      TABLE VI. Benchmark Methodology Critique Across Included Studies

      Study	Benchmar k tool	Replication	Methodological limitation
      Li and Manohara n [4]	Custom scripts	Not disclosed	Uniform data size; limited generalizability to production workloads
      Gyorodi et al. [6]	Custom CRUD	Not configured	No cloud testing; schema not representative of dynamic workloads
      Gyorodi et al. [8]	Custom benchmark	Not configured	Write-skewed test design; limited to document workloads
      Katiyar et al. [9]	Custom inserts	Not disclosed	Insert-only; missing read/update/delet e workload coverage
      Truica et al. [10]	Custom CRUD + rep.	Async replication	Replication lag not measured across variable network conditions
      Singh et al. [11]	Cloud-native	Managed (cloud)	Cloud vendor not disclosed; results may not generalize
      Liyanage et al. [14]	Extended YCSB	Configurabl e	Most rigorous; explicitly addresses payload-size limitation of prior benchmarks
      Gorbenko et al. [17]	Custom + Cassandra API	Tunable (Cassandra)	Single-system; tunable consistency not compared across NoSQL systems

      1. DISCUSSION

         1. Practical Selection Framework (RQJ-RQ3 Synthesis)

            Following the analysis of the answers to research questions one to three, there is a need to consider database selection depending on the workload. Where applications require fast write speeds, cloud-based applications that operate in a distributed manner, where the schema is dynamic and involve working with massive volumes of unstructured data, NoSQL databases can be considered suitable options, including document and column-family types, such as MongoDB and Cassandra, respectively [5][8][11]. On the other hand, where there are transactions that must adhere to the ACID properties, relational and complex queries, and reporting considerations, then SQL databases offer a proper design option for applications ranging from banking, health care records to ERP applications [1][3][7].

         2. Convergence and hybrid architectures

            Another emerging trend is the increasing convergence of SQL and NoSQL functionalities. The support for JSONB by PostgreSQL, MySQL's document-oriented mode [7][8], and the ability of NoSQL systems to use ACID-compliant multi-document transactions show that the binary decision between the two has made room for hybrids. Unfortunately, none of the 18 examined papers provides empirical evidence regarding the benefits of polyglot persistence or hybrid SQL-NoSQL architectures, while such an approach becomes more common nowadays.

         3. lmplications for cloud-native and edge computing

            Both the cloud benchmarking paper [11] and the cluster evaluation study [15] imply that the scaling-out abilities of NoSQL databases will be of more importance in future because of the cloud-native and serverless nature of most future architectures. None of the examined papers have considered any serverless database configurations or evaluated performance on NVMe-SSD cloud storages, which could have been a different scenario compared to using VM-based environments.

         4. Research gaps and future directions

            Respond to RQ4: Six gaps in research have been quantified using evidence in the analysis. They are illustrated systematically in TABLE VII below, together with their severities and research directions.

            Research gap	Description	Studies addressi ng	Severi ty	Recommen ded direction
            Hybrid SQL/NoS QL architectur es	No study evaluates polyglot persistence or combined SQL/NoSQL deployment	0 of 18	Critic al	Empirical evaluation of hybrid persistence under cloud-native workloads
            Standardiz	Inconsistent	14 of 18	High	Adopt
            ed	tools, dataset	use		extended
            benchmar	sizes, and	custom		YCSB
            king	workload	benchma		with
            	profiles limit	rks		variable-
            	cross-study			payload,
            	comparison			production
            				–
            				representat
            				ive
            				workloads
            Energy	Only 1 study	1 of 18	High	Systematic
            efficiency	addressed			energy
            	energy			profiling
            	consumption			across
            	in distributed			SQL/NoS
            	deployments			QL DBs on
            				edge and
            				low-power
            				hardware
            Failure-	No study	0 of 18	High	Chaos
            mode	measures			engineerin
            consistenc	consistency			g
            y	during			experiment
            	network			s testing
            	partitions or			NoSQL
            	node failures			consistenc
            				y under
            				controlled
            				failure
            				scenarios
            Modern	Most studies	2 of 18	Medi	Updated
            cloud-	use on-		um	benchmark
            native	premises or			s on
            benchmar	basic VMs;			contempor
            king	NVMe/serve			ary cloud
            	rless absent			infrastructu
            				re and
            				serverless
            				configurati
            				ons

            TABLE VII. Evidence-Quantified Research Gaps and Future Directions

            Multi-	Only	1 of 18	Medi	Comparati
            system	Cassandra's		um	ve
            tunable	tunable			evaluation
            consistenc	consistency			across
            y	evaluated;			MongoDB,
            	other			DynamoD
            	systems			B,
            	unstudied			Cockroach
            				DB

            The most significant gaps in the literature include hybrid architectures and failure-mode consistency behavior, each of which is examined in 0 out of 18 articles, indicating an absolute gap. Both of these have clear organizational implications for those adopting hybrid architectures or NoSQL in unstable networked environments.

         5. Limitations

TABLE VIII highlights the main limitations along with their description and interpretation implications.

TABLE VIII. Study Limitations and Interpretive Implications

One of the most significant constraints is the scope limitation to 18 articles. Moreover, the exclusion of NewSQL databases, like CockroachDB and Google Spanner that offer the functionality of both SQL query processing capabilities and NoSQL scaling abilities horizontally, reduces the relevance of the study for those interested in assessing such novel database categories.

1. CONCLUSION

This study conducted a systematic literature review regarding SQL and NoSQL database systems based on their performance, scalability, and consistency perspectives by analyzing 18 research articles following the PRISMA methodology. No database system is better than others; the best choice is solely determined by the workload requirements.

NoSQL databases provide obvious write-throughput benefits and scalable architecture in distributed computing environments. Traditional SQL databases continue to provide strengths such as efficient structured queries, transactionally consistent ACID transactions, and complex relational operations; new partitioning strategies continue to close the gap on scalability issues. CAP theorem is validated in practice, and consistency is established as an explicit architectural factor.

There exist six research gaps that can be quantified through existing evidence. These gaps include hybrid SQL and NoSQL architectures and consistency behavior under different failure modes. As distributed computing continues to evolve with emerging applications such as AI inference and global-scale edge computing systems, new architecture paradigms including consistency, power-aware distributed storage, and hybrid persistence layers will emerge.

REFERENCES

1. A. A. Khan, M. Naqvi, M. Rizwan, and M. A. Jan, "SQL and NoSQL database software architecture performance analysis and assessments-A systematic literature review," IEEE Access, vol. 7, pp. 1-15, 2023, doi: 10.3390/bdcc7020097.

2. A. Floratou, N. Teletia, D. J. DeWitt, J. M. Patel, and D. Zhang, "Can the elephants handle the NoSQL onslaught?," Proc. VLDB Endowment, vol. 5, no. 12, pp. 1712-1723, Aug. 2012, doi: 10.14778/2367502.2367513.

3. S. Venkatraman and S. Fahd, "SQL versus NoSQL movement with big data analytics," Int. J. Inf. Technol. Comput. Sci., vol. 8, no. 12, pp. 59-66, 2016, doi: 10.5815/ijitcs.2016.12.07.

4. C. Li and S. Manoharan, "A performance comparison of SQL and NoSQL databases," in Proc. IEEE Pacific Rim Conf. Commun., Comput. Signal Process. (PACRIM), Victoria, BC, Canada, 2013, pp. 15-19, doi: 10.1109/PACRIM.2013.6625441.

5. F. Gessert, W. Wingerath, S. Friedrich, and N. Ritter, "NoSQL database systems: A survey and decision guidance," Comput. Sci. Res. Dev., vol. 32, nos. 3-4, pp. 353-365, 2017, doi: 10.1007/s00450-016-0314-1.

6. C. Gy6rodi, R. Gy6rodi, G. Pecherle, and A. Olah, "A comparative study: MongoDB vs. MySQL," in Proc. IEEE Int. Conf. Eng. Modern Electr. Syst. (EMES), 2015, pp. 1-6, doi: 10.1109/EMES.2015.7158433.

7. C. Gy6rodi, R. Gy6rodi, G. Pecherle, and A. Olah, "Performance impact of optimization methods on MySQL document-based and relational databases," Appl. Sci., vol. 11, no. 15, p. 6794, 2021, doi: 10.3390/app11156794.

8. C. Gy6rodi, R. Gy6rodi, G. Pecherle, and A. Olah, "A comparative study of MongoDB and document-based MySQL for big data application data management," Big Data Cogn. Comput., vol. 6, no. 2, p. 49, 2022, doi: 10.3390/bdcc6020049.

9. A. Katiyar, S. Katiyar, and A. Gupta, "Experiment based performance evaluation of MongoDB and MySQL," Zenodo, 2022, doi: 10.5281/zenodo.6799345.

10. C. O. Truica, F. Radulescu, and A. Boicea, "Performance evaluation for CRUD operations in asynchronously replicated document oriented database," arXiv preprint arXiv:1812.08220, 2018.

11. A. Singh, R. Kumar, and P. K. Gupta, "Cloud based evaluation of databases for stock market data," J. Cloud Comput., vol. 11, no. 1, pp. 1-18, 2022, doi: 10.1186/s13677-022-00334-8.

12. R. Cattell, "Scalable SQL and NoSQL data stores," ACM SIGMOD Rec., vol. 39, no. 4, pp. 12-27, May 2011, doi: 10.1145/1978915.1978919.

13. A. Pavlo, C. Curino, and S. Zdonik, "Skew-aware automatic database partitioning in shared-nothing, parallel OLTP systems," in Proc. ACM SIGMOD Int. Conf. Manage. Data, Scottsdale, AZ, USA, 2012, pp. 61-72, doi: 10.1145/2213836.2213845.

14. D. Liyanage, T. Perera, and M. Silva, "A benchmark for databases with varying value lengths," arXiv preprint arXiv:2501.xxxxx, 2025.

15. D. Pinheiro Gon;alves, J. M. S. Nogueira, and A. R. L. Zucchi, "An evaluation of relational and NoSQL distributed databases on a low-power cluster," Sensors, vol. 23, no. 5, p. 2610, 2023, doi: 10.3390/s23052610.

16. M. Diogo, B. Cabral, and J. Bernardino, "Consistency models of NoSQL databases," Future Internet, vol. 11, no. 2, p. 43, 2019, doi: 10.3390/fi11020043.

17. A. Gorbenko, V. Kharchenko, and O. Tarasyuk, "Interplaying Cassandra NoSQL consistency and performance: A benchmarking approach," in Lecture Notes in Computer Science, vol. 11928, 2020, pp. 229-242, doi: 10.1007/978-3-030-63270-0B18.

18. A. B. M. Moniruzzaman and S. A. Hossain, "NoSQL database: New era of databases for big data analytics," Int. J. Database Theory Appl., vol. 6, no. 4, pp. 1-14, 2013.

19. B. Kitchenham, "Procedures for performing systematic reviews," Keele Univ., Keele, U.K., Tech. Rep. TR/SE-0401, 2004.

20. E. A. Brewer, "Towards robust distributed systems," in Proc. ACM Symp. Principles Distrib. Comput. (PODC), Portland, OR, USA, 2000.