Relational Database Systems Research Articles

Robust and efficient memory management in Apache AsterixDB

SummaryTraditional relational database systems handle data by dividing their memory into sections such as a buffer cache and working memory, assigning a memory budget to each section to efficiently manage a limited amount of overall memory. They also assign memory budgets to memory‐intensive operators such as sorts and joins and control the allocation of memory to these operators; each memory‐intensive operator attempts to maximize its memory usage to reduce disk I/O cost. Implementing such memory‐intensive operators requires a careful design and application of appropriate algorithms that properly utilize memory. Today's Big Data management systems need the ability to handle large amounts of data similarly, as it is unrealistic to assume that truly big data will fit into memory. In this article, we share our memory management experiences in Apache AsterixDB, an open‐source Big Data management software platform that scales out horizontally on shared‐nothing commodity computing clusters. We describe the implementation of AsterixDB's memory‐intensive operators and their designs related to memory management. We also discuss memory management at the global (cluster) level. We conducted an experimental study using several synthetic and real datasets to explore the impact of this work. We believe that future Big Data management system builders can benefit from these experiences.

Monitoring Data Analysis Technology of Smart Grid Based on Cloud Computing

In the smart grid environment, the state data has the characteristics of wide area, panorama, mass and reliability. The traditional storage hardware uses disk arrays and the database management software uses the relational database system. Due to the poor system scalability, high cost and low reliability, it is difficult Adapt to requirements. Based on the above background, the purpose of this article is to study the research of cloud computing-based smart grid monitoring data analysis technology. Aiming at the problem that the power equipment condition monitoring data is getting larger and larger, and traditional storage methods cannot meet the condition monitoring big data storage problem, a cloud platform-based condition monitoring data storage system is designed. Using the distributed file system HDFS and Hbase databases to store status monitoring data, this paper designs a clustering algorithm based on the density cluster structure, DBCLustering, based on the shortcomings of traditional density clustering algorithms. This algorithm first builds the core reachability relationship of the data nodes The index structure CR-Tree is then used to extract a sorted linear table about the reachability relationship of the data. Finally, the clustering results are output according to the results of the linear table. In order to solve the problem of insufficient computing power of the stand-alone version of the algorithm, a clustering algorithm for condition monitoring data based on Spark-RDD-DBClustering algorithm is proposed. This algorithm realizes the parallel application of DBClustering algorithm under Spark platform, and improves the algorithm’s ability to process large-scale data.

Improving Archival Records and Service of Traditional Korean Performing Arts in a Semantic Web Environment

Abstract Purpose This research project aims to organize the archival information of traditional Korean performing arts in a semantic web environment. Key requirements, which the archival records manager should consider for publishing and distribution of gugak performing archival information in a semantic web environment, are presented in the perspective of linked data. Design/methodology/approach This study analyzes the metadata provided by the National Gugak Center’s Gugak Archive, the search and browse menus of Gugak Archive’s website and K-PAAN, the performing arts portal site. Findings The importance of consistency, continuity, and systematicity—crucial qualities in traditional record management practices—is undiminished in a semantic web environment. However, a semantic web environment also requires new tools such as web identifiers (URIs), data models (RDF), and link information (interlinking). Research limitations The scope of this study does not include practical implementation strategies for the archival records management system and website services. The suggestions also do not discuss issues related to copyright or policy coordination between related organizations. Practical implications: The findings of this study can assist records managers in converting a traditional performing arts information archive into a semantic web environment-based online archival service and system. This can also be useful for collaboration with record managers who are unfamiliar with relational or triple database system. Originality/value This study analyzed the metadata of the Gugak Archive and its online services to present practical requirements for managing and disseminating gugak performing arts information in a semantic web environment. In the application of the semantic web services’ principles and methods to an Gugak Archive, this study can contribute to the improvement of information organization and services in the field of Korean traditional music.

Supporting Uncertain Predicates in DBMS Using Approximate String Matching and Probabilistic Databases

Current relational database systems are deterministic in nature and lack the support for approximate matching. The result of approximate matching would be the tuples annotated with the percentage of similarity but the existing relational database system can not process these similarity scores further. In this paper, we propose a system to support approximate matching in the DBMS field. We introduce a `≈' (uncertain predicate operator) for approximate matching and devise a novel formula to calculate the similarity scores. Instead of returning an empty answer set in case of no match, our system gives ranked results thereby providing a glance at existing tuples closely matching with the queried literals. Two variants of the `≈' operator are also introduced for numeric data: `≈+' for higher-the-better and `≈-' for lower-the-better cases. Efficient approximate string matching methods are proposed for matching string-type data whereas numeric closeness is used for other types of data (date, time, and number). We also provide results of our system taken over several sample queries that illustrate the significance of our system. All experiments are performed using the MySQL database, whereas the IMDb movie database and European Football database are used as sample datasets.

$$\varvec{\textsc {Orpheus}}$$DB: bolt-on versioning for relational databases (extended version)

Data science teams often collaboratively analyze datasets, generating dataset versions at each stage of iterative exploration and analysis. There is a pressing need for a system that can support dataset versioning, enabling such teams to efficiently store, track, and query across dataset versions. We introduce OrpheusDB, a dataset version control system that “bolts on” versioning capabilities to a traditional relational database system, thereby gaining the analytics capabilities of the database “for free.” We develop and evaluate multiple data models for representing versioned data, as well as a lightweight partitioning scheme, LyreSplit, to further optimize the models for reduced query latencies. With LyreSplit, OrpheusDB is on average $$10^3\times $$ faster in finding effective (and better) partitionings than competing approaches, while also reducing the latency of version retrieval by up to $$20\times $$ relative to schemes without partitioning. LyreSplit can be applied in an online fashion as new versions are added, alongside an intelligent migration scheme that reduces migration time by $$10\times $$ on average.

Regulatory Architecture of Data Processing for Connected and Automated Driving in Europe

The beginning of the 2020s ought to reflect a steady conclusion of the vast majority of the European Union’s projects with regards to the new era of connectivity and mobility within the European Union dimension. We expect Intelligent Connected Vehicles (ICVs) to step into free circulation within the internal market. Since the operation of the ICVs depends on the number of data processing operations, data processing operations should be precisely determined and framed beforehand. ICVs data operations consist of extraordinarily large volumes and velocity of a data flow which previously existed in traditional relational database systems and could not have been processed within the desired timeframe. Even though the currently adopted database systems are ready to face the new level of data processing, a huge data stream is also faced with complex obstacles and new risks which have never been experienced beforehand.While seeking to ensure safe and secure introduction of a new level of data processing for connectivity and automation at the European Union market, the author precisely examines all potential risks and possibilities of integration into a uniform legal regulation to ensure secured ICVs data processing at all levels. The regulatory framework should document adequate security requirements and defences against ICVs attacks e.g. interference and remote-control interception.

Improved Presentation and Facade Layer Operations for Software Engineering Projects

Nowadays, one of the most challenging situations for software developers is the presence of a mismatch between relational database systems and programming codes. In the literature, this problem is defined as impedance mismatch. This study is to develop a framework built on innovations based on the existing Object Relational Mapping technique to solve these problems. In the study, users can perform operations for three different database systems such as MsSQL, MySql and Oracle. In addition, these operations can be done within the framework of C# and Java programming languages. In this framework, while the developers can define database tables in the interface automatically, they can create relations between tables by defining a foreign key. When the system performs these operations, it creates tables, views, and stored procedures automatically. In addition, entity classes in C# and Java for tables and views, and operation classes for stored procedures are created automatically. The summary of the transactions can be taken as pdf file by the framework. In addition, the project can automatically create Windows Communication Foundation classes to facilitate the handling of database elements created and the interfacing operations, as well. This framework, which supports distributed systems, can be downloaded at this link.

HR-SQL: Extending SQL with hypothetical reasoning and improved recursion for current database systems

In this work we present a formalization, backed with a contrasted implementation, of a relational database language (called HR-SQL) that extends SQL in two aspects. On the one hand, including non-linear and mutual recursion. On the other hand, including hypothetical relations and queries. Regarding expressiveness, HR-SQL allows a novel form of hypothetical reasoning, completely integrated with recursive definitions. In addition, aggregate functions are added. Regarding formalization, the extended language is founded on a stratified fixpoint semantics based on logic programming techniques. We include results of the existence of such fixpoints. An algorithm that transforms a database containing hypothetical definitions into an equivalent one without hypothesis and its correctness proof are presented. Regarding the implementation, we introduce here a system that incorporates the aforementioned benefits and enhances former implementations in several areas. The current HR-SQL system is targeted to several state-of-the-art relational database systems, and could be used with any SQL-based system with minor modifications in the implementation.

Towards Declarative and Data-Centric Virtual Machine Image Management in IaaS Clouds

Virtual machine image (VMI) management has become one of the key infrastructure components in Infrastructure as a Service (IaaS) cloud systems. Any “good” VMI management system should support flexible and efficient VMI services to cloud users, and offer scalable, easy-to-maintain and efficient VMI management for cloud providers. While there have been a number of systems and optimizations for VMI management, this paper investigates a declarative and data-centric approach to VMI management for both cloud users and providers. Specifically, by viewing VMI management as a data-intensive application, we propose Hemera , a novel VMI management system prototype based on relational database systems. Hemera adopts a data-centric approach to VMI management system design, where a VMI is modeled as structured data. With the data-centric approach, the key operations of VMI management can be recast naturally as programs based on SQL language. Moreover, Hemera embraces a series of automatic optimization opportunities with their root from databases. We have developed a system prototype based on MySQL. Our experimental results show the efficiency and feasibility of our declarative and data-centric approach to VMI Management.

Query Optimization in Crowd-Sourcing Using Multi-Objective Ant Lion Optimizer

Nowadays, query optimization is a biggest concern for crowd-sourcing systems, which are developed for relieving the user burden of dealing with the crowd. Initially, a user needs to submit a structured query language (SQL) based query and the system takes the responsibility of query compiling, generating an execution plan, and evaluating the crowd-sourcing market place. The input queries have several alternative execution plans and the difference in crowd-sourcing cost between the worst and best plans. In relational database systems, query optimization is essential for crowd-sourcing systems, which provides declarative query interfaces. Here, a multi-objective query optimization approach using an ant-lion optimizer was employed for declarative crowd-sourcing systems. It generates a query plan for developing a better balance between the latency and cost. The experimental outcome of the proposed methodology was validated on UCI automobile and Amazon Mechanical Turk (AMT) datasets. The proposed methodology saves 30%-40% of cost in crowd-sourcing query optimization compared to the existing methods.

IoT Analysis and Processing System using Sensed Data for Laboratory Safety Management

This study proposes and implements a real-time message analysis processing system using sensed data for safety management of a laboratory composed of Internet of Things (IoT) sensor devices. Existing IoT systems have difficulty storing data in a relational database system due to the different data formats between sensor devices, and they also have preprocessing overhead due to nonstandard communications protocols. Furthermore, they are centralized systems with unstable service continuity, and it is difficult to add new functions or extend the system because of the complex connections among sensor devices and services. Studies on real-time analysis and processing of large volumes of sensor data are also insufficient. To address these problems, the proposed system uses IoT sensor devices, based on standard communications protocols, and message queue brokers based on standard messaging protocols. The message queue brokers can facilitate adding sensor devices or services by delivering the sensor data from the gateway and central server to the services. Services can continue, even in the event of network trouble, because the services that had been provided from the central server are now provided through smart gateways. The central server can provide risk analysis and prevention services for laboratory safety management through Apache Spark, a real-time big data analysis and processing platform. Through the experiments in this study, the proposed real-time message analysis processing system for laboratory safety management was implemented. The message queue brokers between the standard Message Queuing Telemetry Transport (MQTT) protocol and the smart gateway receiving the data were built into sensor devices equipped with network and sensor modules. Exchange routing keys are also specified by topic and service for each sensor device in line with the laboratory safety management environment. Data transmission processing is handled by a federated connection of the gateway brokers and the central server brokers.

RADI-19. EVALUATION OF BRAIN METASTASIS LOCAL CONTROL POST RADIOSURGERY VIA MACHINE LEARNING AND RADIOMICS

Stereotactic radiosurgery can be used to treat multiple, surgically inaccessible, metastatic brain lesions in a single, minimally invasive outpatient procedure. For brain metastasis, stereotactic radiosurgery can provide excellent local control depending on the robustness of the treatment plan. Previous studies have been performed correlating key radiation planning factors to higher local control probability such as tumor size and maximum dose. However, a separate non-inferiority study demonstrated that higher prescription isodose lines (in excess of 70% or higher) did not correlate to local failure. The previous works were limited to shallow feature levels regarding only the dicom plan information and lacked a predictive model. In order to address these conflicting conclusions and to support clinical decision making, we propose a radiosurgery informatics pipeline to support testing these hypotheses with observational data. First, a multidisciplinary team generated a mind-map of relevant information to inform database design. Portions of this mind-map were implemented in a relational database system (PorstgreSQL), and populated with information from 1024 patients treated for brain metastasis via stereotactic radiosurgery. Clinical information were derived from curated databases and the array of intervention variables were mined from the DICOM RT plans, structure sets, images and dose via MATLAB scripts. These factors include, but are not limited to, radiation dosimetry, prior whole brain radiation, radiomic imaging features, prior radiosurgery status, and physician determined local control status. From this pipeline, we plan to use a multi-level feature-based supervised machine learning approach that will be created via boosting to predict local control in patients using local failure timing, or lack thereof, provided by physician. To control for local failure observer bias, an unsupervised machine learning model via random trees will be created to predict clusters of patient parameters with similar local control rates.

CASE STUDY OF TRADITIONAL RDBMS AND NOSQL DATABASE SYSTEM

On the basis of reliability, rational models are useful but not in terms of systems which involve huge amount of data; in such cases, non-relational models are much more useful. To store large chunks of data, NoSQL databases are used. NoSQL databases are scalable and wide ranged because they are non-relationally distributed.
 In relational databases, it was not possible to manage data which involved very large number of Big Data applications hence the concept of NoSQL database was introduced.
 There are a lot of advantages of NoSQL which not only involve its own features but also some features of relational database management system. The severe benefit of NoSQL database is that it is an open source system which helps to adapt many numbers of features for newly generated applications.
 This paper is focused on understanding the concepts of non-relational database system architecture with relational database system architecture and figure out the advantages and disadvantages of both simultaneously.

Beyond Data Capture: A Single Institutional Experience with Establishing a Pelvic Floor Surgery Web Based Database.

The American Urological Association has highlighted the need for complete and long-term data collection in female pelvic medicine and reconstructive surgery. We describe our institution's experience with transitioning our pelvic floor surgery database to a secure web based platform. In this article we share this experience and its challenges to help facilitate the process for other institutions. In 1999 our institution implemented electronic data capture of pelvic floor cases using a stand-alone spreadsheet. By 2001 it was apparent that this modality could no longer sustain growth while maintaining data integrity. We converted to a relational database system but limitations involving multiuser data entry and efficient querying became apparent. After consultation with outcomes research experts across the country we evaluated database alternatives for capacity, ease of use and security. REDCap® (Research Electronic Data Capture), a low cost, server based database software, provided all 3. In 2011 the transition of our electronic outcomes database to REDCap commenced. The transition required approximately 100 person-hours for data dictionary creation and 200 person-hours for the transfer of existing data into REDCap. A database administrator was hired to maintain data integrity, automate the process of generating questionnaires for long-term followup and facilitate data queries. We describe the transition of our institution's outcomes database to REDCap providing the means for secure and streamlined data capture and quality assurance. This platform also facilitates effective multi-institutional collaboration with the goals of developing quality data to advance research and improve care for our patients.

Temporal data exchange

Data exchange is the problem of transforming data that is structured under a source schema into data structured under another schema, called the target schema, so that both the source and target data satisfy the relationship between the schemas. Many applications such as planning, scheduling, medical and fraud detection systems, require data exchange in the context of temporal data. Even though the formal framework of data exchange for relational database systems is well-established, it does not immediately carry over to the settings of temporal data, which necessitates reasoning over unbounded periods of time.In this work, we study data exchange for temporal data. We first motivate the need for two views of temporal data: the concrete view, which depicts how temporal data is compactly represented and on which the implementations are based, and the abstract view, which defines the semantics of temporal data as a sequence of snapshots. We first extend the chase procedure for the abstract view to have a conceptual basis for the data exchange for temporal databases. Considering non-temporal source-to-target tuple generating dependencies and equality generating dependencies, the chase algorithm can be applied on each snapshot independently. Then we define a chase procedure (called c-chase) on concrete instances and show the result of c-chase on a concrete instance is semantically aligned with the result of chase on the corresponding abstract instance. In order to interpret intervals as constants while checking if a dependency or a query is satisfied by a concrete database, we will normalize the instance with respect to the dependency or the query. To obtain the semantic alignment, the nulls (which are introduced by data exchange and model incompleteness) in the concrete view are annotated with temporal information. Furthermore, we show that the result of the concrete chase provides a foundation for query answering. We define naïve evaluation on the result of the c-chase and show it produces certain answers.

Scalable Linear Algebra on a Relational Database System

As data analytics has become an important application for modern data management systems, a new category of data management system has appeared recently: the scalable linear algebra system. In this paper, we argue that a parallel or distributed database system is actually an excellent platform upon which to build such functionality. Most relational systems already have support for cost-based optimization-which is vital to scaling linear algebra computations-and it is well-known how to make relational systems scale. We show that by making just a few changes to a parallel/distributed relational database system, such a system can be a competitive platform for scalable linear algebra. Taken together, our results should at least raise the possibility that brand new systems designed from the ground up to support scalable linear algebra are not absolutely necessary, and that such systems could instead be built on top of existing relational technology. Our results also suggest that if scalable linear algebra is to be added to a modern dataflow platform such as Spark, they should be added on top of the system's more structured (relational) data abstractions, rather than being constructed directly on top of the system's raw dataflow operators.

PrivateSQL

Differential privacy is considered a de facto standard for private data analysis. However, the definition and much of the supporting literature applies to flat tables. While there exist variants of the definition and specialized algorithms for specific types of relational data (e.g. graphs), there isn't a general privacy definition for multi-relational schemas with constraints, and no system that permits accurate differentially private answering of SQL queries while imposing a fixed privacy budget across all queries posed by the analyst. This work presents PrivateSQL, a first-of-its-kind end-to-end differentially private relational database system. PrivateSQL allows an analyst to query data stored in a standard database management system using a rich class of SQL counting queries. PrivateSQL adopts a novel generalization of differential privacy to multi-relational data that takes into account constraints in the schema like foreign keys, and allows the data owner to flexibly specify entities in the schema that need privacy. PrivateSQL ensures a fixed privacy loss across all the queries posed by the analyst by answering queries on private synopses generated from several views over the base relation that are tuned to have low error on a representative query workload. We experimentally evaluate PrivateSQL on a real-world dataset and a workload of more than 3, 600 queries. We show that for 50% of the queries PrivateSQL offers at least 1, 000x better error rates than solutions adapted from prior work.

NIA GENETICS OF ALZHEIMER’S DISEASE DATA STORAGE SITE (NIAGADS): ALZHEIMER’S GENOMICS DATABASE – 2019 UPDATE

The Alzheimer's Genomics Database (https://ww.niagads.org/genomics) provides public access to GWAS summary statistics datasets deposited at NIAGADS, a national genetics data repository for AD. This resource makes available 54 comprehensive summary statistics datasets from 14 GWAS studies for real-time mining with a flexible search interface and exploration via detailed reports and a genome browser. The GenomicsDB is powered by a big-data optimized relational database system. It uses OBO Foundry ontologies to consistently annotate study designs and phenotypes, facilitating data harmonization and efficient real-time data analysis. AD-risk associated variants from GWAS datasets are mapped to up-to-date annotations from dbSNP, ExAC, and similar resources. Variants are also mapped against the GRCh37.p13 assembly and linked to proximal and colocated genes and regulatory regions from functional genomics data repositories (e.g., ENCODE, Roadmap, FANTOM5). The Alzheimer's GenomicsDB efficiently compiles detailed variant and gene reports to assist researchers in interpreting their potential functional or regulatory role in the context of AD neurodegeneration. Updated variant reports highlight those discovered and QC'd by the Alzheimer's Disease Sequencing Project (ADSP) and include ranked predicted variant consequences generated with the ADSP's Annotation Pipeline (Butkiewicz et al. 2018). Other new features include linkage disequilibrium associations, which can be explored via interactive LocusZoom plots. Gene reports have also been updated and now include meta-analysis results from aggregate association tests performed by the ADSP (Bis et al. 2018), allowing us to flag those with genetic-evidence for AD. The system powering the GenomicsDB allows researchers to not only mine, but also analyze and annotate search results or uploaded data. In addition to functional and pathway enrichment analyses of gene lists, researchers can now run INFERNO (Amlie-Wolf et al. 2018) to compare variant lists against transcription factor binding from ENCODE, histone modification sites from Roadmap, FANTOM5 expressed enhancer regions, and GTex eQTL sites to identify potential regulatory elements, tissue contexts, and target genes. Providing access to genetic evidence for AD, the publicly available and user-friendly Alzheimer's GenomicsDB is a rich resource for the AD research community.

From Data to Knowledge: A semantic knowledge graph application for curating specimen data

The landscape of currently existing repositories of specimen data consists of isolated islands, with each applying its own underlying data model. Using standardized protocols such as DarwinCore or ABCD, specimen data and metadata are exchanged and published on web portals such as GBIF. However, data models differ across repositories. This can lead to problems when comparing and integrating content from different systems. for example, in one system there is a field with the label 'determination', in another there is a field with the label 'taxonomic identification'. Both might refer to the same concepts of organism identification process (e.g., 'obi:organism identification assay'; http://purl.obolibrary.org/obo/OBI_0001624), but the intuitive meaning of the content is not clear and the understanding of the providers of the information might differ from that of the users. Without additional information, data integration across isolated repositories is thus difficult and error-prone. As a consequence, interoperability and retrievability of data across isolated repositories is difficult. Linked Open Data (LOD) promises an improvement. URIs can be used for concepts that are ideally created and accepted by a community and that provide machine-readable meanings. LOD thereby supports transfer of data into information and then into knowledge, thus making the data FAIR (Findable, Accessible, Interoperable, Reusable; Wilkinson et al. 2016). Annotating specimen associated data with LOD, therefore, seems to be a promising approach to guarantee interoperability across different repositories. However, all currently used specimen collection management systems are based on relational database systems, which lack semantic transparency and thus do not provide easily accessible, machine-readable meanings for the terms used in their data models. As a consequence, transferring their data contents into an LOD framework may lead to loss or misinterpretation of information. This discrepancy between LOD and relational databases results from the lack of semantic transparency and machine-readability of data in relational databases. Storing specimen collection data as semantic Knowledge Graphs provides semantic transparency and machine-readability of data. Semantic Knowledge Graphs are graphs that are based on the syntax of ‘Subject – Property – Object’ of the Resource Description Framework (RDF). The ‘Subject’ and ‘Property’ position is taken by URIs and the ‘Object’ position can be taken either by a URI or by a label or value. Since a given URI can take the ‘Subject’ position in one RDF statement and the ‘Object’ position in another RDF statement, several RDF statements can be connected to form a directed labeled graph, i.e. a semantic graph. Semantic Knowledge Graphs are graphs in which each described specimen and its parts and properties possess their own URI and thus can be individually referenced. These URIs are used to describe the respective specimen and its properties using the RDF syntax. Additional RDF statements specify the ontology class that each part and property instantiates. The reference to the URIs of the instantiated ontology classes guarantees the Findability, Interoperability, and Reusability of information contained in semantic Knowledge Graphs. Specimen collection data contained in semantic Knowledge Graphs can be made Accessible in a human-readable form through an interface and in a machine-readable form through a SPARQL endpoint (https://en.wikipedia.org/wiki/SPARQL). As a consequence, semantic Knowledge Graphs comply with the FAIR guiding principles. By using URIs for the semantic Knowledge Graph of each specimen in the collection, it is also available as LOD. With semantic Morph·D·Base, we have implemented a prototype to this approach that is based on Semantic Programming. We present the prototype and discuss different aspects of how specimen collection data are handled. By using community created terminologies and standardized methods for the contents created (e.g. species identification) as well as URIs for each expression, we make the data and metadata semantically transparent and communicable. The source code for Semantic Programming and for semantic Morph·D·Base is available from https://github.com/SemanticProgramming. The prototype of semantic Morph·D·Base can be accessed here: https://proto.morphdbase.de.

SolarDB

Efficient transaction processing over large databases is a key requirement for many mission-critical applications. Although modern databases have achieved good performance through horizontal partitioning, their performance deteriorates when cross-partition distributed transactions have to be executed. This article presents SolarDB, a distributed relational database system that has been successfully tested at a large commercial bank. The key features of SolarDB include (1) a shared-everything architecture based on a two-layer log-structured merge-tree; (2) a new concurrency control algorithm that works with the log-structured storage, which ensures efficient and non-blocking transaction processing even when the storage layer is compacting data among nodes in the background; and (3) find-grained data access to effectively minimize and balance network communication within the cluster. According to our empirical evaluations on TPC-C, Smallbank, and a real-world workload, SolarDB outperforms the existing shared-nothing systems by up to 50x when there are close to or more than 5% distributed transactions.