Acid Properties In Distributed Database Transactions

Published Date: 02 Nov 2017

Abstract

ACID (Atomicity, Consistency, Isolation and Durability) properties are widely used to model the design for distributed database transactions. There are many models presented in literature review to support distributed database systems, but they failed in relaxing all the four properties to give reliable user service. The purpose of each research is to develop a model which aims at improving the throughput, while maintaining the ACID properties on user needs. The models are not generalized because they are not applicable to all domains. After the analysis and comparison of all models based on evaluation criteria, it has been analyzed that most of the techniques lack tool support, testability and any improvement in response time. Furthermore, it explains the advantages as well as disadvantages of these techniques and draws a conclusion.

1. Introduction

A central Database Management System (DBMS) controls the distributed database that is dispersed over network of interconnected computers. There are two processes performed on distributed database, replication and duplication. Upon these processes, it must be assured that the data in database should be very consistent and stable. There should be no problem when some transaction is executed. How does the database consistent and stable? The solution of this problem is the ACID (Atomicity, Consistency, Isolation, and Durability) properties of transaction management which guarantees the reliability of transaction over distributed database. ACID are developed for traditional, business oriented applications. It is really important for the database to have ACID properties because it ensures all data in scientific research and business field to be correct and valid status, without, these properties database transaction will be mess. This research paper will review the other research studies to explain and analyze the usefulness and effectiveness of ACID properties for transaction management.

2. Distributed Database System Techniques

2.1) Concurrency Issues of Distributed Advance Transaction Process (Sheetlani Jitendra et al, 2012).

Sheetlani Jitendra et al [1] reviewed the basic concepts of a distributed database system. Further, the idea of advance nested transaction is presented which provides much flexibility and high performance when used for complex application such as web-services transactions and heterogeneous distributed systems. The drawback of this research is that, nested transaction are not implemented and testified to check the effects of response time of transactions.

2.2) A Distributed Transaction Model for a Multi-database Management System (Omar Baakeel et al, 2012).

B. Omar et al [2] has discussed the distributed transaction issues that are present in multi-database management systems (DBMSs) and how the distributed transaction in database technology differs from other distributed processing systems. Furthermore, transactions problems such as network problem, site failure and time failure are also discussed. This research paper proposed a new distributed transaction model for performing and managing transaction details more efficiently.

2.3) Countermeasures against Consistency Anomalies in Distributed Integrated Databases with Relaxed ACID Properties (Lars Frank, 2011).

L. Frank [3] used ACID properties of a Database Management System (DBMS) in distributed and mobile databases and made the availability of data with optimization. The design of Consistency anomalies is used for the inconsistencies caused by the missing consistency property in distributed integrated databases with relaxed ACID properties. Countermeasures against the consistency anomalies are also discussed to make it possible to reduce or avoid the consequences of the missing distributed consistency property. The countermeasures which are described have not been discussed in the scientific literature before but they may have been used in practice as the problems of the missing distributed integrity control may have been dealt with in practical implementations of extended transaction models.

2.4) Relaxation of ACID properties in AuTrA, The adaptive user-defined transaction relaxing approach (Rose Tshoganetso Khachana et al, 2011).

Rose Tshoganetso Khachana et al [4] proposed a new model Adaptive User-defined Transaction Relaxing Approach (AuTrA) is introduced for supporting transactions in a web services environment. The model offers business benefits by increasing throughput of transactions. This is achieved by supporting the relaxation of various combinations of ACID properties according to the user's requirements and the business needs. The model aims at improving transaction throughput at the same time as maintaining a reliable user service. A distinctive feature is that any or all of the ACID properties may be maintained or relaxed depending on user need. The model is tested through various experiments and results which show that the proposed model AuTrA is capable of increasing transaction throughput that makes AuTrA a truly valuable technology. The strength of this tool is that, it widely supports user and business needs for web-based applications with a disadvantage that it is domain specific.

2.5) Transaction Management for Distributed Database using Petri Nets (Bidyut Biman Sarkar et al, 2010).

Bidyut Biman Sarkar et al [5] focused that distributed database environment must ensure that the sequence of updates is safe and reliable when committed on the stable storages at different locations. For this purpose, Two Phase Commit (2PC) protocol is proposed for distributed transactions with the help of a timed Petri net model to analyze the ACID property for consistent commitment of distributed transactions. The proposed protocol lacks implementation to executable model.

2.6) Dynamic Management of Transactions in Distributed Real-Time Processing System (Y. Jayanta Singh et al, 2010)

Y. Jayanta et al [6] has discussed the transactions problems in real time distributed processing system. In this paper a series of simulation study have been performed to evaluate the performance under different transaction management situation such as different workloads, distribution methods, impact of dynamic slack factors to throughput. The data accesses is arranged in order to meet their deadlines and to minimize the number of transactions that missed deadlines. The advantage of this paper is that, the system gives a significant improvement in performance.

2.7) On Performance Evaluation and Design of Atomic Commit Protocols for Mobile Transactions (Nadia Nouali-Taboudjemat et al, 2010).

N. N. Taboudjemat et al [7] have studied the impact of the various and fluctuant parameters of wireless and mobile systems on a set of ACPs for mobile environment. It highlights performance indices which give orientations to the design of an adaptable approach that supports different atomicity notions satisfying a wide range of applications and environment requirements [7]. This research paper focused on the atomic commit protocols for mobile transactions designed for infrastructure based mobile networks. Furthermore, evaluating a set of such protocols are also discussed and highlighted their performance indices.

2.8) A Case Study for Distributed Business Transaction Coordination across Heterogeneous Platforms (Bin Mu et al, 2010).

B. Mu et al [8] have been presented the technology of Web Service Atomic Transaction (WS-AT) for performance, workload and the architecture of distributed system. With WS-AT support we can completely ignore the differences of the heterogeneous platforms to reuse the legacy IT assets in modern systems and make existing EIS’s architecture more extendable, flexible, reliable and interoperable to enhance the system’s quality of service in an open SOA environment. The disadvantage of this technology is that it is not suitable for processing the transactions with long time duration and will in some cases increase the system’s workload.

2.9) Queue Sensing Distributed Real-time Commit Protocol: A New Dimension for Distributed Database System (Sylvia et al, 2009).

Improvement in the performance of transactions in real time distributed system has been proposed by Sylvia et al [9]. The protocol used in the research paper is Queue Sensing Distributed Real-time Commit Protocol (QSDRCP) which ensures the maximum number of transactions is processed before the deadline expires and the ACID property of the transactions executed is preserved. Comparison of QSDRCP with other protocols which are working to improve the performance of transactions is also discussed.

Table 1: Evaluation Criteria for ACID properties in Distributed Database

Evaluation Parameters

Meaning

Possible Values

Response Time/Performance

Throughput/time required in completing one transaction

Increased, Decreased

Consistency

To check the integrity in Distributed Database

Yes, No

Reliability

Run efficiently

Yes, No

Tool

Proposed models supporting distributed transactions

Yes ,No

Testability

Proposed model is tested or not

Yes, No

Techniques

Techniques used for the distributed transaction management

ACID Properties, Two phase Commit

Interface

Environment

Environment Type

Case study

Support of examples

Yes, No

Validity

Tested through experiments

Yes, No

2.10) Distributed Real Time Database Systems: Background and Literature Review (Udai Shanker et al, 2008).

Udai Shanker et al [10] described basic concepts and definitions of the real time database system and reviewed the work carried out in the areas of priority assignment policies for the scheduling of transactions, commit processing and memory optimization in non-replicated/replicated environment. Most of these protocols try to improve the performance by allowing a transaction to use a data item locked by some other transaction. There is disadvantage of complete lack of protocols which can be well suited to efficient use of memory by creating lesser number of temporary objects. Some of the highlighting factors about the work reported in this study are analysis of the static two phase locking mechanism, priority assignment policy, and commit protocols suitable for handling huge volume of data and large number of transactions.

2.11) A Survey on the History of Transaction Management: from Flat to Grid Transactions (TingWang et al, 2008).

TingWang et al [11] presents a historic overview of numerous transaction models ranging from the classical flat transactions, via advanced and work flow transactions to the Web Services and Grid transaction models organized in several ages of transaction management. The evolution and relationship between the transactional models is depicted graphically. The research also focuses that the future trend of the transaction is correspondingly following a direction to address the need of more functionality and better performance. The fundamental idea is to provide a reliable approach in designing new model that goal remains same rather than complex requirements.

2.12) SWIFT-A New Real Time Commit Protocol (Udai Shanker et al, 2006).

Udai Shanker et al [12] designed an efficient commit protocol SWIFT for real time distributed database systems because the existing protocols improves their performance by allowing a committing cohort and an executing cohort where committing cohort provides it data to an executing cohort. As a result, it reduces data inaccessibility. SWIFT is ideal for fast and timeliness commit protocol because it reduces time needed for processing and is free of cascaded aborts. Simulations show that SWIFT improves response time as compared to previous protocols which results in increasing performance.

2.13) Electronic Commerce Using Distributed ERP-Systems with Approximated ACID Properties (Lars Frank, 2004).

L. Frank [13] gave an overview of how it is possible to incorporate different heterogeneous internet systems without losing the consistency of the data involved. In this research paper relaxed ACID properties are used. The main purpose of using these properties is to give the users a view of the data that may be inconsistent across different locations but in practice functions as if the traditional ACID properties were implemented. That means, the users can trust the data they use.

2.14) Transaction Management in Distributed Database Systems: the Case of Oracle’s Two-Phase Commit (Ghazi Alkhatib et al, 2002).

Ghazi Alkhatib et al [14] explains transaction management in Distributed Database Systems (DDBS) and how Oracle implements tow phase commit to maintain the consistency and reliability of the database. In one section, fundamental of transaction managements are introduced whereas the other section explains an example of oracle database management system is given to demonstrate the step involved in executing the two-phase commit which assists when installing Oracle DBMS, or encourage organizations to migrate from centralized to distributed DBMS. The research ignored the implementation of transaction management platform which is its limitations whereas the comprehensive description of the Oracle Database Management System is its strength.

2.15) Dictatorial Transaction Processing: Atomic Commitment without Veto Right (Maha Abdallah et al, 2002).

Maha Abdallah et al [15] gave, for the first time, a precise abstract specification of the Dictatorial Atomic Commitment (DAC) problem, resulting from removing veto rights from the traditional Atomic Commitment (AC) problem. This study took advantage of existing protocols that solve the DAC problem, and proposed a new protocol that broadens the applicability of dictatorial transaction processing in order to meet the requirements of today’s distributed environments. The strength of this protocol is performance tradeoffs, and described the implementation of Dictatorial atomic commit protocol in the context of current transactional standards, initially designed with 2PC in mind.

2.16) Architecture for Integration of Distributed Enterprise Resource Planning Systems and E-Commerce Systems (Lars Frank, 2004).

Enterprise Resource Planning (ERP) software’s with high performance and reliability have been designed by L. Frank [16]. The global atomicity property and durability property is implemented, consistency and isolation properties are also managed. The method used in this research paper is extended replication evaluation method. The Advantage of this method is that if the analyzed ERP system had used a flexible transaction model, countermeasures against isolation anomalies, and short duration locks, then it would have been relatively easy to combine such a system in the distributed architecture and reuse already existing application code.

2.17) A Transaction Model to Improve Data Availability in Mobile Computing (Sanjay Kumar Madria et al, 2001).

S.K. Madria et al [17] have proposed a technique of prewritten values for mobile transactions. Prewritten values help in increasing accessibility as the transactions can be executed during disconnections both at Mobile Host (MH) and Mobile Service Station (MSS) without blocking. The proposed technique allows a transaction’s execution to shift from the MH to MSS for database updates. Therefore it reduces the computing expenses. In this research paper algorithms for the proposed transaction processing model and the locking protocols are also discussed.

2.18) Mechanical Verification of Transaction Processing Systems (Dmitri Chkliaev et al, 2000).

D. Chkliaev et al [18] have presented formal specifications and mechanical verifications of transaction processing system aimed at distributed database system and the relationship of ACID properties among themselves. The protocols which are used for formal verification is two-phase locking, two-phase commit and undo/redo recovery. The locking and undo/redo mechanism at distributed sites is characterized by state machines, whereas the interaction between sites according to the two-phase commit protocol is precise by assertions.

2.19) Application specific Transaction Management in Multi-database Systems (Peter Muth, 1997).

Peter Muth [19] presented a transaction model for multi database known as heterogeneous 3-level transactions, including a formal definition of serializability and recovery. The heterogeneous 3-level transaction model assumes component database systems to support ACID transaction properties. This research also describes a prototype implementation of heterogeneous 3-level transactions in the object-oriented database system VODAK. It is verified that heterogeneous 3-level transactions guarantee all consistency constraints that are ensured in serial executions. But the proposed model lacks the parameter of response time.

2.20) Overview of Multi-Database Transaction Management (Yuri Breitbart et al, 1992).

Yuri Breitbart et al [20] provided a brief review of the most current work in the area of multi-database transaction management. The paper work focuses a major problem that exists in the heterogeneous computing environments, distributed over various DBMSs is that traditional techniques for ensuring transaction atomicity and consistency which are easily implemented in homogeneous distributed database systems are not appropriate for an heterogeneous environment. So, it is very important to develop methods that do not require major modifications to existing DBMS software but are able to support users data in a consistent and reliable manner.

3. Analysis

Table 2 shows the results the analysis of evaluation parameters defined in evaluation criteria table 1. Twenty techniques are surveyed and it used nine parameters for their evaluation. Response time is an important factor while implementing ACID properties. An analysis of the table 2 reveals that A. James et al [4], Sylvia et al [9], U. Shankar et al [12] has improved response time of transactions but others did not pay much attention to it. All of the techniques satisfy the parameter consistency which is important property of database transaction management. All most all of the techniques provide reliability of database to the user except B. Omar et al [2], B. Mu et al [8] and M. Abdallah et al [15]. Some of the techniques proposed models to support their experiments are as follows: B. Omar et al [2], L. Frank [3], A. James et al [4], B.B. Sarkar et al [5], G. Alkhatib et al [14], S. K. Madria et al [17], P. Muth et al [19], Y. Breitbart et al [20].

Table 2: Analysis of existing techniques for ACID Properties in Distributed System

S#

Techniques

Response Time

Consistency

Reliability

Tool

Testability

Techniques

Interface

Case Study

Validity

1

S. Jitendra et al.,2012

No

Yes

Yes

No

No

ACID Properties

Distributed Environment

No

No

2

B. Omar e tal., 2012

No

Yes

No

Exceptional distributed transaction model

No

ACID properties

Multi-database management system

No

Yes

3

L. Frank, 2011

No

Yes

Yes

Countermeasure Transaction model

No

Relaxed ACID Properties

Distributed Integrated Databases

E-commerce system

Yes

4

A. James et al., 2011

Decreased

Yes

Yes

AuTra

Yes

AICD Properties

Web environment

Travel Scenario

Yes

5

B.B. Sarkar

et al.,2010

No

Yes

Yes

Timed Petri Nets

No

Two Phase Commit, ACID Properties

Distributed Virtual Data warehouse

No

Yes

6

Y. Jayanta et al., 2010

No

Yes

Yes

No

Yes

Real time distributed processing system

No

Yes

7

N. N. Taboudjemat et al., 2010

No

Yes

Yes

No

No

Atomic commit protocol

Mobile Transaction Processing system

Mobile wireless system

Yes

8

B. Mu et al., 2010

No

Yes

No

No

Yes

ACID properties

Distributed business transaction

yes

9

Sylvia et al., 2009

Decreased

Yes

Yes

No

Yes

Two phase commit protocol, queue sensing distributed real time commit protocol

Distributed real time database system

Transactions over open network

Yes

10

M.Misra et al.,2008

No

Yes

Yes

No

No

Real Time commit Protocols

Distributed Real Time Database systems

No

No

11

T. Wing etal.,2008

No

Yes

Yes

No

No

Grid Transactions

Web Services

No

No

12

U.Shankar etal.,2006

Decreased

Yes

Yes

No

Yes

SWIFT

Distributed Real Time Database Systems

No

Yes

13

L. Frank, 2004

No

Yes

Yes

No

No

ACID properties

Distributed ERP system

E-commerce System

yes

14

G. Alkhatib etal.,2002

No

Yes

Yes

Oracle

No

Two Phase Commit

Distributed Database System

Yes

No

15

M. Abdallah etal.,2002

No

Yes

No

No

Coordinator Logical Log(CLL)

Distributed systems and applications

No

Yes

16

L. Frank , 2001

Decreased

Yes

Yes

No

No

ACID Properties

Web environment

Supplier integrated E-commerce system

Yes

17

S. K. Madria etal., 2001

No

Yes

Yes

Pre-write transaction model

Yes

ACID properties, serializibility

Mobile Transaction Processing system

No

Yes

18

D. Chkliaev etal., 2000

Decreased

Yes

Yes

No

Yes

Two Phase commit, Two Phase Locking

Distributed Transaction Processing System

No

Yes

19

P. Muth etal.,1997

No

Yes

Yes

VODAK

Yes

Heterogeneous 3-level transactional model

Object Oriented Database

Flight Reservation System

Yes

20

Y. Breitbart

etal.,1992

No

Yes

Yes

Multi-database Model

No

Serializability

Heterogeneous databases

No

No

The proposed techniques in table 2 are only tested and verified by A. James et al [4], Y. Jayanta et al [6], B. Mu et al [8], Sylvia et al [9], U.Shankar et al [12], S. K. Madria et al [17], D. Chkliaev et al [18] and P. Muth et al [19]. B. Omar et al [2] used Exceptional distributed transaction model but in contrast A. James et al [4] presented AuTra technique in web environment which results decrease in response time. All the techniques follow two phase commit protocol for implementation of ACID properties in Distributed Database transactions except N. N. Taboudjemat et al [7], T. Wing et al [11], U. Shankar et al [12], M. Abdallah et al [15] and P. Muth et al [19], however rest of techniques have their own significance.

L. Frank [3] , A. James et al [4], N. N. Taboudjemat et al [7], Sylvia et al [9],L. Frank [13], L. Frank [16] and P. Muth et al [19] proposed techniques are supported with the case study and examples while all the rest nor include any case studies neither examples to support their techniques. The variety of user needs in any environment means that there is much need for customization, including relaxation of ACID properties. Many models have been proposed for such support but, due to this variety, there are still some problems that need to be faced like the introduction of a mode that customizes relaxation of all ACID properties according to user requirements. Furthermore there is a need for a transaction management model that incorporates application specific criteria as well as the well known generic ACID criteria to produce a better user service.

Future work could involve developing a model in order to empower users with the knowledge gained through experimentation in order that they can make informed and secure choices about relaxation of ACID properties in the context of seeking improved service.

4. Conclusion

ACID properties in distributed database transactions are very important and difficult to handle, when the data is distributed on different sites. Generalization is important aspect in implementing ACID properties to design models. Distributed database system must cater response time which is of supreme important as far as performance of the transactions is concerned.

In this paper, various techniques and protocols for implementing ACID properties in Distributed database transactions are explained in detail. An evaluation criterion for comparison of techniques has been defined and analysis of these techniques has also been discussed. After analysis, it is concluded that majority of techniques tend to follow ACID properties through two phase commit protocol. The techniques lack response time, testability, tool support and their applications in other domains which are the major limitations of these researches. Therefore, it is suggested that there is need of developing new models which should be generalized. Keeping in view the limitation, performance of the distributed system must be improved which is very important in distributed database.