An asynchronous replication model to improve data available into a heterogeneous system

Computer Engineering and Intelligent Systems www.iiste.org
ISSN 2222-1719 (Paper) ISSN 2222-2863 (Online)
Vol 3, No.8, 2012

An Asynchronous Replication Model to Improve Data Available
into a Heterogeneous System

Mohammad Naderuzzman (Corresponding Author)
Department of Computer Science & Engineering
Dhaka University of Engineering & Technology,
Gazipur, Dhaka
E-mail: nader_u@yahoo.com

Dr. Md. Nasim Akhtar
Department of Computer Science & Engineering
Dhaka University of Engineering & Technology,
Gazipur, Dhaka
E-mail: nasim_duet@yahoo.com
Abstract
Data replication is an important technique for high level data availability, especially applied for distributed
systems. To enhance the data access and data reliability, data replication is one of the most important aspects.
This paper is proposed a asynchronous replication model. This model also supports heterogeneous system which
is currently a very promising system. In our proposed model the main server and replication servers are loosely
coupled. Since the structure is loosely coupled and asynchronous model thus there is less dependency on each
server. Moreover the proposed model is supporting heterogeneous system, so it will be highly cost minimizing
solution for efficient data replication.
Keywords: Data replication, Data Persistency, Asynchronous, Heterogeneous Replication.

I. INTRODUCTION
The heterogeneous computing system is very promising platform. Since to exploit the parallelism, the single
parallel architecture based systems may not be sufficient for a running application. In some cases, heterogeneous
distributed computing (HDC) systems can achieve better performance than the super computer systems;
moreover it involves much lower cost than the super computer. However the HDC system is more exception
oriented which may have negative impact on the running application [1]. Basically the homogeneous computing
system is easier to control, because the processing times are not dependent and identically with an arbitrary
identically distribution [2].
A distributed heterogeneous computing (DHC) system is a collection of autonomous and dissimilar computing
machines those are linked by a network; these computing machines are synchronized by software to function like
a single powerful computing facility. Computer’s tasks are broken into different parts, and because of this it is
possible to distribute the task for parallel execution. A DHC system has some advantages over homogeneous
computing, some parts of an application perform in one system while some parts may perform better in another
system [3].
Heterogeneous computing system may achieve both capability based (aimed at minimizing the completion time
of one big job) and capacity based (aimed at maximizing the number of completions of small jobs within a given
time) jobs [4].
Data replication is computing process that copy and maintain database objects, such as tables, simultaneously in
multiple databases. A change in a main database is reflected, forwarded and applied in each of the replicated
servers which might be located in a remote location [5]. Replicated database and a distributed database may
sounds same. But the difference between them is that, in a distributed database, data are available at many
locations, but a particular table resides only at one location [6]. For example, the employees table resides at only
the pah.employee database in a distributed database system which also includes the kl.employee and kn.world
databases. In replication a 100% same data in different locations [7]. Replication balances the data transactions
and provides fast, local access to shared data over multiple sites through network [8]. Replication works as to
balancing load.
Data replication can be driven by using application programs which transports data to and from other locations
and then load at the receiving location. During the replication, data might be needed to be filtered and
transformed. This replication should not interfere with existing applications and must have minimal impact on

14

Vol 3, No.8, 2012

production system. Replication processes should be under managed and monitored [9]. So, data replication
improves the transaction time, data access time and improves fault tolerance by maintaining and managing
multiple copies of data (e.g. files, objects, databases or part of databases etc.) at different locations [10].
A replication process could be either asynchronous or synchronous replication for copying data. In asynchronous
replication, changes are made after a certain time from the master site to different other site. And in synchronous
replication, changes are made immediately after some data transaction occurs to the master site. Using
asynchronous replication, updates of transactions results efficiently update at all other sites [5]. So asynchronous
replication should be appropriate solution for organizations: seeking the fastest data recovery, minimize data
loss, and protection against database integrity problems. It ensures that a copy of data must be identical to the
primary copy which was created at the time the primary copy updated. The primary or main benefit of
synchronous replication is that, data can be recovered quickly. Operation on remote or mirrored site begins
immediately at any time if primary site stops, all the operations should be disrupted. Only few operations in
process at that instant of disruption may be lost. Because the primary site and the remote site neither of them will
have record of those transactions, at this situation the database rolls back to the last confirmed state [11].
The architectural layer which is persistence layer, whose job is to provide an abstract interface to storage
mechanisms. Such an interface in the API should be abstract and independent of storage technologies. It should
typically have the features like: Store/ retrieve of whole database object by key, Logical database cursor
abstraction for accessing some / all instances of a given type, Transaction support: open, commit, and abort,
rollback and Session management. Also some basic query support, like: how many instances of a given type
exists etc. such layer usually are built from at least two internal layers of software: first being the abstract
interface and second being a set of bindings, one for each targeted database. In practice, there are three layers;
there may be internal division between the logic of object and relational (or other) storage mechanisms [12].
Our proposed layer will use a multi-threaded application and will act like an interface between database and the
main system. In the persistence layer there will have a single thread which will be responsible for making
communication with the main server and will have a different thread running to manage the replicated databases.
It will help the entire system to reduce dependency of the replicated server on the main server. If the main server
crashes it will intelligently start using replicated server. In this way such a system will make no down time for
the entire system and the persistence layer will be like an interface between the database and the entire remote
system. In this way, adding more replicated server will be just like plug and play.

II. RELATED WORK
In grid community, lots of work has focused on providing efficient and safe replication management services for
distributed and clustering system through designing of algorithm and systems. Enterprise business or industrial
business uses replication for many reasons. Replication technology creates the replication of data on the node
that is right for it and from where the data transmission becomes faster. For example in a network some remote
nodes may be far from the main server and in this case data transmission rate will be too high. This case a
replication server can be created on to a remote location which in terms helps the remote system reduce data
transmission impediments and thus improve visit delay, bandwidth consumption and system reliability [13].
Hitoshi Sato et. al [14] proposed an approach for file clustering based replication for working in a grid
environment. His goal was to create a technique to automatically determining the optimal file replication
strategies. The approach outperformed stored group files in a grid file system according to the relationship of
simultaneous file access and it determines locations and movement of replicas of file clusters from the observed
performance data of file access and the implementation specification was in Linux OS. The authors did not
consider the heterogeneous systems and the replication in grid environment needs lot of inter connection speed
e.g. in gigabytes.
In [15] an intelligent replication framework in data grid was proposed by the author. The goal of their approach
was to create a replica management service that interrogate replica’s placement optimization mechanisms and
dynamic replication techniques, which coupled with the algorithms of computations and job scheduling for better
performance in data grid. They used dynamic ordinary replication strategies and replica placement schemes and
their result shows that their approach improves the job execution time by 10% to 23%. However their work
replica management is only coupled with computational job scheduling which actually perform better in
Symmetric Multi-Processors Server (SMP).
In a study a reflected persistence data layer framework based on O/R mapping were designed and implemented
by Yuan-Sheng Lou et. al [16]. In the information system design, persistence data layer is the most important
part. This is the foundation for the performance of the system and its migration ability. In the paper the author
presented five modules: data loadable module, data write module, database services module, primary key cache
module and paging cache module for persistence layer. In this case the reflection is not native to the OS. A lot of

15

Vol 3, No.8, 2012

execution handling mechanism needed to be included in to the system. Besides the replication by using reflection
mechanism is a very slow process and takes a lot of memory involvement and might cause buffer overflow.
For a peer-to-peer network a load sharing technique was proposed by using dynamic replication [17]. For
providing and improving access performance they have proposed two load sharing techniques, which use data
replication. In the first technique they have used a periodic push-based replication (PPR) to reduce the hop count
(the number of legs traversed by a packet) and in the second technique it uses on demand replication (ODR),
which performs and improves access frequency. They have proposed two algorithms to improve access
performance on a P2P network. A P2P network is typically not very much secure and they didn’t have any
strategy for large distributed application.

III. FRAMEWORK OF HETEROGENEOUS ASYNCHRONOUS REPLICATION
A. Multithreading in persistence layer
In the proposed framework, the persistence layer will queue all the transactions of database for the
replicated server that will run in a single thread. On other hand, the persistence layer will have another
thread that got higher priority will make a sound transaction with the main server. All the database
server connection will be configurable by using XML configuration. This way if the system needs
another replicated server, it just needs to modify the configurable XML and plug it to the system and
the data replication will start immediately. In case of database server crashes, the persistence layer will
be intelligent enough to start using sound transaction from one of the replicated server and it will send
system an alert message to the administrator. This will make no down time for the system. This paper
presents a new framework for using in the persistence layer for asynchronous data replication (Figure
1), and they implement the corresponding function. The various modules are describing as follows:
(1) Exception: The Exception module/layer will handle all kind of exception of the system. The
database exception handler will incorporate with this proposed layer. Typically in a system an
exception handler is defined by the occurrence of exceptions, e.g. special conditions that change
the normal flow of program execution. Thus, null pointer, memory overflow, dividing by zero are
those special conditions which in terms handled by the Exception layer. In this proposed system, if
a database server become overloaded or crashes (which will be known as exception) it will look out
of the box and create an exceptions thus send it to exception handling layer. For example, if the
main server crash, the exception layer will get acknowledged immediately and it will then pass the
exception type to the main engine which then take the proper action.
(2) Global Configuration: Global Configuration is a system resource where all the data, classes etc.
will be defined by its values. For an example in a web enterprise application the data flow/routing
class, the url(s) will be defined in the global configuration sections.
(3) Connection String: Connection string is a character string expression which uniquely identifies
the data-store for using by a particular query or set of queries, and methods for connecting. In this
proposed system, connection string will be used to hold all the connection expression between
servers and persistence layer.
The entire system will help any enterprise system more secure and reliable by replicating the
database. Data will be replicated asynchronously to different database server. The entire system
will be much flexible to handle any kind of database server. For example, the system might be
using Oracle server as the main server and other replicated servers can be SQL, MySQL, DB2,
Informix or even MS Access server. The proposed system will develop such system, because a
company might use SQL Server in Windows for the security reason and it will cost a lot of money,
so if the company wants to make the replication through all the SQL Server then it will cost lots of
money for each. But if it is possible to replicate with the open source database like MySQL, then it
will drastically reduce the cost. Besides, in this proposed paper, it is possible to have something
like plug-and-play ability into the system. For example in this system any new replicated server can
be added whenever anyone wants and for this its not necessary to stop for long to all the servers,
just reconfigure the replication and then restart the server. If a replication server is added to the
system, it needs to add some connection string and the API details in the global configuration and
will also have a lookup service on that configuration file. The lookup service will update the
system the system as soon anything added or removed from the configuration file and also it will
have another service that will be known as replication asynchronous service. The job of replication
asynchronous service is to update the newly added replication server with the bulk data from the
main server. This service collects all the data from the main server and update the newly added to
the replication server.

16

Vol 3, No.8, 2012

A proposed algorithm will be use to run the asynchronous replication. In this replication, by using
multi-threading we will make the replication asynchronous. A higher priority thread will keep
consistent connection with the main server. After that the persistence layer will create other many
threads for each of the replication servers. Among all threads a single thread will have the higher
priority and all other will have the lower priority on the system. So, the record will be kept in queue
of those data of saved, deleted or modified, then it will be make its own copy and do the transaction
with that thread.
B. Heterogeneous Replication Process
In the heterogeneous replication system the network architecture is built with different OS. In this
proposed system the software’s main engine(SME) resides in the main server which contains the
independent DLLs (dynamic link library in C# ) or JAR files (in JAVA) to solve complex business
logics. The data connectivity, complex query execution and ORM (Object Relational Mapping)
executes on the persistence layer. This layer also responsible for more different service like Data Queue
(DQ) and replication asynchronous service (RAS) that supports the heterogeneous system (Figure 2).
In the heterogeneous system, through the DQ, all the replication servers are connected from persistence
layer. A multithreading is introduced from the DQ to this system. So, in this system data will flow
through a high priority thread to the server from persistence layer and lower priority thread will be used
for the flow of data to different replication server.
The high level model has been shown in figure 1.
IV. Result & Discussion
The conception of faster replication of data and use of lesser bandwidth is the principle of the persistence layer
framework. The job of the persistence layer is to create an interface with main server, which communicates with
different database servers and relates with them. The procedure ensures efficient way for data replication within
a very short time with better and faster data transmission rate. A Host to Host and Disk to Disk is considered in
this replication. The Persistence Layer Replication’s data transfer rate can be compared with other existing
replication methods, and it can be determined that it works faster than other replication techniques. All different
replication algorithms applied and found different performance found on different data sets, although they took
different time interval for transmission of various data length. In our paper we have proposed that all the data
will be of fixed length. Different many replication mechanisms applied on them fixed length data and the
resultant comparison shows that the asynchronous Persistence Layer Replication takes significantly less time
than other replication methods. The comparison of various replication algorithms were based on the bit rate and
data length. The comparison is shown in figure 2.
From the Figure 2, we can compare very easily that our proposed model is much faster than other replication
models available.

17

Vol 3, No.8, 2012

Figure 1: Asynchronous replication model

Bit Rate
Data Length

Figure 2: Graph of various replication algorithms

18

Vol 3, No.8, 2012

REFERENCES

[1] Xiaoyong Tanga, Kenli Li a,_, Renfa Li a, Bharadwaj Veeravalli,“ Reliability-aware scheduling strategy for
heterogeneous distributedcomputing systems” J. Parallel Distrib. Comput. 70 (2010) 941_952
[2] Xiaonian Tong, Wanneng Shu, “An Efficient Dynamic Load Balancing Scheme for Heterogenous Processing
System”IEEE Conference on Computational Intelligence and Natural Computing, 2009, pp. 319 – 322
[3] Wayne F. Boyera, Gurdeep S. Hura, “Non-evolutionary algorithm for scheduling dependent tasks in
distributed heterogeneous computing environments” J. Parallel Distrib. Comput. 65 (2005) 1035 – 1046
[4] Guest editorial, “Heterogeneous computing” Parallel Computing 31 (2005) 649–652
[5] S Rajasekhar, B Rong, K Y Lai, I Khalil and Z Tari, “Load Sharing in Peer-to-Peer Networks using Dynamic
Replication” Advanced Information Networking and Applications, AINA 2006. 20th International Conference,
April 2006,pp. 1011 – 1016
[6] Randy Urbano, “Oracle Database Advanced Replication”, Part No.B10732-01,Oracle Corporation, 2003,ch.
1, Retrieved February 8, 2010, from http://www.databasebooks.us/oracle_0003.php
[7] Yijie Wang , Sijun Li, “ Research and performance evaluation of data replication technology in distributed
storage systems” August 2006, Computers & Mathematics with Applications 51 (2006) 1625-1632
[8] Bost, Bernadette Charron, Fernando Pedone, Andr´e Schiper,“ Replication Theory and Practice” Berlin
Heidelberg NewYork,
2009, Springer, ISBN-10 3-642-11293-5 Springer, ch. 2, Retrieved 14 February 2010, from
http://www.ebook3000.com/Replication --Theory-and-Practice_36872.html
[9] Yi lin, ,“Practical and consistent database replic ation”, Mc Gill University Montreal, Quebec, 2007,ch. 1-2,
Retrieved February 15,2010, from http://proquest.umi.com.ezproxy.ump.edu.my
[10] Lijun (June) Gu, LIoyd Budd, Aysegul Caycl, Colin Hendricks, Micks Purnell, Carol Rigdon,” Practical
Guide to DB2 UDB Data Replication V8 ”.Durham, NC, USA, 2002, IBM ,ch.1,Retrieved February12, 2010,
from http://site.ebrary.com/lib/kuktemlib/Doc?id=10112697&ppg=23
[11] Uros C ibej , Bostjan Slivnik, Borut Robic, “The complexity of static data replication in data grids”, 2005,
Parallel Computing 31 (2005) 900–912
[12] Hitachi data system, Retrieved February 18, 2010, from
http://www.hds.co.uk/assets/pdf/sb-synchronous-data-replication.pdf
[13] Open Ehr, Retrieved February 18, 2010, from http://www.openehr.org/208- E.html? branch=1&language =
1
[14] Tian Gao, Fang’ai Liu, “A Dynamic Data Replication Technology in Educational Resource Grid”,
Information Technologies and Applications in Education, 2007. ISITAE '07. First IEEE International
Symposium , Nov. 2007, pp. 287 – 291
[15] Hitoshi Sato, Satoshi Matsuoka, and Toshio Endo , “File Clustering Based Replication Algorithm in a Grid
Environment”, 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, June 2009, pp. 204
– 211
[16] Ali Elghirani, Albert Y. Zomaya, Riky Subrata “An Intelligent Replication Framework for Data Grids”,
Computer Systems and Applications, AICCSA '07. IEEE/ACS International Conference, June 2007, pp. 351 –
358
[17] Yuan-sheng Lou, Zhi-jian Wang, Long-da Huang and Lu-lu Yue, “The Study of A Reflected Persistence
Data Layer Framework”, Software Engineering, 2009. WCSE '09. WRI World Congress , November 2009, pp.
291 – 294

19

This academic article was published by The International Institute for Science,
Technology and Education (IISTE). The IISTE is a pioneer in the Open Access
Publishing service based in the U.S. and Europe. The aim of the institute is
Accelerating Global Knowledge Sharing.

More information about the publisher can be found in the IISTE’s homepage:
http://www.iiste.org

The IISTE is currently hosting more than 30 peer-reviewed academic journals and
collaborating with academic institutions around the world. Prospective authors of
IISTE journals can find the submission instruction on the following page:
http://www.iiste.org/Journals/

The IISTE editorial team promises to the review and publish all the qualified
submissions in a fast manner. All the journals articles are available online to the
readers all over the world without financial, legal, or technical barriers other than
those inseparable from gaining access to the internet itself. Printed version of the
journals is also available upon request of readers and authors.

IISTE Knowledge Sharing Partners

EBSCO, Index Copernicus, Ulrich's Periodicals Directory, JournalTOCS, PKP Open
Archives Harvester, Bielefeld Academic Search Engine, Elektronische
Zeitschriftenbibliothek EZB, Open J-Gate, OCLC WorldCat, Universe Digtial
Library , NewJour, Google Scholar

An asynchronous replication model to improve data available into a heterogeneous system

More Related Content

What's hot (19)

Viewers also liked (7)

Similar to An asynchronous replication model to improve data available into a heterogeneous system (20)

More from Alexander Decker (20)

Recently uploaded (20)

An asynchronous replication model to improve data available into a heterogeneous system