A Survey of File Replication Techniques In Grid Systems

International Journal of Computer Applications Technology and Research
Volume 4– Issue 7, 491 - 506, 2015, ISSN:- 2319–8656
www.ijcat.com 491
A Survey of File Replication Techniques In
Grid Systems
Moslem kaviani
Department of
Computer Engineering
Iauksh University
Kermanshah, Iran
Faezeh Pournaghdali
Department of Computer
Engineering
Razi University
Kermanshah, Iran
Mohmood Ahmadi
Department of
Computer Engineering
Razi University
Kermanshah, Iran
Abstract: Grid is a type of parallel and distributed systems that is designed to provide reliable access to data
and computational resources in wide area networks. These resources are distributed in different geographical
locations. Efficient data sharing in global networks is complicated by erratic node failure, unreliable network
connectivity and limited bandwidth. Replication is a technique used in grid systems to improve the
applications’ response time and to reduce the bandwidth consumption. In this paper, we present a survey on
basic and new replication techniques that have been proposed by other researchers. After that, we have a full
comparative study on these replication strategies.
Keywords: Grid environment, Data Grid, Replication, Replication Strategy, Replication Techniques.
1. Introduction
Nowadays, the management of the huge distributed
and shared data resources efficiently around the
wide area networks becoming an important part of
shared recourses. In many fields, which are totally
diverse in nature, like high energy physics,
bioinformatics, earth observations, global climate
changes, image processing, and data mining; the
volume of data of interest is measured in terabytes
and some time in petabytes Which need to be shared
and analyzed[1].
Storing such amount of data in the same location is
difficult, even impossible. Moreover, an application
may need data produced by another geographically
remote application. For this reason, a grid is a large
scale resource sharing and problem solving
mechanism in virtual organizations and is suitable
for the above situation [2]. In addition, users can
access important data that is available only in
several locations, without the overheads of
replicating them locally. These services are
provided by an integrated grid service platform so
that the user can access the resource transparently
and effectively [3].Managing this data in a
centralized location increases the data access time
and hence much time is taken to execute the job. So
to reduce the data access time, "Replication" is used
[4]. When a user generates a request for a file, large
amounts of bandwidth could be consumed to
transfer the file from the server to the client.
Furthermore the latency involved could be
significant considering the size of the files involved.
The main aims of using replication are to reduce
access latency and bandwidth consumption.
Replication can also help in load balancing and can
improve reliability by creating multiple copies of
the same data [5].
The replication is the process of creation and
placement of the copies of entities software. The
phase of creation consists in reproducing the
structure and the state of the replicated entities,
whereas the phase of placement consists in choosing
the suitable slot of this new duplication, according
to the objectives of the replication. So, replication
strategy can shorten the time of fetching the files by
creating many replicas stored in appropriate
locations [6], [7]. By storing the data at more than
one site, if a data site fails, a system can operate
using replicated data, thus increasing availability
and fault tolerance. At the same time, as the data is
stored at multiple sites, the request can find the data
close to the site where the request originated, thus

www.ijcat.com 492
increasing the performance of the system. But the
benefits of replication, of course, do not come
without overheads of creating, maintaining and
updating the replicas [8].
There is a fair amount of work on data replication in
grid environments. Most of the existing work
focused on mechanisms for create, decision and
delete replicas. The purpose of this document is to
review various replication techniques and compare
these techniques which have been presented by
other researches in different distributed
architectures and topologies.
The rest of this paper is organized as follows. In the
third section, we present an overview of grid
systems, types of grids and topologies that exist for
grid systems. The third section describes replication
scenario, challenges and parameters of evaluating
replication techniques. Section four takes a closer
look on basic and new existing data replication
strategies in grid environment. In section five, we
present a comparative study on the replication
techniques that were discussed in the previous
Section. Finally, section six will be reserved for the
conclusion.
1. Motivation
In highly distributed environment of a grid,
availability of data, response time, access cost,
bandwidth consumption, reliability, scalability are
some very important metrics to be considered.
The motivation of this survey is to explore the
existing dynamic replication strategies so that the
researchers can include all the necessary metrics in
their works in this domain and the limitations of the
existing ones can be overcome.
2. Grid Systems
When a user generates a request for a file, large
amounts of bandwidth could be consumed to
transfer the file from the server to the client.
Furthermore the latency involved could be
significant considering the size of the files involved.
Our study investigates the usefulness of creating
replicas to distribute these huge data sets among the
various scientists in the grid. The main aims of
using replication are to reduce access latency and
bandwidth consumption. The other advantages of
replication are that it helps in load balancing and
improves reliability by creating multiple copies of
the same data [9]. In [10] Foster introduced us to a
definition of a grid as follows “coordinated resource
sharing and problem solving in dynamic, multi-
institutional virtual organizations”. There are
different types and topologies of Grid developed to
emphasize special functions that will be defined in
the two next sections.
2.1. Types of Grid
Grid computing can be used in a variety of ways to
address various kinds of application requirements
and it has three primary types. Of course, there are
no hard boundaries between these grid types and
often grids may be a combination of two or more of
these [11]. Types of grids are summarized below:
• Computational grid: Computational grid is
focused on setting aside resources specifically for
computing power. Such as most of the machines
are high-performance servers [11].
• Scavenging grid: Scavenging grid is most
commonly used with large numbers of desktop
machines that are scavenged for available CPU
cycles and other resources. Owners of the desktop
machines are usually given control over when
their resources are available to participate in the
grid [11].
• Data grid: A data grid is a collection of
geographically distributed computer resources.
These resources may be located in different parts
of a country or even in different countries. A grid
connects all these locations and enables them to
share data and other resources [9]. For example,
you may have two universities doing life science
research, each with unique data. A grid connects
all these locations and enables them to share their
data, manage the data, and manage security issues
such as who has access to which data [12].
2.2. Grid Topologies
In this section we present an overview of major grid
topologies. The performance of replication
strategies is highly dependent on the underlying
architecture of grid [13].Graph and tree models are
used where there is a single source for data and the
data has to be distributed among collaborations
worldwide [13]. The Figure 1 and Figure 2, shows
the Graph and tree models respectively.

www.ijcat.com 493
A tree topology also has shortcomings. The tree
structure of the grid means that there are specific
paths to the messages and files can travel to get to
the destination. Furthermore, data transference is
not possible among sibling nodes or nodes situated
on the same tier [13].
Figure 1.An example of Graph topology.
Figure 2.An example of Tree topology.
Peer to Peer (P2P) systems overcome these
limitations and offer flexibility in communication
among components. A P2P system is characterized
by the applications that employ distributed
resources to perform functions in a decentralized
manner. From the viewpoint of resource sharing, a
P2P system overlaps a grid system. The key
characteristic that distinguishes a P2P system from
other resource sharing systems is its symmetric
communication model between peers, each of which
acts as both a server and a client [13]. The Figure 3
shows an example of the P2P structure.
Hybrid Topology is simply a configuration that
contains an architecture consisting of any
combination of the previous mentioned topologies.
It is used mostly in situations where researches
working on projects want to share their results to
further research by making it readily available for
collaboration [13]. A hybrid model of a graph grid
with peer linkages at the edges is shown in Figure 4.
A hybrid topology can carry features of both tree
and P2P architectures and thus can be used for
better performance of a replication strategy [12].
Figure 3.An example of Peer to Peer topology.
Figure 4.An example of Hybrid topology.
2.3. Classification of existing data
replication techniques
The data replication algorithm has to answer critical
questions such as which data must be replicated and
where the replica must be placed. The dynamic
behavior of grid users makes it difficult to make
decisions regarding the data replications to attain
the target of maximum availability.
Replication techniques can be classified into two
main streams, static replication and dynamic
replication. In a static replication strategy, the
number of replicas and the host node is chosen
statically at the start of the life cycle, no more
replicas are created or migrated after that [14]. On
the other hand, dynamic strategies adapt to changes
in user request pattern, storage capacity and

www.ijcat.com 494
bandwidth and can create replicas on new nodes and
can delete replicas that are no longer required
depending upon the global information of the data
grid [9]. The dynamic strategies are better than the
static ones because they can make intelligent
decisions about the placement of data depending
upon the information of the grid environment.
Simultaneously, there are drawbacks as well; a
replication decision center is required in a data grid
which needs to collect the runtime information of
all the nodes in a complex grid infrastructure. The
overload of this central decision center further
increases if the nodes in a data grid enter and quit
frequently. In case of the decentralized approach,
further synchronization is involved making the task
difficult. Hence, the focus of this paper is only the
dynamic replication strategies keeping in mind that
the static replication strategies, though very simple
to implement, are less useful.
2.4. Data Replication challenges
Using replication strategies in grid environment
may cause some challenges. The four important
challenges in replicated environments areas follow
[8]:
• Time of creation of a new replica: If strict data
consistency is to be maintained, performance is
severely affected if a new replica is to be created.
As sites will not be able to fulfill request due to
consistency requirements.
• Data Consistency: Maintaining data integrity and
consistency in a replicated environment is of
prime importance. High precision applications
may require strict consistency of the updates
made by transactions.
• Lower write performance: Performance of write
operations can be dramatically lowered in
applications requiring high updates in replicated
environment, because the transaction may need to
update multiple copies.
• Overhead of maintenance: If the files are
replicated at more than one site, it occupies
storage space and it has to be administered. Thus,
there are overheads in storing multiple files.
2.5. Data Replication evaluation
All replication strategies use subset of these
parameters [15]:
• Access latency: Almost all the replications
strategies try to reduce the access latency thus
reducing the job response time and hence increase
the performance of the grids.
• Bandwidth consumption: Similarly almost all the
replication strategies try to reduce the bandwidth
consumption to improve the availability of data
and performance of the system. The target is to
keep the data as close to the user as possible, so
that data can be accessed efficiently.
• Balanced workload: Some of the replication
strategies explicitly target to provide a balanced
workload on all the data servers. This helps in
increasing the performance of the system and
provides better response time.
• Maintenance cost: With more number of replicas
in a system the cost of maintaining them becomes
an overhead for the system. Some of the
strategies aim to make only an optimal number of
replicas in the data grid. This ensures that the
storage is utilized in an optimal way and the
maintenance cost of replica is minimized.
• Strategic placement: Some strategies target the
strategic placement of the replicas along with an
optimal number of replicas. The strategic
placement of replicas is a very important factor
because it is integrated with few other very
important factors. For example, if the replicas are
placed on the optimal locations it helps to
optimize the workload of different servers. It is
also related with the cost of the maintenance.
• Job execution time: another very important
parameter is Job execution time. Some replication
strategies target to minimize the job execution
time with optimal replica placement. The idea is
to place the replicas closer to the users in order to
minimize the response time, and thus the job
execution time. This will increase the throughput
of the system.
• Fault tolerance, quality assurance: Only a few
replication strategies have considered replication
as an option to provide fault tolerance and quality
assurance.
3. Replication Techniques
The role of a replication strategy is to identify when
a replica should be created, where to place replicas,
when to remove replicas and how to locate the best
replica. Several replication replacement strategies

www.ijcat.com 495
have been proposed in the past and they are the
basics of other replication algorithms.
3.1. Techniques for peer to peer
architecture
Ranganathan et al. [16] have presented a Dynamic
Model-Driven Replication strategy which proposes
that data availability can be improved in large peer
to peer communities. In this approach the peers can
automatically produce the replicas in a
decentralized fashion whenever it is required to
improve the availability of data. Each peer has a set
of tools by which it can find out the state of the
system to take the replication decision.
The proposed replication strategy considers the
following parameters: (1) Average probability of a
node being up. (2) The transfer time to replicate
data from one node to another node. (3) The storage
cost of file F at a given node. (4) The accuracy of
replica location method. The algorithm works in
following steps:
1)Collect the above mentioned parameters.
2)Using those parameters it finds out a required
number of replicas (r) for a file.
3)Using a replica locator service finds out existing
number of replicas (M).
4)If M >r wait and check again.
5)If M <r use a resource discovery procedure to
find a host for replica.
6)Send replica to the selected remote host. The
target of this approach is to find an optimal
number of replicas and determine the best host
for a new replica.
Abdullah [17] presented a P2P model in 2008 for
higher availability, reliability, and scalability. In
proposed models all the peers operate independently
within a peer group. All peers working in a group
agrees upon a common set of services. Peers can
join or leave a group at any time. When a peer joins
a group it can share the data of other peers, and its
own data is shared by others. A peer can be a
member of more than one group at a time. Peers can
share the data sets with each other without knowing
from which peer they are getting the data. The
process of data discovery starts when a request is
forwarded to all the neighbors depending upon
information stored in the routing table. Two are
newly proposed in this research: ‘‘path and
requestor node placement strategy’’, and ‘‘N-hop
distance node placement’ ’strategy. In the
‘‘requestor node placement strategy’’ a required file
when found is replicated to the requestor node only.
In the ‘‘path node placement strategy’’ the file is
replicated to all the nodes on the path from
requestor node to provider node. The newly
proposed strategy ‘‘path and requestor node
placement strategy’’ is a combination of the first
two strategies. In ‘‘N-hop distance node
placement’’ a file is replicated to all neighbors of
provider nodes within an n hop distance.
Cashing plus Cascading combines cascading and
plain cashing strategies. The client caches file
locally, and the server periodically identifies the
popular files and propagates them down the
hierarchy. Note that the clients are always located at
the leaves of the tree but any node in the hierarchy
can be a server. Specifically, a Client can act as a
Server to its siblings. Siblings are nodes that have
the same parent [12].
3.2. Techniques for multi-tier tree
architecture
The multitier topology provides a very economical
and efficient way to share the storage,
computational and the network resources. It allows
hundreds and thousands of users to share the
common resources efficiently. Ranganathan et al.
[9] have proposed six different replication
techniques for three different access patterns in
2001. The main aims are reduced access latency and
bandwidth consumption.
1)No Replication: in this case only the root node
contains the replicas.
2)Best Client: a replica is created for the client who
accesses the file the most.
3)Cascading: a replica is created on the path of the
best client.
4)Plain Caching: a local copy is stored upon initial
request.
5)Caching plus cascading: combines plain caching
and cascading strategies.
6)Fast Spread: file copies are stored at each node on
the path to the best client.
• NO Replication strategy will not create replica
and therefore, the files are always accessed
remotely. One example of the implemented

www.ijcat.com 496
strategy is the Simple Optimizer algorithm, which
never performs replication; rather it reads the
required replica remotely. Simple Optimizer
algorithm is simple to implement and performs
the best relative to other algorithms in terms of
the storage space usage, but performs the worst in
terms of job execution time and network usage
[12].
• Best client creates replica at the client that has
generated the most requests for a file, this client
is called the best client. At a given time interval,
each node checks to see if the number of requests
for any of its file has exceeded a threshold, then
the best client for that file is identified [12].
• Cascading Replication supports tree architecture
the data files generated in the top level and once
the number of accesses for the file exceeds the
threshold, then a replica is created at the next
level, but on the path to the best client, and so on
for all levels, until it reaches to the best client
itself [12].
• Plain Cashing: The client that requests a file
stores a copy locally. If these files are large and a
client has enough space to store only one file at a
time, then files get replaced quickly [12].
• Fast Spread: In this method a replica of the file is
stored at each node along its path to the client.
When a client requests a file, a copy is stored at
each tier on the way. This leads to a faster spread
of data. When a node does not have enough space
for a new replica it deletes the least popular file
that had come in the earliest [12].
• Simple Bottom-Up (SBU) and Aggregate
Bottom-Up (ABU) are two dynamic replication
mechanisms that are proposed in the multi-tier
architecture for data grids. The SBU algorithm
replicates the data file that exceeds a pre-defined
threshold for clients. The main shortcoming of
SBU is the lack of consideration to the
relationship with historical access records. For the
sake of addressing the problem, ABU is designed
to aggregate the historical records to the upper
tier until it reaches the root. The results shown
improvements against Fast Spread strategy. The
values for interval checking and threshold were
based on data access arrival rate, data access
distribution and capacity of the replica servers
[18].
• Proportional Share Replica (PSR) policy is an
improvement in Cascading technique. The
method is a heuristic one that places replicas on
the optimal locations by assuming that the
numbers of sites and the total replicas to be
distributed are already known. Firstly an ideal
load distribution is calculated and then replicas
are placed on candidate sites that can service
replica requests slightly greater than or equal to
that ideal load [19].
• Multi-objective approach is a method exploiting
operations research techniques that is proposed
for replica placement. In this method, replica
placement decision is made considering both the
current network status and data request pattern.
The problem is formulated in p-median and p-
center models to find the p replica placement
sites. The p-center problem targets to minimize
the max response time between user site and
replica server whereas the p-median model
focuses on minimizing the total response time
between the requesting sites and the replication
sites [20].
• Latest Access Largest Weight (LALW) is a
dynamic data replication mechanism. LALW
selects a popular file for replication and calculates
a suitable number of copies and grid sites for
replication. By associating a different weight to
each historical data access record, the importance
of each record is differentiated. A more recent
data access record has a larger weight. It indicates
that the record is more pertinent to the current
situation of data access [21].
• Adaptive Popularity Based Replica Placement
(APBRP) is a dynamic replica placement
algorithm, for hierarchical data grids which is
guided by “file popularity”. The goal of this
strategy is to place replicas close to clients to
reduce data access time while still using network
and storage resources efficiently. The
effectiveness of APBRP depends on the selection
of a threshold value related to file popularity.
APBRP determines this threshold dynamically
based on data request arrival rates [22].

www.ijcat.com 497
• Predictive hierarchical fast spread (PHFS) is a
dynamic replication method in multi-tier data grid
environments which is an improve version of
common fast spread. The PHFS tries to forecast
future needs and pre-replicates the min hierarchal
manner to increase locality in accesses and
improve performance that consider spatial
locality. This method is able to optimize the
usage of storage resources, which not only
replicates data objects hierarchically in different
layers of the multi-tier data grid for obtaining
more localities in accesses. It is a method
intended for read intensive data grids. The PHFS
method use priority mechanism and replication
configuration change component to adapt the
replication configuration dynamically with the
obtainable condition. Besides that, it is developed
on the basis of the concept that users who work
on the same context will request some files with
high probability [23].
• Dynamic Hierarchical Replication (DHR) is a
dynamic replication algorithm for hierarchical
structure that places replicas in appropriate sites.
Best site has the highest number of access for that
particular replica. This algorithm minimizes
access latency by selecting the best replica when
various sites hold replicas. The replica selection
strategy of DHR algorithm, selects the best
replica location for the users running jobs by
considering the replica requests that waiting in
the queue and data transfer time. It stores the
replica in the best site where the file has been
accessed most, instead of storing files in many
sites [24].
• Modified Latest Access Largest Weight
(MLALW) is a dynamic data replication strategy.
This strategy is an enhanced version of Latest
Access Largest Weight strategy. MLALW deletes
files by considering three important factors:
1. Least frequently used replicas
2. Least recently used replicas
3. The size of the replica
MLALW stores each replica in an appropriate site
in the region that has the highest number of
access in future for that particular replica. The
experiment results show that MLALW strategy
gives a better performance compared to the other
algorithms and prevents unnecessary creation of
replica which leads to efficient storage usage
[25].
• The combination of Modified DHR Algorithm
(MDHRA) and Combined Scheduling Strategy.
MDHRA considers three factors for replacing
replicas: (1) the last time the replica was
requested; (2) number of accesses and (3) size of
replica. It computes the response time based on
the data transfer time, storage access latency, the
unprocessed replica requests and the distance
between nodes. The computed response time is
used to select the best replica location [26].
3.3. Techniques for general grid
topology
In 2004 [27] Park et al. proposed an internet
hierarchy based replication strategy called BHR to
reduce the data access time by avoiding network
congestions. In proposed model there can be
different network regions combined with each other.
If a required file is present within a region there will
be less number of routers in path, but if the file has
to be fetched across the region from another region,
there will be more number of routes in the path.
Within a region there will be broad bandwidth
available. Network level locality means that if the
required file is fetched from the site having broad
bandwidth, it will reduce the response time
significantly.
Bandwidth Hierarchy Replication (BHR) is a novel
dynamic replication strategy which reduces data
access time by avoiding network congestions in a
data grid network. With BHR strategy, we can take
benefits from “network-level locality” which
represents that required file is located in the site
which has broad bandwidth to the site of job
execution. BHR strategy was evaluated by
implementing in OptorSim simulator and the results
show that BHR strategy can outperform other
optimization techniques in terms of data access time
when hierarchy of bandwidth appears in Internet.
BHR extends current site-level replica optimization
study to the network-level [27].
3.4. Techniques for general graph
architecture
In 2005 Rehman et al. [28] presented six dynamic
replication strategies based on utility and risk for

www.ijcat.com 498
two different kinds of access patterns. Before
placing a replica at a site they considered both
utility and risk index for each site according to the
current network load and user requests. A site with
optimized utility and risk index is than chosen for
replication. They have used Graph as architecture to
represent a data grid, which is closer to the real life
grid scenario than a typical hierarchal architecture.
Their model uses average response time a basis for
comparison among various replication strategies.
The best replication strategy has lower response
time. The algorithms proposed based on utility
select a replica site assuming that the future requests
and current loads and user requests. Algorithms
proposed based on risk index expose sites far from
all other sites and assume a worst case whereby
future requests will primarily originate from here.
The replication algorithm selects one site per
iteration to host a replica.
Bsoul et al. [29] in 2010 have presented an
Enhanced Fast Spread replication technique for data
grid. EFS consider the number and frequency of
requests, size of replica and last time the replica was
requested while making the replication decision.
They have considered a network topology (which is
a complete graph) in which there is one server node
and many clients. The server node has the main
storage with all the data and the clients have less
storage space available as compared to the server.
Whenever a file is required it is first searched
locally, and is used if found. If it is not available
then the client traverses the shortest path until it
finds the required file. The fast spread strategy
replicates the file on all nodes along the path. If
storage on any node is not enough if remove some
file(s) to make room for new replica using LRU or
LFU.
• Agent-based replica placement algorithm is
proposed to determine the candidate site for the
placement of replica. For each site that holds the
master copies of the shared data files will deploy
an agent. The main objective of an agent is to
select a candidate site for the placement of a
replica that reduces the access cost, network
traffic and aggregated response time for the
applications. Furthermore, in creating the replica
an agent prioritizes the resources in the grid based
on the resource configuration, bandwidth in the
network and insists for the replica at their sites
and then creates a replica at suitable resource
locations.
• Least Frequently Used (LFU) strategy always
replicates files to local storage systems. If the
local storage space is full, the replica that has
been accessed the fewest times is removed and
then releases the space for new replica. Thus,
LFU deletes the replica which has less demand
(less popularity) from the local storage even if the
replica is newly stored [30].
• Least Recently Used (LRU) strategy always
replicates files to local storage system. In LRU
strategy, the requested site caches the required
replicas, and if the local storage is full, the oldest
replica in the local storage is deleted in order to
free the storage. However, if the oldest replica
size is less than the new replica, the second oldest
file is deleted and so on [30].
• Weight-based dynamic replica replacement
strategy calculates the weight of replica based on
the access time in the future time window on the
last access history. After that, calculate the access
cost which embodies the number of replicas and
the current bandwidth of the network. The
replicas with high weight will be helpful to
improve the efficiency of data access, so they
should be retained and then the replica with low
weight will not make sense to the rise of data
access efficiency, and therefore, should be
deleted. The access history defines based on the
zipf-like distribution [31].
• Efficient Replication strategy is a replication
strategy for dynamic data grids, which take into
account the dynamic of sites. This strategy can
increase the file availability, improved the
response time and can reduce the bandwidth
consumption. Moreover, it exploits the replicas
placement and file requests in order to converge
towards a global balancing of the grid load. This
strategy will focus on read-only-access as most
grids have very few dynamic updates because
they tend to use a "load" rather than "update"
strategy.
There are three steps provided by this algorithm,
which are:

www.ijcat.com 499
1)Selection of the best candidate files for
replication; Selected based on requests number
and copies number of each files.
2)Determination of the best sites for files placement
which are selected in the previous step; Selected
based on requests number and utility of each site
regarding to the grid.
3)Selection of the best replica; Taking account the
bandwidth and the utility of each site [32].
• Value-based replication strategy (VBRS) is
proposed to decrease the network latency and
meanwhile to improve the performance of the
whole system. In VBRS, threshold was made to
decide whether to copy the requested file, and
then solve the replica replacement problem.
VBRS has two steps; At the first steps, the
threshold will be calculated to decide whether the
requested file should be copied in the local
storage site. Then at the second stage, the
replacement algorithm will be triggered when the
requested file needs to be copied at the local
storage site does not have enough space. The
replica replacement policy is developed by
considering the replica’s value which is based on
the file’s access frequency and access time. The
experiment results show that the effectiveness of
VBRS algorithm can reduce network latency
[33].
• Enhance Fast Spread (EFS) is an enhanced
version of Fast Spread for replication strategy in
the data grid. This strategy was proposed to
improve the total of response time and total
bandwidth consumption. Its takes into account
some criteria such as the number and frequency
of requests, the size of the replica and the last
time the replica was requested. EFS strategy
keeps only the important replicas while the other
less important replicas are replaced with more
important replicas. This is achieved by using a
dynamic threshold that determines if the
requested replica should be stored at each node
along its path to the requester [34].
4. Comparative Study
In this section, we present a full comparative study
on the replication techniques that were discussed in
the previous section.Benefits of data replication
strategies are:
Availability: All the replication strategies aim to
provide maximum availability. Rather, it would be
better to say that replication is the only way to
improve availability of data: generally in all
distributed database environments and specifically
in data grids.
Reliability: When replication increases the
availability, the reliability is improved as well. The
more the number of replicas more is the chance that
user’s request will be serviced properly, and hence
systems is more reliable.
Scalability: It is another important metric that must
be considered by a replication algorithm. The extent
to which scalability can be provided depends upon
the architecture chosen for the data grid. Different
architectural models support different levels of
scalability. That means, scalability is more
dependent on model than replication algorithm.
Adaptability: This is a very important parameter
which must be provided by a replication strategy.
The nature of the grid is very dynamic. Nodes keep
on entering and leaving the grid very frequently.
The replication algorithm must be adaptive to
provide support to all nodes present in a data grid at
any given time.
Performance: As the availability of data increases
the performance of the data grid environment
increases.
All replication strategies use any subset of these
parameters.
• Reduced access latency.
• Reduced bandwidth consumption.
• Balanced workload.
• Less maintenance cost.
• Strategic replica placement.
• Job execution time.
• Increased fault tolerance.
• Quality assurance.
These replication strategies that discussed in the
previous section are compared in the Table 1, Table
2, and Table 3.

www.ijcat.com 500
Table 1: Replication Technique for Tree topology
Replication
Technique
Year Performance Metric Method Additional Feature
Best client 2001
Response time ,
Bandwidth conservation
Replicates file to site that
generates maximum number
of requests
Need to compute
number of request
for each file
Cascading 2001
Response time ,
If number of requests exceeds
threshold then replica trickles
down to lower tier
Need to define a
threshold for
number of requests
Plain Cashing 2001
Response time ,
A requesting client receives
the file and stores a replica of
it locally
Fast Spread 2001
Response time ,
If a client requests a file then
a replica of file stores at each
node along the path toward
the client
Need to storing
request history to
avoid of double
replicating
Simple Bottom-
Up (SBU)
2005
Replication frequency,
Bandwidth cost, Response
time
Creates replicas as close as
possible to the clients that
request the data files with
high rates exceeding the pre-
defined threshold
Need to process
records in the access
history individually
Aggregate
Bottom-Up
(ABU)
2005
Replication frequency,
Bandwidth cost, Response
time
Aggregates the history records
to the upper tier step by step
till it reaches the root
Need to access
history
Proportional Share
Replica (PSR)
2004 Mean of response time
Calculates an ideal workload
and distributes replicas
Need to define ideal
workload
Multi-objective 2006 Average response time
Reallocates replicas to new
candidate sites if a
performance metric degrades
significantly over best k-time
periods
Need to calculate
replica relocation
cost
Latest Access
Largest Weight
(LALW)
2008
Network usage, Mean job
execution time
Selects a popular file for
replication and calculates a
suitable number of copies and
grid sites for replication
Need to find out a
popular file and
suitable site
Adaptive
Popularity Based
Replica Placement
(APBRP)
2010
Storage cost, Average
bandwidth cost, Job
execution time
Selects a threshold value
related to file popularity and
places replicas close to clients
to reduce data access time
while still using network and
storage resource efficiency
Need to determines
threshold value
dynamically, based
on data request
arrival rates
Predictive
hierarchical fast
spread (PHFS)
2011 Average access latency
Tries to forecast future needs
and pre-replicates the min
hierarchical manner. Uses the
hierarchical replication to
optimize the utilization of
resources
Need to considering
spatial locality and
using predictive
methods
Dynamic
Hierarchical
2012 Mean job execution time
Selects best replica when
various sites hold replicas.
Need to access
history

www.ijcat.com 501
Replication
(DHR)
Places replicas in appropriate
sites that has the highest
number of access for that
particular replica
`Modified Latest
Access Largest
Weight
(MLALW)
2012
Effective network usage,
Mean job execution time
Stores each replica in an
appropriate site. Deletes files
by considering least
frequently used replicas, least
recently used replicas and the
size of the replica factors
Need to LRU lists of
replicas, LFU lists
of replicas and
access history
Modified
Dynamic
Hierarchical
Replication
Algorithm
(MDHRA)
2013
minimizes the bandwidth
consumption thus
reducing the network
traffic
replaces replicas based on the
last time the replica was
requested, number of access,
and file size of replica
Need to the last
time the replica was
requested, number
of accesses and size
of replica
Table 2: Replication Technique for peer to peer & general grid topology
Replication
Technique
Year
Performance
Metric
Method Additional Feature
Cashing plus
cascading
2001
Response time ,
Bandwidth
conservation
Joining two replication
techniques:
Cashing and cascading
techniques
Need to define a
threshold for
number of requests
Bandwidth Hierarchy
Replication (BHR)
2004
Total job execution
time
Replicates files which
are likely to be used
frequently within the
region in near future
Need to define
network-level
locality and regions
5. Conclusion
Replication enables faster access to files, decreases
bandwidth consumption, and distributes server load.
In contrast to static replication, dynamic replication
automatically creates and deletes replicas according
to changing access patterns, and thus ensures that
the benefits of replication continue even if user
behavior changes.
In this paper, a review and a comparative study has
been done on basic and new replication techniques
that have been implemented in grids. After a brief
introduction, an overview of grid systems, types of
grids and grid topologies were presented in Section
3. In Section 3, replication scenario, challenges and
ways of evaluating replication techniques were
described. In Section 4, we have presented a
classification of dynamic replication strategies for
grid environment. In Section 5, a full comparative
study was presented on the replication techniques
that were discussed in Section 4. And finally, in
this section, Table 4 is presented that shows the
results of discussed replication techniques.
From this survey it can be seen that there is still a
lot of work to be done in the field of data replication
in grid environment.

www.ijcat.com 502
Table 3: Replication Technique for Graph topology
Replication
Technique
Year Performance Metric Method Additional Feature
Agent-based
replica
2009
Execution time test,
Data availability test
By an agent for each site that
holding the master copies,
select a candidate site for the
placement of replica that
exceeds the conditions
Need to define agents
Least Frequently
Used(LFU)
2003
Job execution time Always replicates files to
local storage , if no space :
delete least accessed files
Need to files access
history
Least Recently
Used (LRU)
2003
Job execution time Always replicates files to
local storage , if no space :
delete oldest file in the storage
Need to files access
history
Weight-based
dynamic replica
2008
Effective network
usage, Mean job
execution time
Calculates the weight of
replica based on the access
time in the future time
window, based on the last
access history
Need to access history
that define based on
zip-like distribution
Efficient
Replication
2010
Response time,
Effective Network
Usage
Takes into account the
dynamic of sites. Exploits the
replicas placement and file
request in order to converge
towards a global balancing of
grid load
Need to considering
dynamicity of sites
Value-based
replication
strategy (VBRS)
2010
Mean job time,
Effective Network
Usage
Calculates the ideal threshold
to decide whether the file
should be copied or not.
Chooses the replica that
should be replaced based on
the values of the local replicas
Need to define
threshold
Enhance Fast
Spread (EFS)
2011
Total response time,
Total bandwidth
consumption
Uses a dynamic threshold that
determines if the requested
replica should be stored at
each node along its path to the
requester. Keeps only the
important replicas while other
less important replicas are
replaced with more important
replicas
Need to frequency of
requests, the size of
the replica and the last
time that the replica
was requested

www.ijcat.com 503
Table 4 : SUMMARIZES THE MAJOR RESULTS OF REPLICATION TECHNIQUES IN
GRIDS
Replication Technique Result proper strategies
Cascading[12] Has an small degree of locality Tree
Cashing[12] High response time Tree
Cascading plus Cashing[12] Better performance than cascading
Better performance than cashing
Peer to Peer
Fast Spread[12] High I/O and CPU load
High storage request
Tree
Proportional Share Replication
(PSR)[19]
Better results over cascading technique
Tree
Bandwidth Hierarchy
Replication (BHR)[27]
Better total job times than LRU and LFU
Peer to Peer
Simple Bottom-Up (SBU)[18]
Aggregate Bottom-Up (ABU
)[18]
Better results over Fast Spread technique
Tree
Multi-objective approach[20] Good performance in dynamic
environments
Tree
Weight-based replication[31] Better performance than LRU and LFU Graph
Least Access Largest Weight
(LALW)[21]
Better job execution time and effective
network usage than LRU, LFU and BHR
Tree
Adaptive Popularity Based
Replica Placement (APBRP)[22]
Better performance than Best client,
Cascading, Fast Spread, ABU and LRU
Tree
Efficient replication strategy[32] Improves the response time
Increases data availability
Reduces bandwidth consumption
Graph
Value Based Replication
Strategy (VBRS)[33]
Decreases network latency
Improves performance of the hole system Graph
Enhanced Fast Spread (EFS)[34] Improves total of response time
Improves total bandwidth consumption
Enhanced version of Fast Spread for
replication strategy in data grid
Graph
Predictive Hierarchical Fast
Spread (PHFS)[23]
Optimizes the utilization of resources
Decreases access latency in multi-tier data
grids
Improved version of common Fast Spread
Lower latency and better performance
compared with common Fast Spread
Tree
Dynamic Hierarchical
Replication (DHR)[24]
Prevents unnecessary creation of replica
Efficient storage usage
Minimizes access latency
Tree
Modified Least Access Largest
Weight (MLALW)[25]
Modified version of LALW strategy
Better performance than LRU, LFU, BHR,
LALW and DHR
Tree
Modified Dynamic Hierarchical
Replication Algorithm
(MDHRA)[26]
Modified DHR Algorithm is compared
with LFU, LRU, BHR, Modified BHR,
3LHA and DHR and was proved the
best
Tree

www.ijcat.com 504
6. References
[1] M. Tang, B.S. Lee, C.K. Yeo, X. Tang,
Dynamic replication algorithms for the
multi-tier data grid, Future Generation
Computer Systems 21 (5) (2005) 775–
790.
[2] N. Mohd. Zin, A. Noraziah, A. Che Fauzi,
and T. Herawan, Replication Techniques
in Data Grid Environments, in Intelligent
Information and Database Systems, vol.
7197, Eds. Springer Berlin, Heidelberg,
pp. 549–559, 2012.
[3] F.B. Charrada, H. Ounelli, and H.
Chettaoui, Dynamic period vs static
period in data grid replication, in
International Conference on P2P, Parallel,
Grid, Cloud and Internet Computing
(3PGCIC), pp. 565–568, 2010.
[4] N. Mohd. Zin, A. Noraziah, A. Che Fauzi,
and T. Herawan, Replication Techniques
in Data Grid Environments, in Intelligent
Information and Database Systems, vol.
7197, Eds. Springer Berlin, Heidelberg,
pp. 549–559, 2012.
[5] K. Ranganathan and I. Foster, Identifying
dynamic replication strategies for a high-
performance data grid, Grid Computing,
pp. 75–86, 2001.
[6] S. Venugopal, R. Buyya, and K.
Ramamohanarao, A taxonomy of data
grids for distributed data sharing,
management, and processing, in Acm
Computing Surveys, vol. 38, no. 1, p. 3,
2006.
[7] K. Ranganathan, A. Iamnitchi, and I.
Foster. Improving data availability
through dynamic model-driven replication
in large peer-to-peer communities, in 2nd
IEEE/ACM International Symposium on
Cluster Computing and the Grid, pp. 376–
376, 2002.
[8] S. Goel and R. Buyya, Data replication
strategies in wide area distributed
systems, in Enterprise Service
Computing: From Concept to
Deployment, vol. 17, 2006.
[9] K. Ranganathan, I. Foster, Design and
evaluation of dynamic replication
strategies for a high performance data
grid, in: International Conference on
Computing in High Energy and Nuclear
Physics, vol. 2001, 2001.
[10] I. Foster, C. Kesselman, and S. Tuecke,
The Anatomy of the Grid: Enabling
Scalable Virtual organizations, in
International Journal of High Performance
Computing Applications, vol. 15, no. 3,
pp. 200–222, 2001.
[11] B. Jacob, Grid computing: what are the
key components, IBM Developer Works,
2003.
[12] K. Ranganathan and I. Foster, Identifying
dynamic replication strategies for a high-
performance data grid, Grid Computing,
pp. 75–86, 2001.
[13] Q. Rasool, J. Li, G. Oreku, E. Munir, and
D. Yang, A Comparative Study of Replica
Placement Strategies in Data Grids, in
Advances in Web and Network
Technologies, and Information
Management, vol. 4537, Springer Berlin,
Heidelberg, pp. 135–143, 2007.
[14] O. Tatebe, Y. Morita, S. Matsuoka, N.
Soda, S. Sekiguchi, Grid datafarm
architecture for petascale data intensive
computing, in: CCGrid, 2002, p. 102.
[15] T. Amjad, M. Sher, and A. Daud, A
survey of dynamic replication strategies
for improving data availability in data
grids, in Future Generation Computer
Systems, vol. 28, no. 2, pp. 337–349,
2012.
[16] K. Ranganathan, A. Iamnitchi, I. Foster,
Improving data availability through
dynamic model-driven replication in large
peer-to-peer communities, in: CCGrid,
2002, p. 376.

www.ijcat.com 505
[17] A. Abdullah, M. Othman, H. Ibrahim,
M.N. Sulaiman, A.T. Othman,
Decentralized replication strategies for
P2P based scientific data grid, in:
Information Technology, 2008, ITSim
2008, International Symposium on, vol. 3,
pp. 1–8.
[18] M. Tang, B.S. Lee, C.K. Yeo, and X.
Tang, Dynamic replication algorithms for
the multi-tier Data Grid, in Future
Generation Computer Systems, vol. 21,
no. 5, pp. 775–790, 2005.
[19] J. Abawajy, Placement of File Replicas in
Data Grid Environments, in
Computational Science ( ICCS), vol.
3038, Springer Berlin, Heidelberg, pp.
66–73, 2004.
[20] R.M. Rahman, K. Barker, and R. Alhajj,
Replica placement designs with static
optimality and dynamic maintainability,
in Sixth IEEE International Symposium
on Cluster Computing and the Grid
(CCGRID), vol. 1, pp.4, 2006.
[21] R.S. Chang and H.P. Chang, A dynamic
data replication strategy using access-
weights in data grids, in The Journal of
Supercomputing, vol. 45, no. 3, pp. 277–
295, 2008.
[22] M. Shorfuzzaman, P. Graham, and R.
Eskicioglu, Adaptive popularity-driven
replica placement in hierarchical data
grids, in The Journal of Supercomputing,
vol. 51, no. 3, pp. 374–392, 2010.
[23] L.M. Khanli, A. Isazadeh, and T.N.
Shishavan, PHFS: A dynamic replication
method, to decrease access latency in the
multi-tier data grid, in Future Generation
Computer Systems, vol. 27, no. 3, pp.
233–244, 2011.
[24] N. Mansouri and G.H. Dastghaibyfard, A
dynamic replica management strategy in
data grid, in Journal of Network and
Computer Applications, vol. 35, no. 4, pp.
1297–1303, 2012.
[25] N. Mansouri, An Effective Weighted Data
Replication Strategy for Data Grid, in
Australian Journal of Basic and Applied
Sciences, vol. 6, no. 10, pp. 336–346,
2012.
[26] N. Mansouri , Gh. Dastghaibyfard,
Combination of data replication and
scheduling algorithm for improving data
availability in Data Grids, in Journal of
Network and Computer Applications, pp.
711–722, 2013.
[27] S.M. Park, J.H. Kim, Y.B. Ko, W.S.
Yoon, Dynamic data grid replication
strategy based on Internet hierarchy, in:
Grid and Cooperative Computing, 2004,
pp.838–846
[28] R.M. Rahman, K. Barker, and R. Alhajj,
Replica placement in data grid:
Considering utility and risk, 2005.
[29] M. Bsoul, A. Al-Khasawneh, E. Eddien
Abdallah, Y. Kilani, Enhanced fast spread
replication strategy for data grid, Journal
of Network and Computer Applications
(2010).
[30] W.H. Bell, D.G. Cameron, A. P. Millar, L.
Capozza, K. Stock-inger, and F. Zini,
Optorsim: A grid simulator for studying
dynamic data replication strategies, in
International Journal of High performance
Computing Applications, vol. 17, no. 4,
pp. 403–416, 2003.
[31] W. Zhao, X. Xu, N. Xiong, and Z. Wang,
A weight-based dynamic replica
replacement strategy in data grids, in
IEEE Asia-Pacific Services Computing
Conference,(APSCC), pp. 1544–1549,
2008.
[32] F.B. Charrada, H. Ounelli, and H.
Chettaoui, An efficient replication
strategy for dynamic data grids, in
International Conference on P2P,
Parallel, Grid, Cloud and Internet
Computing (3PGCIC), pp. 50–54,
2010.

www.ijcat.com 506
[33] W. Zhao, X. Xu, Z. Wang, Y. Zhang,
and S. He, Improve the performance
of data grids by value-based
replication strategy, in Sixth
International Conference on
Semantics Knowledge and Grid
(SKG), pp. 313–316, 2010.
[34] M. Bsoul, A. Al-Khasawneh, E.E.
Abdallah, and Y. Kilani, Enhanced
fast spread replication strategy for
data grid, in Journal of Network and
Computer Applications , vol. 34, no.
2, pp. 575–580, 2011.

A Survey of File Replication Techniques In Grid Systems

More Related Content

What's hot (17)

Viewers also liked (6)

Similar to A Survey of File Replication Techniques In Grid Systems (20)

More from Editor IJCATR (20)

Recently uploaded (20)

A Survey of File Replication Techniques In Grid Systems