Consistency of data replication

International Journal on Information Theory (IJIT),Vol.3, No.4, October 2014
CONSISTENCY OF DATA REPLICATION
PROTOCOLS IN DATABASE SYSTEMS: A
REVIEW
Alireza Souri1, Saeid Pashazadeh*2 and Ahmad Habibizad Navin3
1, 3Department of Computer Engineering, Tabriz Branch, Islamic Azad University,
Tabriz, Iran.
2* Faculty of Electrical and Computer Engineering, University of Tabriz, Tabriz, Iran.
Abstract:
In this paper, a review for consistency of data replication protocols has been investigated. A brief
deliberation about consistency models in data replication is shown. Also we debate on propagation
techniques such as eager and lazy propagation. Differences of replication protocols from consistency view
point are studied. Also the advantages and disadvantages of the replication protocols are shown. We
advent into essential technical details and positive comparisons, in order to determine their respective
contributions as well as restrictions are made. Finally, some literature research strategies in replication
and consistency techniques are reviewed.
Keywords:
Database system, consistency, data replication, update propagation.
1. Introduction
Consistency of Replication models is essential to abstract away execution particulars, and to
classify the functionality of a given system. Also a consistency model is a method for come to a
joint considerate of each other’s rights and responsibilities.
Database system attracts lots of consideration. A large-scale database storage system [1,2] is
among the fundamental conveniences in the cloud, unstructured peer-to-peer (P2P) networks [3],
grid environment [4] or in similar systems. The system with large-scale database system typically
assigns computing replicas near their input data[5]. A good data management develops very
important conditions in such a scenario. Data in a distributed database system [6] is replicated for
increasing reliability, availability and performance. There are two mechanisms for locations of
data replicas such as static and dynamic replicated system[7,8], which regulates replica locations
based on session information of requests [9].
DOI : 10.5121/ijit.2014.3402 19

On the one hand, in these consistency models, its performance suitability for data replication
architecture are not specified exactly. On the other hand, a consistency model dos not guarantees
the high performance and high scalability for a data replication mechanism [10].
Consistency, accessibility, scalability, security, fault tolerant and performance [11] are areas for
system implementation[12]. High accessibility and performance are basics for such a system with
large-scale distributed database system. We have to make a tradeoff between consistency and
replication. There are dissimilar levels of weak and strong consistency. A distributed database
system may deliver levels of consistency weaker than one-copy-serializability but stronger than
eventual consistency. Also there are levels of consistency such as data-centric and client-centric
models for a data replication mechanism. So the recognizing usage of consistency models in each
data replication mechanism is necessary. In Section 2, we introduce consistency models in client
view and server view, adapted from the theory of database concurrency control. Then, we depict
on consistency protocols in Section 3. With these discussions, we can represent a comparison
among eventual consistency and client-centric consistency models. Properties of their
implementations can also be deduced accordingly. In section 4, we discuss replication models and
propagation techniques. We show all of the replication protocols according to the update
propagation and replication mechanisms in Section 5. Section 6 shows a classified review for
replication and consistency techniques in some research strategies. Section 7 is the conclusion
and describes future work finally.
20
2. Consistency models
In this section, a series of consistency models are considered. We discuss about differences of
consistency models. Variant methods to categorizing the consistency models can be originated
from [13] and [14].One of the important properties of a system design is consistency model. This
property can typically offered in relations of a state that can be true or false for different
implementations. Consistency models are referred to as the contracts between process and data
for ensuring correctness of the system. Consistency models are presented through a number of
consistency criteria to be satisfied by assessments of operations [15].For standard consistency
conditions of the ACID properties [16], there exists some methods for consistency guarantee. In
ACID consistency method, database is in a consistent state when a transaction is finished. In the
client level there are four component:
• DS is a storage system.
• PA is the process operation for each read or write by DS.
• PB is sovereign of process PA that performs each read and write operation from the DS.
• PC is sovereign of process PA that performs each read and write operation from the DS.
•
In the client level consistency, it is important that how and when an observer is occurred. The PA,
PB and PC processes see updates with a data item in the storage system. There are two
consistency types such as Data-Centric consistency and Client-Centric consistency[17].
In Data-Centric consistency there are:
• Strict consistency. All of the A, B and C send back the result of update value when the
update procedure is completed.

• Sequential consistency. The level of sequential consistency is lower than strict
consistency. Each read and write operation is performed by all replicas on their data item
x sequentially. Also each discrete procedure operations execute the identified order.
• Causal consistency. This consistency is weaker than strict and sequential consistency[18].
If transaction T1 is influenced or caused on an earlier transaction T2, each replica should
be first see T2, and then see T1.
• FIFO consistency. FIFO consistency is relaxed to implement because it is being
guaranteed two or more writes from a single source must arrive in the order issued.
Basically, this means that with FIFO consistency, all writes generated by different
processes are concurrent.
21
In Client-Centric consistency models there are:
• Eventual consistency. This model guarantees that if a updates are complete to the item
eventually [19], then all accesses on this data item send back the previous updated value
[20].
• Monotonic Reads. In this model if an operation reads the data item x, always each
following read operation on data item x send back same value x or a more recent value.
• Monotonic Writes. In this model if an operation writes on the data item x, always each
following write operation on data item x comes after related write operation on the data
item x.
• Read-your-Write. The result of a write operation on the data item x always will be
realized by a following read operation on x by the same value.
• Write follow read. In this model the effect of a write operation on a data item x following
a previous read operation on data item x by the same value that is guaranteed to take
place on the same or a more recent value of x that was read.
3. Consistency Protocols
In this section, we describe the consistency protocols according to[21]. A consistency protocol
explains as an implementation of a specific consistency model. We track the group of our
conversation on consistency models[22].
3.1 Primary Replica Based Protocol
In this protocol, All write operations to a data item x is attended by one specific replica that called
primary replica. This primary replica [23] is accountable for updating other replicas, the client
just cooperates by this primary replica[16].
Two requirements should be happen for this generous of protocol [24]:
• All read and write operations for updating a data item x should spread and be executed all
replicas at some time.
• These operations should be executed in the same order.

22
3.2 Replicated Write Protocol:
In this protocol, each write operations are sent to each replica to update procedure. There are two
types for replicated write protocols.
3.2.1. Active Replication:
In active replication, each replica contains a concomitant procedure that transports out the update
operations. Unlike other protocols, update operations are normally propagated through the write
operation. This propagation causes the operation is sent to each replica. Also there is required a
total order for all write operations that each replica execute the same order of write commands
[25].
3.2.2. Quorum Based:
This protocol specifies that the clients obtain the authorization of several servers before any
reading or writing a replicated data item x [26]. For example, the write operations only want to be
executed on fragment of all replicas before return to the client. It use elections to avoid write-read
conflict and write-write conflict [27]:
• R is the number of replicas of each data item.
• Rr is number of replicas that a client should contacts by them for reading a value.
• Rw is number of replicas that a client should contacts by them for writing a value.
• For preventing the Write-Write and Write-Read conflicts, Rr + Rw > R and Rw +
Rw > R should be satisfied.
4. Update propagation strategies
Update propagation can be measured in two methods [28]
• The update operations are applied to all replicas as part of the unique contract.
• Each replica is updated by the originating transaction. Update operations send to other
replicas asynchronously as a discrete transaction for each node [29].
There are two update propagation methods: Eager techniques and lazy techniques. Normally the
eager protocols are identified as read-one/write-any (ROWA) protocols. First, they have not
transactional inconsistencies. Second, an update transaction can read a local copy of the data item
x and be sure that a refresh value is read. Consequently, there is no essential to executing a remote
read. Finally, the variations to replicas are completed atomically. When we use to a 2PC
execution, the update speed is restricted and it cause that the response time performance of the
update transaction is low. When one of the copies is inaccessible, the update transaction cannot
terminate meanwhile all the copies updated essentially. Lazy protocol is used to new mechanisms
for guaranteeing strong mutual consistency [30]. These mechanisms may be bright to endure
some inconsistency between the replicas for better performance[31].

In a distributed database system the requests permit to access data from local and remote
databases [32]. Distributed methods spread on the update procedure to the local copy where the
update transaction creates, then the updates are broadcasted to the other replica. If distributed
techniques are attached by eager propagation approaches, then the distributed concurrency control
approaches can sufficiently report the concurrent updates problem [13]. Table 1 is shown the
comparisons of update propagation and propagation techniques [33].
23
Table 1. Comparisons of update propagation techniques
Consistency Updating Performance Failure
Eager
update
Strong
consistency
Up-to-date with
High response time
Not transactional
inconsistency,
Changes are
atomic
Restricted update
speed, transaction
crash and
Lower availability
Lazy update Weak
consistency
Out-of-date problem
and Low response
time
Not fault tolerant,
good
response time
Dirty read problem,
Data inconsistency
and transaction
inversion
Centralized
techniques
-
Up-to-date with
Update without
synchronization
Appropriate for
few master sites
High overload and
bottleneck
problems
Distributed
techniques
-
Up-to-date with
Concurrency control
methods
Highest system
availability
Management
problem, Copies
need to be
synchronized
5. Replication Protocols
[13] presented a categorization for replicas data protocols. This is important that When one of the
update propagation mechanisms such as eager or lazy incomes and who should complete updates
mechanism such as primary copy or update-everywhere. In eager propagation mechanism, the
propagation of updates is contained by the restrictions of a transaction. The client does not
receive the notification of commit message up to necessary duplicates have been updated in the
system. In the Lazy mechanism, the update procedure of a local copy is committed. the update
propagation be accomplished [34]. There is an expensive way for providing response time and
message overhead in consistency of eager mechanism. An optimization for prevent from these
problems is using Lazy replication approach. However, the update procedure is executed
separately, therefore inconsistency conditions might occur. [35]. When the updates are broadcast
to replicas in eager or lazy mechanism, two architectures are needed for updates such as
centralized and distributed. Table 2 shows the four replication mechanisms such as eager
distributed and eager centralized for eager mechanism, lazy distributed and lazy centralized for
lazy mechanism [36].

24
Table 2. Update propagation vs. propagation techniques
Centralized Distributed
Eager
Eager Primary Copy
Eager One Master by Restricted Transparency Update Everywhere
One Master with Full Transparency
Lazy
Lazy Primary Copy
Single Master with Limited Lazy Update Everywhere
Transparency
5.1 Eager Centralized Protocol
In eager centralized protocol, there is a site as a master that navigates the read and writes
operations on a data item (x). This protocol guarantees strong consistency techniques for update
propagation. In update procedure, all updates are applied to a logical data item (x) by using the
perspective of the update transaction. This applying is committed by using the 2 Phase Commit
protocol. So, when the update procedure is completed in its transaction, all copies return the
similar values to the updated data items. The result of this mechanism is one-serializability-replication[
37]. The categories of eager centralized include eager primary copy, single master by
restricted transparency and single master by full transparency. In eager primary copy, any data
item (xi) has a master. One replica specified as the primary copy. In this case, there is no single
master for controlling serializability condition. In the single master by restricted transparency, all
of the updates have been sent to the specified master directly. For a read operation, a read lock
occurred on data item x and the read operation is executed. The result of the operation is returned
to the client. Also for a write operation, a write lock occurred on data item x and this operation is
executed. The result of write operation is returned to the client. In the single master by full
transparency, the replica coordination level has been performed by a router. The router sends the
entire read and writes operations to the master directly. The master executes each operation and
returns the result of execution to the client.
5.2 Eager Distributed Protocol
In eager distributed protocol, first the update applied to the local replica, then the update
procedure is propagated to other replicas. The eager update everywhere is a type of eager
distributed protocol.
5.3 Lazy Centralized Protocol
Lazy centralized protocol is like to eager centralized protocol. In this protocol, first the updates
are applied to a master and then propagated to the clients. The significant alteration is that the
propagation procedure does not take place via the update process. However, after the
commitment of transactions, if a client executes a read operation (x) on its local copy, it may read
a non-refresh data, then data item x may have been updated at the master, nevertheless the update
may not have been propagated to the clients yet. The categories of lazy centralized include lazy
primary copy and single master by restricted transparency. In lazy primary copy, each read and
write operation sends to a master. All of the updating results have been send back to the client. In

single master by restricted transparency, the update procedure is executed to the master directly.
When one update has committed, the new transaction is sent to the clients.
25
5.4 Lazy Distributed Protocol
In Lazy distributed protocol, the update transactions can execute on each replica. Also these
updates are propagated to the other replicas lazily. Lazy update everywhere is a type of lazy
distributed protocol. In this type, each read and write operation are performed on the local copy
and the update transactions commit locally. Comparison of replication protocols about
consistency conditions is shown in Table 3.
Table 3. Comparison of replication protocols
Replication strategies Advantages Disadvantages
Eager Centralized
The coordination do not
needs for Update
transactions, there is no
inconsistencies
Extensive response time,
Local copies are can only be
Read, Only useful with few
updates
Lazy Centralized
The coordination do not
needs for Update
transactions, there is
diminutive response times
Inconsistencies, Local copies
are not refresh
Eager Distributed No inconsistencies
Updates need to be
coordinated, Long response
times
Lazy Distributed
Shortest response times, No
centralized coordination
Inconsistencies, Updates can
be lost
6. Comparison of Consistency and replication classification
In this section, some popular and applicable research strategies of replication and consistency
techniques in database systems are discussed. However, Amjad, et al. [38] presented a survey for
dynamic replication strategies in data grid. But, they just considered replication protocols without
consistency models. We discuss the consistency models and replication methods in each research
approach.
In a distributed system for providing and handling extremely available service via no single point
of failure, Lakshman and Malik [39] proposed a quorum-based protocol. This system replicates
data by using in replicated-write group. Also they present three quorum values for guarantee
eventual consistency model. [40] presented the design of a highly available key-value storage
system (Dynamo) which is supports eventual consistency model via quorum-based protocol hat it
allows for better availability in presence of failure. A new dimension of different cloud providers
(MetaStorage) based on quorum strategy was presented by [41]. They proposed a new
consistency model based on static approach. Dingding, et al. [42] proposed a new I/O model to
achieve a good tread-off between scalability and consistency problems. Their model based on
static replication and guarantee eventual consistency model. A new model based on generic

broadcast was proposed by Pedone and Schiper [43] that support causal consistency model. Also
Aguilera, et al. [44] considered the problem of generic broadcast in asynchronous systems with
crashes and presented a new thrifty generic broadcast based on dynamic replication approach that
support causal consistency model. Sousa, et al. [45] proposed a technique for native clocks and
the constancy of network suspensions to decrease the faults in the ordering of cautious deliveries
in wide area networks. They present their model based on static replication approach that
guarantee strong consistency model. An algorithm that handles replication efficiently in active
replication was presented by [46]. Their algorithm is based on static replication approach that
focused on strong consistency model. They could not manage their algorithm when the rollback
problem is occurred. Xinfeng [47] presented a middleware for using a timestamp-based protocol
to maintain the replica consistency. Their algorithm is based on static replication approach that
focused on strong consistency model for improving scalability problem.A static distributed data
replication mechanism of cloud in Google file system was proposed through[48]. They
considered some features when creating conclusions on replicas of data: 1-insertion the new
replicas on mass servers by choosing lower-average disk space consumption, 2-limiting the sum
of replica establishments on each mass server and 3- spreading replicas of a mass crossways
stand. Their algorithm is based on static replication approach that support eventual consistency
model.
Wenhao, et al. [49] proposed a novel cost-effective dynamic data replication strategy named CIR
in cloud data centers. They applied an incremental replication approach to minimizing the number
of replicas while meeting the reliability condition in order to facilitate the cost-effective data
replication management goal. Their approach could reduce the data storage cost substantially,
especially when the data are only stored for a short duration or have a lower reliability
requirement. Also their strategy is based on dynamic replication approach that support causal
consistency model.
Qingsong, et al. [50] proposed a dynamic distributed cloud data replication algorithm CDRM to
capture the relationship between availability and replica number. They focused on dynamic
replication approach that supports a causal consistency model. Ranganathan and Foster [51]
presented six different replication strategies for three different access patterns: Best Client,
Cascading Replication, No Replication or Caching, Plain Caching, Caching plus Cascading
Replication, and Fast Spread. They guarantee the reduction of access latency and bandwidth
consumption based on dynamic replication approach. A centralized data replication algorithm
(CDRA) for Grid sites was presented by [52]. Their algorithm reduced the total file access time
with the consideration of limited storage space of Grid sites. Choi and Youn [53] proposed a
dynamic hybrid protocol (DHP) which effectively combines the grid and tree structure. This
protocol can detect read-write conflict and write-write collision for consistency maintaining.
Their protocol is based on dynamic replication approach that supports an eventual consistency
model.
An evolutionary algorithm to find the optimal replication strategy was proposed by [54]. They
optimized reliability, latency and storage of the system. Because they considered static replication
approach, their protocol did not take total data center energy cost as the primary optimization
target. Lloret, et al. [55] presented a protocol for exchanging information, data, services,
computing and storage resources between all interconnected clouds. Their protocol is based on
26

static replication approach that guarantees an eventual consistency model.Table 4 summarizes the
discussed research strategies and introduces their advantages and disadvantages.
27
Table 4. A collection of research strategies on consistency and replication
Articl
e
Main idea
Consistenc
y method
Replicatio
n scheme
Advantages Disadvantages
[39]
Presenting a distributed
system for handling and
providing highly available
service by no single point
of failure.
Eventual Dynamic
Providing good
scalability and
supports dynamic
control over data
layout and format.
The main
consistency model
is restricted to
eventual
consistency.
[40]
presenting the design of a
highly available key-value
storage system (Dynamo)
Eventual Dynamic
providing a novel
interface for
developers to using
the large e-commerce
operations
The response time
for replicas not
considered
[41]
Presenting a new
dimension of different
cloud providers
(MetaStorage) based on
quorum strategy
Eventual Static
MetaStorage has a
highly available and
scalable distributed
hash table for control
consistency-latency
The strategy can
only guarantee
single consistency
model
[42]
proposing a new I/O
model to reach a good
tread-off between
Scalability and
Consistency
Eventual Static
this new model has
many advantages
over the conventional
asynchronous-synchronous
model
Limiting
consistency
maintenance to
eventual
consistency model
[43]
Ordering the delivery of
messages only if needed,
based on the semantics of
the messages.
Causal Static
Showing better
scalability via
optimizing the
atomic broadcast
protocol with relaxed
causal consistency
Static consistency
model, semantic of
data is difficult to
identify without
knowing the
environment.
[44]
considering the problem of
generic broadcast in
asynchronous systems
with crashes
Causal Dynamic
By defining a
parsimonious
approach for the set
of messages in
generic broadcast
ensures can have
optimal scalability
The availability
has not considered
and the number of
replicas are not
shown
[45]
Proposing a technique for
local clocks and the
stability of network delays
to reduce the mistakes in
the ordering of tentative
deliveries in wide area
networks
Strong Static
Improves scalability
based on the
assumption that data
conflict is rarely
occurring.
Fixed consistency
model, expensive
cost process
[46]
Presenting an algorithm
that handles replication
efficiently in active
replication
Strong Static Improves scalability
Fixed consistency
model
[47]
Presenting a middleware
for using a timestamp-based
protocol to maintain
the replica consistency
Strong Static Improves scalability
Fixed consistency
model, expensive
roll back process.

28
[48]
propose a static distributed
data replication
mechanism in cloud
Eventual Static
insertion the new
replicas on mass
servers by choosing
lower-average disk
space consumption,
Fixed replica
number is used for
all files which may
not be the best
solution for data.
[49]
proposing a novel cost-effective
dynamic data
replication strategy named
CIR in cloud data centers
Causal Dynamic
applies an
incremental
replication approach
to minimize the
number of replicas
and it can reduce the
data storage cost
substantially
their approach is
only based on the
reliability
parameters and
pricing model of
Amazon S3 which
makes it is not
suitable for Google
cluster
[50]
Proposing a dynamic
distributed cloud data
replication algorithm
CDRM to capture the
relationship between
availability and replica
number.
Causal Dynamic
maintains the
minimum replica
number for
a given availability
requirement,
Improves scalability
The scalability
approach is not
proposed
[51]
presenting six different
replication strategies
for three different access
patterns
Eventual Dynamic
Reduction in access
latency and
bandwidth
consumption.
The fixed
consistency model
and limited
number of replica
[52]
presenting a centralized
data replication algorithm
(CDRA) and designing a
distributed caching
algorithm
wherein Grid sites
Eventual Dynamic
reduce the total file
access time with the
consideration
of limited storage
space of Grid sites
The limitation of
the algorithm is
that it considers
only the access
cost.
[53]
proposing a dynamic
hybrid protocol (DHP)
which effectively
combines the grid
and tree structure
Eventual Dynamic
The protocol can
detect read/write
conflict and
write/write collision
for consistency
maintaining.
The grid and tree
structure can only
support read-one/
write-all
mechanism but
hybrid protocol
can have read-all/
write-all
[54]
Presenting an evolutionary
algorithm to find the
optimal replication
strategy
Eventual Static
optimize latency,
storage and
reliability of the
system
This algorithm
cannot take total
data center energy
cost as the primary
optimization
target. Also it
doesn’t take into
account the load
balancing of the
replicas.
[55]
Presenting a protocol for
exchanging information,
data, services, computing
and storage resources
between all interconnected
clouds
Eventual Static
highly scalable and
load balancing
approaches
The resource cost
is not considered
in replicas

Table 5 displays a summarized form of structures of all research strategies studied in above.
These structures include availability, scalability, reliability, response time, bandwidth, load
balancing, number of replicas and storage cost.
29
Table 5. The popular factors of replication and consistency techniques
Article Availability Scalability Reliability
Response
time
Bandwidth
consumption
Load
balancing
Optimal
number
of
replicas
Storage
cost
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
[49]
[50]
[51]
[52]
[53]
[54]
[55]
7. Conclusion
This paper presents a review for data replication protocols in the database systems. Also it
discusses consistency models of replication mechanisms in different update propagations. By
comparing propagation approaches we can use to type of consistency methods for implementing
various data replication mechanisms By notice to comparison of replication protocols, a
consistent replication protocol have important issue in managing and implementing database
systems. In future work, we discuss efficient factors of consistency protocols in distributed
databases that extended in distributed database systems.
References
[1] S. Çokpınar and T. . Gündem, Positive and negative association rule mining on XML data streams
in database as a service concept, Expert Systems with Applications, vol. 39, pp. 7503-7511, 6/15/
2012.
[2] X. Wang, X. Zhou, and S. Wang, Summarizing Large-Scale Database Schema Using Community
Detection, Journal of Computer Science and Technology, vol. 27, pp. 515-526, 2012/01/01 2012.
[3] G. Gao, R. Li, K. Wen, and X. Gu, Proactive replication for rare objects in unstructured peer-to-peer
networks, Journal of Network and Computer Applications, vol. 35, pp. 85-96, 1// 2012.

[4] N. Saadat and A. M. Rahmani, PDDRA: A new pre-fetching based dynamic data replication
30
algorithm in data grids, Future Generation Computer Systems, vol. 28, pp. 666-681, 4// 2012.
[5] A.-H. Wu, Z.-J. Tan, and W. Wang, Annotation Based Query Answer over Inconsistent Database,
Journal of Computer Science and Technology, vol. 25, pp. 469-481, 2010/05/01 2010.
[6] F.-C. Tseng, Mining frequent itemsets in large databases: The hierarchical partitioning approach,
Expert Systems with Applications, vol. 40, pp. 1654-1661, 4// 2013.
[7] N. Mansouri, G. H. Dastghaibyfard, and E. Mansouri, Combination of data replication and
scheduling algorithm for improving data availability in Data Grids, Journal of Network and
Computer Applications, vol. 36, pp. 711-722, 3// 2013.
[8] N. Mansouri and G. H. Dastghaibyfard, A dynamic replica management strategy in data grid,
Journal of Network and Computer Applications, vol. 35, pp. 1297-1303, 7// 2012.
[9] M. Tang, B.-S. Lee, X. Tang, and C.-K. Yeo, The impact of data replication on job scheduling
performance in the Data Grid, Future Generation Computer Systems, vol. 22, pp. 254-268, 2// 2006.
[10] X. Liao, H. Jin, and L. Yu, A novel data replication mechanism in P2P VoD system, Future
Generation Computer Systems, vol. 28, pp. 930-939, 6// 2012.
[11] J. E. Westfall, C. M. Kim, and A. Y. Ma, Locking the virtual filing cabinet: A researcher's guide to
Internet data security, International Journal of Information Management, vol. 32, pp. 419-430, 10//
2012.
[12] R. Garcia, R. Rodrigues, N. Pregui, and #231, Efficient middleware for byzantine fault tolerant
database replication, presented at the Proceedings of the sixth conference on Computer systems,
Salzburg, Austria, 2011.
[13] J. Gray, P. Helland, P. O'Neil, and D. Shasha, The dangers of replication and a solution, SIGMOD
Rec., vol. 25, pp. 173-182, 1996.
[14] M. Wiesmann, F. Pedone, A. Schiper, B. Kemme, and G. Alonso, Understanding replication in
databases and distributed systems, in Distributed Computing Systems, 2000. Proceedings. 20th
International Conference on, 2000, pp. 464-474.
[15] W. He and I.-R. Chen, A proxy-based integrated cache consistency and mobility management
scheme for client–server applications in Mobile IP systems, Journal of Parallel and Distributed
Computing, vol. 69, pp. 559-572, 6// 2009.
[16] A. S. Tanenbaum and R. V. Renesse, Distributed operating systems, ACM Comput. Surv., vol. 17,
pp. 419-470, 1985.
[17] D. B. Terry, A. J. Demers, K. Petersen, M. Spreitzer, M. Theimer, and B. W. Welch, Session
Guarantees for Weakly Consistent Replicated Data, presented at the Proceedings of the Third
International Conference on Parallel and Distributed Information Systems, 1994.
[18] P. W. Hutto and M. Ahamad, Slow memory: weakening consistency to enhance concurrency in
distributed shared memories, in Distributed Computing Systems, 1990. Proceedings., 10th
International Conference on, 1990, pp. 302-309.
[19] Y. Zhu and J. Wang, Client-centric consistency formalization and verification for system with large-scale
distributed data storage, Future Generation Computer Systems, vol. 26, pp. 1180-1188, 10//
2010.
[20] T. Chen, R. Bahsoon, and A.-R. H. Tawil, Scalable service-oriented replication with flexible
consistency guarantee in the cloud, Information Sciences, 2014.
[21] D. Imbs and M. Raynal, Virtual world consistency: A condition for STM systems (with a versatile
protocol with invisible read operations), Theoretical Computer Science, vol. 444, pp. 113-127, 7/27/
2012.
[22] H. Yu and A. Vahdat, Efficient Numerical Error Bounding for Replicated Network Services,
presented at the Proceedings of the 26th International Conference on Very Large Data Bases, 2000.
[23] N. Budhiraja and K. Marzullo, Tradeoffs in implementing primary-backup protocols, presented at
the Proceedings of the 7th IEEE Symposium on Parallel and Distributeed Processing, 1995.
[24] N. Budhiraja, K. Marzullo, F. B. Schneider, and S. Toueg, The primary-backup approach, in
Distributed systems (2nd Ed.), ed: ACM Press/Addison-Wesley Publishing Co., 1993, pp. 199-216.

[25] L. E. T. Rodrigues, H. Fonseca, and P. Verissimo, Totally ordered multicast in large-scale systems,
in Distributed Computing Systems, 1996., Proceedings of the 16th International Conference on, 1996,
pp. 503-510.
[26] D. K. Gifford, Weighted voting for replicated data, presented at the Proceedings of the seventh
31
ACM symposium on Operating systems principles, Pacific Grove, California, USA, 1979.
[27] P. Jalote, Fault tolerance in distributed systems: Prentice-Hall, Inc., 1994.
[28] H. Garcia-Molina, Performance of update algorithms for replicated data in a distributed database,
Stanford University, 1979.
[29] W. Zhou, L. Wang, and W. Jia, An analysis of update ordering in distributed replication systems,
Future Gener. Comput. Syst., vol. 20, pp. 565-590, 2004.
[30] E. A. Boiten and J. Derrick, From ODP viewpoint consistency to Integrated Formal Methods,
Computer Standards Interfaces, vol. 35, pp. 269-276, 3// 2013.
[31] Y. Saito and M. Shapiro, Optimistic replication, ACM Comput. Surv., vol. 37, pp. 42-81, 2005.
[32] S. Goswami and C. Kundu, XML based advanced distributed database: implemented on library
system, International Journal of Information Management, vol. 33, pp. 28-31, 2// 2013.
[33] M. T. Ozsu, Principles of Distributed Database Systems: Prentice Hall Press, 2007.
[34] J. Ma, W. Liu, and T. Glatard, A classification of file placement and replication methods on grids,
Future Generation Computer Systems, vol. 29, pp. 1395-1406, 8// 2013.
[35] B. Kemme, R. J. Peris, and M. Patio-Martnez, Database Replication: Morgan and Claypool
Publishers, 2010.
[36] E. Atsan and Ö. Özkasap, SCALAR: Scalable data lookup and replication protocol for mobile ad hoc
networks, Computer Networks, vol. 57, pp. 3654-3672, 12/9/ 2013.
[37] B. Kemme and G. Alonso, A new approach to developing and implementing eager database
replication protocols, ACM Trans. Database Syst., vol. 25, pp. 333-379, 2000.
[38] T. Amjad, M. Sher, and A. Daud, A survey of dynamic replication strategies for improving data
availability in data grids, Future Generation Computer Systems, vol. 28, pp. 337-349, 2// 2012.
[39] A. Lakshman and P. Malik, Cassandra: a decentralized structured storage system, SIGOPS Oper.
Syst. Rev., vol. 44, pp. 35-40, 2010.
[40] G. DeCandia, D. Hastorun, M. Jampani, G. Kakulapati, A. Lakshman, A. Pilchin, et al., Dynamo:
amazon's highly available key-value store, SIGOPS Oper. Syst. Rev., vol. 41, pp. 205-220, 2007.
[41] D. Bermbach, M. Klems, S. Tai, and M. Menzel, MetaStorage: A Federated Cloud Storage System
to Manage Consistency-Latency Tradeoffs, in Cloud Computing (CLOUD), 2011 IEEE International
Conference on, 2011, pp. 452-459.
[42] L. Dingding, L. Xiaofei, J. Hai, Z. Bingbing, and Z. Qi, A New Disk I/O Model of Virtualized Cloud
Environment, Parallel and Distributed Systems, IEEE Transactions on, vol. 24, pp. 1129-1138, 2013.
[43] F. Pedone and A. Schiper, Generic Broadcast, in Distributed Computing. vol. 1693, P. Jayanti, Ed.,
ed: Springer Berlin Heidelberg, 1999, pp. 94-106.
[44] M. Aguilera, C. Delporte-Gallet, H. Fauconnier, and S. Toueg, Thrifty Generic Broadcast, in
Distributed Computing. vol. 1914, M. Herlihy, Ed., ed: Springer Berlin Heidelberg, 2000, pp. 268-
282.
[45] A. Sousa, J. Pereira, F. Moura, and R. Oliveira, Optimistic total order in wide area networks, in
Reliable Distributed Systems, 2002. Proceedings. 21st IEEE Symposium on, 2002, pp. 190-199.
[46] P. Felber and A. Schiper, Optimistic active replication, in Distributed Computing Systems, 2001.
21st International Conference on., 2001, pp. 333-341.
[47] Y. Xinfeng, Providing Reliable Web Services through Active Replication, in Computer and
Information Science, 2007. ICIS 2007. 6th IEEE/ACIS International Conference on, 2007, pp. 1111-
1116.
[48] S. Ghemawat, H. Gobioff, and S.-T. Leung, The Google file system, SIGOPS Oper. Syst. Rev., vol.
37, pp. 29-43, 2003.
[49] L. Wenhao, Y. Yun, and Y. Dong, A Novel Cost-Effective Dynamic Data Replication Strategy for
Reliability in Cloud Data Centres, in Dependable, Autonomic and Secure Computing (DASC), 2011
IEEE Ninth International Conference on, 2011, pp. 496-502.

[50] W. Qingsong, B. Veeravalli, G. Bozhao, Z. Lingfang, and F. Dan, CDRM: A Cost-Effective
Dynamic Replication Management Scheme for Cloud Storage Cluster, in Cluster Computing
(CLUSTER), 2010 IEEE International Conference on, 2010, pp. 188-196.
[51] K. Ranganathan and I. Foster, Identifying Dynamic Replication Strategies for a High-Performance
Data Grid, in Grid Computing — GRID 2001. vol. 2242, C. Lee, Ed., ed: Springer Berlin
Heidelberg, 2001, pp. 75-86.
[52] D. T. Nukarapu, T. Bin, W. Liqiang, and L. Shiyong, Data Replication in Data Intensive Scientific
Applications with Performance Guarantee, Parallel and Distributed Systems, IEEE Transactions on,
vol. 22, pp. 1299-1306, 2011.
[53] S. Choi and H. Youn, Dynamic hybrid replication effectively combining tree and grid topology,
32
The Journal of Supercomputing, vol. 59, pp. 1289-1311, 2012/03/01 2012.
[54] O. Al-Haj Hassan, L. Ramaswamy, J. Miller, K. Rasheed, and E. R. Canfield, Replication in Overlay
Networks: A Multi-objective Optimization Approach, in Collaborative Computing: Networking,
Applications and Worksharing. vol. 10, E. Bertino and J. D. Joshi, Eds., ed: Springer Berlin
Heidelberg, 2009, pp. 512-528.
[55] J. Lloret, M. Garcia, J. Tomas, and J. J. P. C. Rodrigues, Architecture and protocol for intercloud
communication, Information Sciences, vol. 258, pp. 434-451, 2/10/ 2014.

Consistency of data replication

More Related Content

What's hot (20)

Similar to Consistency of data replication (20)

Recently uploaded (20)

Consistency of data replication