International Journal of Computer Science, Engineering and Information Technology (IJCSEIT)

International Journal of Computer Science, Engineering and Information Technology (IJCSEIT), Vol.3,No.5,October 2013
DOI : 10.5121/ijcseit.2013.3503 17
SURVEY ON CACHING AND REPLICATION
ALGORITHM FOR CONTENT DISTRIBUTION
IN PEER TO PEER NETWORKS
C.SriguruLakshmi1
, G.sivakumar2
, V.Venkatachalam3
1
PG Scholar, Dept of CSE, Erode Sengunthar Engg College, India
2
Assistant Professor-SG, Dept of CSE, Erode Sengunthar Engg College, India
3
Principal, Dept of CSE, The Kavery Engg College, India
Abstract
In this paper, we focuses on caching and replication algorithm for content distribution in peer to peer
networks. Caching and replication is a key strategy for improve the reliability, availability, and
performance in peer to peer networks. This paper gives a brief introduction to caching, replication and
various algorithms have been discussed and a detailed study has been performed. The comparison table
shows it clearly that an algorithm satisfies the caching and replication requirement.
Key words
Caching, Replication, P2P networks, Comparison of algorithm
1. INTRODUCTION
P2P systems are classified into two types namely: 1). Centralized P2P system and 2).
Decentralized P2P system. The paper mainly deals with decentralized P2P system. Decentralized
P2P system can be classified into decentralized structured and unstructured system. In
Decentralized Structured P2P architecture network topology is tightly controlled, whereas
Decentralized Unstructured P2P systems do not have any control over the network topology, and
placement files over the network. The most general P2P system is the decentralized unstructured
system .Peers form a network among them on top of the existing inter infrastructure, which is
known as the Overlay Network. The challenges facing in P2P systems are scalability, reliability,
access latency, network traffic, fault tolerance, bandwidth utilization, security, and load
balancing. Data replication and caching techniques are the important two services in peer to peer
networks. It increases data availability by creating local or nearly available copies of popularly
used items, by forwarding each query to its nearest copy; the query search latency can be
effectively reduced. It also reduces communication overhead, increased system performance,
achieves fault-tolerance, and enhances reliability and load balancing.[17]. 1)Reliability: In p2p
results in the elimination of the single point of failure that has dogged all timesharing systems. It
also supports replication for all of its network services. If one of the servers becomes unavailable,
a client automatically switches over to one of the replicated servers. 2)Availability: The wining

18
feature of content distribution components are available to users at all times. Use of replication
enables an administrator to do file system backups while the system is up and running. The
replicated copy remains stable even while the user is changing the original file. 3)Fault
Transparency or Tolerance: Components in a distributed system can fail independently. A file
may be made more available in the face of failures of a file server if it appears on more than one
file server. Availability is one of the main measures of the advantage of a replication algorithm.
4)Load Balancing : Replication of files in a p2p networks allows better load balancing.
5)Performance: Another advantage of a p2p system is the ability to share information with many
diverse users. DFS is an efficient, extensible system. In addition, the server tracks which clients
have cached copies of files, reducing the need for the client to constantly query the server, as well
as reducing the network and server load.
2. CACHING TECHNIQUE
Caching is processed on the media chunks. The caching is done in media chunks in two ways.
The first way is to cache the chunks in a predefined manner when no data regarding the
popularity of the media is available. The second way caching is work working based on the
popularity of the video. For example, when a server receives a request from a client, it
immediately responds to the client if the object is in its local cache storage space. In cache storage
space storing frequently used videos. Cache using various algorithms to store the video contents.
So Caching increase the searching speed, hit rate and reduces the client waiting time and
communication among peers.
3. REPLICATION TECHNIQUES
Replication is the process of creation and maintenance of duplicate copies of objects in internet-
scale distributed system. Replication improves the system performance, fault tolerance, reduces a
network bandwidth usage and increases the availability of popular data objects by distributing the
source of information in globally. To efficiently use the server storage we need to replicate
objects that will yield the best performance. Replication is needed in the case of System failure,
network traffic and to increase system scalability, load balancing, and to reduce access cost. For
example, users can access a local object rather than origin server to minimize network traffic,
access latency and provide location transparency.[17] There are different models of object
replication. We mainly deal with a distributed replication group. A distributed replication group
contains several servers dedicating some storage for the replicas. A server has to serve requests
from its clients and also from other servers in the group. When a server receives a request from a
client, it immediately responds to the client if the object is in its local storage. Otherwise, the
object is fetched from other servers within the group at a higher access cost or from the origin
server, at an even higher cost; in the case no server within the group stores a replica of the object.
The purpose of the replication group is to achieve minimum access cost.
4. RESEARCH ISSUES IN CACHING AND REPLICATION
TECHNIQUES
In this paper we have analyzed only some methods of caching and replication algorithms in p2p
networks as follows:

19
A. Cache Optimization On Hot-Point Proxy Caching Using Weighted-Rank Cache
Replacement Policy(WRCP)
Existing Approach the HP-Proxy model using Dual Cache Replacement Policy(DCRP). DCRP
function, only calculate the access frequency percentage and other costs to calculate the weighted
cost(WC). Based on the WC, the three-level cache replacement is performed, and even the parts
of the objects that have not been accessed recently or are very old are kept. The DCRP does not
consider the age or access gap of the objects currently available in the proxy server.In Proxy
Caching schemes to provide better media content delivery. It store the initial part or interleaved
parts of video objects into a proxy server’s cache. If the cache is done, first access by a client
from the media server. Not possible to cache an entire object in proxy. The media objects need to
divided into smaller chunks. The cache is processed on the media chunks.[3]
Weighted-Rank Cache Replacement Policy(WRCP):
• Access Frequency of object – no.of times accessed compared with all available
objects.
• Mean age of the object - mean age of object compared with all available objects.
• Mean access Gap ratio of the object - it calculate at every mean arrival time.
• Other cost functions- It says fetching latency and size of the object.
Weighted Rank Calculation (WR): The WR is used to determine object to the users. The objects
with a high ranking are kept in the cache without replacement. Each object with a medium or low
ranking is selected as a victim object and replaced with either an entire object or a partial object
from the cache. The WR of an object is assigned based on the calculated WC value of the object.
B. Design and Analysis of Streaming For P2P VOD with MDC
P2P video streaming system reduced server load compared to traditional client server
system. More and more video on demand system using p2p service. The user requirement is
satisfied by another user. So it reduced the server loading. Existing System using Cache and
Relay Scheme. Each peer cache the most recent video segment in its buffer and then relay on late
joining peers. This scheme depends on buffer size. User require service to the server, it increase
the server load.MDC: (Multiple Description Coding). It uses MDC. It divided the video file into
segments, divided those segments again by using MDC that encoded the several description. It
solve the package losing problem, and allowed users with different bandwidth can all watch the
video.It satisfying user’s QOS(Quality Of Service). Even if the pocket loss can still view the
lower video quality. And it reduced high churn rate.[1]Video locality Based Buffering: video
divided into segments. Classified each segments into two classes. Leading, non leading. Leading
segment store the popular videos. And non leading segment store some consecutive videos. When
the buffer is full, it implements LRU(least recently used) mechanism that release non popular
video segments. Balanced Dynamic Buffering: It using inter group balancing and intra group
balancing. Intra group balancing determines with the least aggregate bandwidth. Inter group
having most aggregate bandwidth. When the peer buffer is full, it drops the cached description
with most required bandwidth. Dynamic buffering achieves a more efficient utilization of peer
resources. Streaming Mechanism: MDC provides the complete process. It defines i) from peer
joining the system to leaving after viewing the whole video , ii)after peer joining the system,

20
peers are grouped by server to received leading description number. Caching Mechanism: It
caches the frequently watching videos. It gets leading segment description of group from peer
zof other group until quality of the segment is satisfied to continue caching the next segment.
When the buffer is full, buffer needs to be adjusted by buffering mechanism. Buffering
Mechanism: This mechanism using non-leading, secondary, leading segment. When the buffer is
full, the first released non leading buffer. It using FIFO mechanism to released non-leading
segment. But buffer is still full and can’t receive the subsequent segment description. After it
released secondary segment. It using MRU(Most Recently Used) mechanism to released
secondary segment.
C. Design of Distributed Prefetching Protocol in Push to Peer Video on Demand
System(FSS & CBS)
In p2p streaming system, the upstream bandwidth of peers are larger than video playback rate,
this p2p system does not overcome the upstream bandwidth limitation. In Existing proxy prefix
caching technique require upstream bandwidth of peer larger than video play back rate. So it goes
for push to peer system. In push to peer system does not relay on content server. So this system
can overcome the bandwidth limitation. It using content placement and associated pull policies.
In content placement increase the content availability and improve the client bandwidth
utilization. The mostly wanted videos are proactively fetched to the set of boxes in the subscriber
during time of low network utilization(eg: Early morning). The client can easily download and
playing the videos in set top box. So it reduced server load and client waiting time.It using 2
approaches: Full Striping scheme, Code based placement scheme.[2] Push-to-Peer System: It
willing to have content proactively push to them before video distribution among the cooperating
peers begins.Full Striping Scheme: Videos are strip into video blocks and push the different video
block into the set top box. This system provide high quality video streaming and it reduce the
client waiting time, network congestion. In this scheme stripes each window of overall movie M
boxes. Every window divided M blocks. Each of size is W/M. Each block pushed to only one
box. Each box stored different blocks of window. Code Based Placement Scheme: Videos are
encoded into coded symbol, by using the rate less code algorithm. This algorithm eliminate the
box of failure in full striping scheme. Rate less code algorithm generate an infinite number of
coded symbol. It fixes the rate of each video content. To assign the symbol of each video content.
This system easy to find the most probable video. Distributed Prefetching Protocol: When the
client’s required video not available in the set top box, then the distributed prefetching protocol
used to directly connect client to the video server and streaming the video server to client.
D. Popularity Aware Limited Caching for Reliable on Demand P2P video
Streaming(PALC)
The peer to peer network as a good environment for resource sharing over the Internet. Compared
with traditional file sharing provokes a significant amount of today internet traffic . It has
insufficient memory, high bandwidth utilization for large scale media objects, and lack of
cooperation between proxies and their clients. Very challenging task to achieve efficient content
delivery under the increased availability of continues media streaming.[5]Existing System:
Proxy based caching, Segment Based-proxy caching Proposed System: Hybrid Proxy caching
Proxy Based Caching: The proxy caching provides significant performance improvement in video
streaming with asynchronous demands and highly localized access interests. proxy caching that
can achieve efficient streaming for small size video content. In large video content, caching a

21
complete media objects is not practically feasible in the case of limited proxy cache space.
Segment-Based Proxy Caching: To handle the large size media objects. It caching the entire
media content, the segment based approach partially caches the segments of media objects. The
segment-based proxy caching technique provides efficient streaming with high scalability.
Reason’s)The small size of the reference locality of video objects leads to a cache space problem
in media proxy systems.ii)The single point failure may occur in proxy based caching system.iii) It
is highly vulnerable to overload.iv) Reservation of proxy bandwidth for continuous
streaming.Hybrid Proxy Caching: An efficient hybrid proxy caching system in order to reduce the
network/ server load. It improving the end user video quality. It also an efficient Popularity
Aware Limited Caching (PALC) algorithm that cooperatively utilizes the memory space of the
proxy servers and clients to achieve efficient and cost effective P2P video streaming. The proxy
maintains the popular media segments for global sharing. In order to maintain the unpopular
segments, a small space is left for each peer in the network. The overall hit ratio is improved due
to the effective and adaptive maintenance of popular and unpopular segments. It uses as follows.
• Video Segmentation
• Distributed hash table(DHT)
• Announcing of video media segments
• Cache Lookup Retrieval Algorithm
E. On Replication Algorithm in P2P VoD(RLB & ARLB)
In p2p VoD contributes some storage to store the videos. Peers have the sufficient bandwidth for
given playback rate.[4] The relationship between the storage capacity at each peer? This storage
capacity based on the number of videos, number of peers, and off-loading of video server
bandwidth. We use simple statistical model to drive the relationship. It using generic replication
algorithm random with load balancing. It balances the both Homogenous and Heterogeneous
peers. Proposed System: Replication algorithm with Random load balancingExisting System:
RLB-Centralized algorithm.RLB-Centralized Algorithm: It is expansive to use a centralized
algorithm, RLB to push movies to peers for storage, It implement Reactive and Distributed
replication algorithm, Replication at each peer is changed only after the peer finished viewing a
movie. Replication decision whether stores the movie locally or discard it. Reactive algorithm is
adapting the system dynamics. (Eg: movie popularity churn, and peer churn).The centralized
algorithm like RLB is static and not very practical in system dynamics.Replication
Algorithm:(RLB): It produce best load balance among the peers, terms of average load and
correlation of load. We use a generic algorithm called Random with Load Balancing. It balances
the average load and reduced the different correlation load. It efficient to utilize the bandwidth. It
provides the scaling properties of what p2p replication can achieve. It servers the thumb rule:
needed peers storage for given number of movies. These peers are Homogenous and
Heterogeneous. It using RLB Distributed and Adaptive Algorithm(ARLB) and how to coverage
to a balanced state given movie popularity. Random Demand Model: Random algorithm only
generates the replication allocation. That balances the load for all movies.Adaptive RLB: Movie
replication viewed movies only. After a peer i viewed a movie j , the peer runs the Adaptive RLB
replication algorithm. It achieves balanced bandwidth for all movies. And it easily achieved by
centralized algorithm in RLB.It achieves the adaptation.

22
F. Exploring the Optimal Replication Strategy in P2P-VoD Systems: Characterization
and Evaluation(ORA)
We address an important open problem: what is the “optimal replication ratio” in a P2P-VoD
system such that peers will receive service from each other and at the same time, reduce the
access to the content server? Issues: 1) what is the optimal replication ratio of a movie if we know
its popularity, and 2) how to achieve these optimal ratios in a distributed and dynamic fashion. In
Existing approach using the proportional replication strategy is “suboptimal”. In Proposed
approach expands the design space to both “passive replacement policy” and “active push policy”
to achieve the optimal replication ratios. It shows how to greatly reduce server’s workload and
improve streaming quality via our distributed algorithms. Two types of Deployment: 1.software
based and 2.set-top box based. 1.Software Based- Active peers 2.Set top box Based- Non active
peers[7]Peer Scheduling Policy: A peer first seeks help from concurrent peers; if the playback
rate cannot be satisfied, the peer seeks help from replication peers; if the playback rate still cannot
be satisfied, the server S will upload the data to this peer so as to satisfy the playback
rate.Suboptimality of the Proportional Replication Strategy:This replicate more unpopular
movies. Optimal ratio of the Optimal Replication in set top box system:It only replicates popular
movies. It achieves optimal ratio.
G. Home-Box-assisted Content Distribution Network for Internet Video-on-Demand
Services(HB)
The paper an adaptive popularity-based video caching strategy among the introduced HB layer
and compares it through simulation to existing CDNs’ one. In Existing approach CDNs, the video
contents are pushed from a Content server to multiple powerful servers.[6] so called surrogates.
In replicas placement, content clustering, client request redirection consider pure CDN. It
consumes high storage and uploads connectivity. In proposed approach introduces a new
equipment in the Home-Box virtual layer, capable of content caching and with streaming
capabilities, permitting through the deployment of an overlay virtual layer to overcome scalability
and deployment cost issues of CDNs.HB Assisted CDN: i).An efficient and adaptive video
content caching that brings video contents more closer to the End-Users permitting thus short
end-to-end distances for more controlled risk of service failures.ii).Infinitely scale services by
relying on the HB storage and uplink bandwidth capabilities.iii)Dynamically managing the user
requests.iv).HB can act as a Content/Service Provider for the contents published by its related
users.v).HBs are organized in an overlay and act both as proxies for their “home serving” users
when they request a service and as a server for supplying other HBs related clients with the video
contents stored in their caches.Popularity Based Video Caching Strategy: HB maintain the
particular video replication strategy. It make available the video contents to the possible users’
demand, reducing delays and network bandwidth consumption.Each participating HB contributes
with persistent disk storage capacity and upload bandwidth. All HBs have the same storage
capacity and upload connectivity. Important to replicate videos in the overlay according to their
popularity. How to better distribute them among the HBs? An efficient distribution, we have
used the K- means algorithm to determine the distance to respect between any two HBs that will
cache the same video content.

23
H. On Incentiving Caching For P2P-VOD Systems(RBS)
Incentive scheme that peers to cache the needed video data. A reward-based incentive scheme for
caching, where the content provider decides the reward price of each video and peers decide what
videos to cache in a distributed manner. We apply a mean field model to characterize the steady-
state of the caches in a large scale P2P-VoD system. We formulate a pricing problem using an
optimization framework and solve the optimal prices that minimize the content provider’s
operational cost. Our contributions are: 1).We develops a stochastic model to characterize the
peers’ caching behaviors, and use the mean field technique to characterize the system state in a
limiting steady state.2).We formulates an optimal pricing problem of the content provider and
derives the optimal reward prices for each video. Our scheme incentivizes the peers to cache
enough replicas of various videos in the system and minimizes the content provider’s operational
cost.[9]. In Existing approach A straight-forward way is using a stochastic model where the
system’s state is determined by all peers’ cache states, such a system can be computationally
expensive. When the number of peers is large, the state space of the system becomes huge. In our
incentive scheme to overcome this difficulty, we model the peers in the system using a mean field
approach, where we are more interested in the fractions of peers in certain states in steady state,
rather than the exact number of peers in the system.Pricing Scheme: The content provider
proposes the incentive scheme in order to reduce its operational cost. We define two kinds of
pricing strategies.1).Conservative pricing problem (CPP): The content provider would like to set
the prices such that the number of cached replicas can satisfy all peers’ demand.2).Strategic
pricing problem(SPP): when increasing operational cost and upload cost of cache storage, content
provider to set lower prices so as to reduce the reward cost and balance the overall utility.
I. Two-Level Result Caching for Web Search Queries on Structured P2P
Networks(TLC)
Each peer maintains a LRU result cache (RCache) used to keep the answers for queries. To
achieve a short-term fair distribution of queries we introduce in each peer a location cache
(LCache) ,which keeps pointers to peers that have already requested the same queries in the very
recent past. These peers share their query answers with newly requesting peers.[8] This process is
fast as these popular queries are usually cached in the first DHT hop. Then requesting peer which
quickly tends to redistribute load among more and more peers. It achieves better load balance,
Significantly smaller communication volume among peers, Larger cache hit ratio.DHT
(Distributed Hash Table): It provides high data availability and persistence through
replication. Every peer associated with unique identifier, which defines the peer’s position in the
structure, and the range of keys it is responsible for. DHT overlays maintain a strong topology.
Every peer assigned a unique nodeID in a space of 128-bits identifiers generated using a
cryptographic hash SHA-1. Leafset: that contains the L numerically closest peers. It handles
replicas to improve fault tolerance in an environment, where peers join and leave the network
without warning. In order to maintain these two sets of peers, a “keep alive” message is used
periodically to detect peer failures. Bubble : grows in relation intensity of flood, until no peer in
the boundary. It observes high.RCache : The RCache is used to store answers to queries in a
normalized format. This cache stores answers of queries for which the peer is responsible, and
answers of queries delivered to users that submitted queries in the peer. Each cache entry has a
state flag indicating whether the respective query answer is being retrieved from some other place
such as a peer or a Web search engine. A timeout flag indicating the time instant from which the
respective query answer is no longer valid. RCache entries are administered with the LRU

24
replacement policy.LCache: The LCache stores data used to support short-term scheduling. Each
entry stores terms of a query, an IP address of a peer that contains (or is to contain soon) the
answer for the query, and an expiration counter used to limit the number of times the entry is used
to route incoming queries with the same terms. These cache entries are also administered with the
LRU replacement policy though their overall size is much smaller than in the RCache as they do
not store query answers.
J. Proportional Replication in Peer-to-Peer Networks(PRA)
In p2p networks, that has number of replicas of each object proportional to the request rate for
this object and analyzes the how many nodes per advantages. It achieves proportional replication
in decentralized system. It using LRU (Least Recently Used) algorithm achieves near
proportional replication and replica distribution achieved LRU very close to optimal. If the
objective is to minimize the network bandwidth used, the download source should be the nearest
replica if multiple replicas of the file are available.Link Distance to the Nearest Replica: To
replicate object based on the distance. A client request object to another client. The receiver of
client in far away distance, so network bandwidth utilization in increased. The replicate objects
on near optimal to reduce bandwidth consumption.Cache Replacement Algorithms for
Proportional Replica Distribution: A peer-to-peer system where the peers had finite storage space
and if space was needed for a newly requested file then a previously obtained file was deleted
using the LRU file replacement policy (except that the last replica of a file is never deleted). The
steady-state distribution of the number of replicas of each file against the file request rate. The
LRU cache replacement policy obtains near linear proportionality except for high request rate
files. Cache management algorithms such as FIFO (First-In, First-Out: replace the oldest file),
LFU (Least Frequently Used: replace the least frequently used file) and Random-Delete
(randomly select the file to be replaced). All these algorithms generate a replica distribution
similar to LRU: LFU is closer to the optimal distribution than LRU while FIFO and Random-
Delete are slightly further from it than LRU. LRU: cache replacement algorithms such as LRU
are able to achieve near-proportional distribution. Since the user access patterns may change over
time. The time taken by LRU to converge to the new steady-state replica distribution after a
change in the user access pattern was found to be very close to the minimum required by any
cache replacement algorithm. LRU cache replacement algorithm is a very attractive in network
system.
K. RESTREAM – A Replication Algorithm for Reliable and Scalable Multimedia
Streaming
In this paper, an adaptive replication algorithm that relies on replication to achieve reliable and
scalable streaming in resource constrained system. It dynamically adapts replica replacement to
maximize the number of consumer and reduce latency. Restream supports partitioning method. It
performs 1). Serving Consumers 2). Adapting Replication Scheme. In serving consumer, when
new consumer enter in the system replication joining on that consumer and leaving in other one.
It informs that out of our replication. Adapting Replication, it moves from one state to another
state, replication will be growing or shrinking .[10]Node or Link Failure: If node failure case, the
failure node eliminate, reconnect with another node. If link failure case, it will not reconnect in
structure of current tree immediately. But may be used later. Dynamically Adapting Replication
Scheme maximize the number of consumers being concurrent served, without exceeding
bandwidth capacity and latency. In partitioning helps to recover the failure and achieve good
performance in dynamic changing model.

25
L. QoS-Aware Intelligent Replica Management Architecture for Content Distribution
in P2P Overlay Networks (QIRMA)
In this paper, using Intelligent Replica placement algorithm, which the requested contents are
classified into class I and class II. In class I most frequently accessed contents are replicated in
strong cluster which is having high weight values and in number copies. In class II least
frequently accessed contents are replicated in weak cluster which is having low weight values and
in less number of copies.Routing Technique: Routing is performed hierarchically by broadcasting
the query only to the strong clusters. This method has a caching technique, to reduce the search
latency. Using this method System scalability can be improved by distributing the load across
multiple servers which is proposed by replication. The demanded contents can be brought much
closer to the clients through replication techniques, thus reducing both the access latency and
network traffic and increase the system performance by avoiding hot peers which becomes a
bottleneck. Limitation: The limitation in this method is that in strong clusters a few replicated
items have more number of copies. So, this consumes more cost, memory space, bandwidth and
redundancy. In weak clusters a few replicated items have less number of copies. Increased user
response time and performance are the bottleneck of the system.[12]
M. Distributed Algorithm for the Replica Placement Problem (DARPP)
Caching and Replication of most popular data objects results in reduction of network bandwidth
usage and the access latency by designing a distributed approximation algorithm. The main focus
is to improve the efficiency of popular object replication within a distributed replication
group.[15] the group may consist of several servers that allocate certain amount of memory to
place the replicated object which is requested by their clients. For example: When a server
receives a request from its client, it response if the object is in its local storage otherwise it
fetches from other replication servers within the group or from the origin server. By analyzing the
request rates for objects and the server capacities the replicas are placed by considering the
availability of server capacities.Performance: Replication of popular data objects at a server
closer to the user reduces the access time and network bandwidth. Thus it increases the overall
system performance with all objects. By comparing with centralized algorithm this shows only
one percentage degradation of system performance and not decided how many copies of objects
should be replicated in the group.
N. A Framework for Lazy Replication in P2P VoD
Lazy replication is postpones replication, trying to make efficient use of bandwidth. Lazy
replication can decrease server load. Two predicators are plugged to create a lazy replication
algorithm. This replication is more efficient than eager replication. It reduces departure misses
and exist misses if chunks are replicated to a peer with free cache space. Lazy Replication: It
uses of two sub algorithms. 1). A peer departure predictor 2). A chunk request predictor. The peer
departure predictor predicts if a peer will depart in the next time interval. The chunk request
predictor predicts the future popularity of a chunk. The replication algorithm decides which
chunks to replicate, to which peers, and when, given uncertainty about the later arrival request. It
is impossible to know exactly when peers will depart and what chunks will be requested. The
predictors approximate this knowledge. Lazy replication uses a peer departure predictor to choose
when to replicate (just before a peer leaves) and to whom (peers that are predicted to be most
unlikely to leave) and uses a chunk request predictor to choose what to replicate (the chunks that

26
will be the most popular in the coming sessions). [16]Lazy replication uses the same greedy
approach. Replicate the chunk with the highest popularity to the peer with the lowest probability
of departing. The two diﬀer in when a chunk is replicated. Lazy replication uses the peer
departure predictor to decide both when to replicate and to where. Eager replication is more
aggressive than lazy replication because it immediately replicates whenever a peer fetches a
chunk.
O. An Efficient Caching Scheme and Consistency Maintenance in Hybrid P2P
System(ACMA & TC)
P2P networks can be divided into two categories: 1). structured peer-to-peer networks in which
peers are connected by a regular topology, and 2). unstructured peer-to-peer networks in which
the topology is arbitrary. The objective is to design a hybrid peer-to-peer system for distributed
data sharing. Consistency maintenance is propagating the updates from a primary file to its
replica. Adaptive consistency maintenance algorithm (ACMA) maintains that periodically polls
the file owner to update the file due to minimum number of replicas consistency overhead is very
low. Top Caching (TC) algorithm helps to boost the system performance and to build a fully
distributed cache for most popular information. Caching scheme can reduce query delay, better
load balance and higher cache hit ratios. It utilizes both the efficiency of the structured peer-to-
peer network and the flexibility of the unstructured peer-to-peer network, and good balance
between the efficiency and flexibility.[13] To maintain consistency, using file consistency
algorithm for hybrid P2P system so that periodically the file owner to update the file due to
number of replicas consistency overhead is very low. It boosts the performance of hybrid P2P,
Top Caching (TCS) algorithm is used to build a fully distributed cache for popular information in
P2P systems. It effectively relieves the over- caching problems for the most popular objects.
Table : Comparison of Various Caching and Replication Algorithms
Algorit
hm
(or)
Techni
ques
Advantage Result Improved
Parameters
Limitation
WRCR
[3]
It reduces cache
space
Min of 31 object
& max of 177
object
Network bandwidth
Delay
MDC
&
MRU[1
]
Reduce package
losing problem
& server load
Maintain
leading object
only
Uses for high churn
rate
No. of Video
watching rate is
limited
FSS &
CBS[2]
Reduces Client
waiting time,
and consistency
FSS more
consistency than
CBS
Availability of
object
Missing contents on
server
PALC[
5]
Achieves high
hit ratio, increase
cache contents
Achieve 95% of
efficiency,
cache hit ratio
90%
Cache capacity will
improve
Only maintain
popular videos

27
ARLB[
4]
It adapt all the
video object in
sufficient
bandwidth
Adapt
Homogeneous
,heterogeneous
Peers
dynamic
changing movie
popularity
Computational load
will be increased
ORA[7
]
Accurate to find
popular objects
Pop=81%,
unpop=10%
Improved optimality Suboptimal in
software based
HB[6] Calculate
distance of end-
user to replicate
same video
Saves 60% of
server load
Time complexity Delay
RBS[9] High cost
reduction
Cu=0.2,1,2
2=high price
Access Latency Dynamically adjust
price for changing
movie popularity
TLC[8] Better load
balance,
High hit ratio
Replication
reduced 96%,
10% of
improved cache
hit compare
other
Waiting time Delay is increased
PRA[1
0]
It replicate
object based on
user access
pattern
Iteration
increased to 20
to 50, maintain
steady state
Network bandwidth Can't adapt
dynamic user
request
RESTR
EAM[1
2]
Good
performance and
recover link and
node failure
Probability by
sending req
leave message
0.5
Improve access rate Bandwidth
consumption
LAZY
Replica
[16]
Decrease server
load
Decrease server
load at 15%
Throughput and
latency
Cost
ACMA
&
TC[13]
Reduce delay,
better load
balance, high hit
ratio
Decrease join
latency 2.5 only
other LFR 7.0,
LRU 7.8
Speed Network traffic
DARP
P[15]
Minimizes
Response Time,
Cost & Increase
Availability
Only 1%
degradation of
system
performance
Cost, access latency,
Network bandwidth
Not provide
individual system
performance, &
number of copies
QIRM
A[14]
Access latency,
Network traffic,
Scalability, Fault
Tolerance,
Search Latency
bandwidth
utilization of
QIRM is nearly
80-90%, when
compared to
VIRAT, which
is 60-70%.
Throughput, Delay,
Bandwidth, Query
Efficiency,
High availability,
Redundancy, Low
Availability,
Bottleneck
Problem,

28
5. COMPARISON OF PERFORMANCE EVALUATION
Popularity Aware Limited Caching(PALC)
Figure: 1 Cache hit ratio(PALC)
Popularity Aware Limited Caching Obtained upto 90% of Cache hit Ratio
Two Level Result Caching:(TLC)
Figure2: Cache hit ratio(TLC)
In TLC improved Cache hit ratio 10% extra increased for previous one
Home Box Assisted Content Distribution Networks
Figure3: Cache hit ratio (HB)
Home Box Assisted Content Distribution Networks Increase Cache hit at 4%
0
50
100
Cachehitratio%
Cached Media Content(in MBYTES)
PROP
PALC
COPACC
0
200
400
0 20 40 60 80 100
CacheHtiRatio(%)
Simulation Progress
Two Level Result Cache
TLC
L+L
B+L
Leafset
0
0.2
0.4
0.6
100
200
300
400
500
600
700
800
900
1000
1100
1200
1300
1400
1500
CacheHitRatio(%)
Time (S)
Hit Ratio

29
In above all caching hit ratio improved is better for Popularity Aware Limited Caching (PALC).
It improved up to 90% cache hit ratio.
Replication Results
Figure4: proportional Replication
Figure5: Optimal Replication
In proportional and Optimal Replication algorithm, optimal replication distributes more popular
videos to consumers. So optimal replication is better than Proportional replication.
Average of Added Additional Objects: WRCR Vs DCRP
Figure6: Minimum percentage of more object allowed
0
5
0 50 100 150 200 250
Replica
Distribution
Movie ID
Proportional Replication
Proportional
Replication
0
10
0 50 100 150 200 250
Replica
Distribution
Movie ID
0ptimal Replication
0ptimal Replication
0
5
10
0 1 2 3
Minimum
PercentageOfMore
Object(%)
Replacement Level
Cache Policy
WRCP
DCRP

30
Figure7: Maximum percentage of more object allowed
Compared with DCRP and WRCP, WRCP is the added additional objects. Minimum percentage
of adding 31 objects and maximum percentage of adding 177 objects.
Average Bandwidth Utilization: (Home-Box)
Figure8: HB Bandwidth Utilization
Home box assisted content distribution networks, compared CDN with HB-support decrease
server load(10%) than Pure CDN server load(50%)
On Replication Algorithm
Figure9: Server Load
RLB(Random Load Balancing) balances the server load(50%), to use sufficient bandwidth for
objects replication.
0
20
40
0 0.5 1 1.5 2 2.5 3
Maximum
Percetanageofmore
objectallowed(%)
Replacement Level
Cache Policy
WRCP
DCRP
0
1
2
3
0
100
200
300
400
500
600
700
800
900
1000
1100
1200
1300
1400
1500
NormalizedCDN
ServerLoad
Time (s)
Average Server Bandwidth
pure CDN
CDN with HB-support
0
200
400
0 8 16 24 32 40 48 56 64 72 80
AverageServer
Load
No .of users in the System
Average bandwidth
RLB
Lower bound
Upper bound

31
ACMA and TC: Join Latency Comparison
Figure10: Latency
Compared with Hybrid LRU, TC, LFR, TC & LFR increase latency, LRU lower query delay,
better load balance.
6.CONCLUSIONS
This paper conducts a theoretical analysis study on caching and replication strategies in peer to
peer network. A brief discussion of those techniques is summarized. The advantages and
limitations of caching and replication techniques are summarized with reference to various issues
related to caching and replication techniques in content distribution of peer to peer networks.
7. REFERENCES
[1] Chung- yan chen and yu-wei chen, “Design And Analysis of Streaming for P2P VOD with MDC”,
Business and Information, Vol.3, No.1. (2013)
[2] R.Thiruselvan,”Design of Distributed Prefetching Protocol in Push to Peer Video on Demand
System”, International Journal of advanced Computer Technology, Vol.2, No.6. (2013)
[3] S.P.Ponnusamy and E.Karthikeyan,“Optimization Cache on Hot-Point Proxy Caching Using
Weighted- Rank Cache Replacement Policy”,ETRI Journel, Vol.35, No.4. (2013)
[4] Yipeng Zhou, Tom Z.J.Fu, and Dah Minh Chiu, “On Replication Algorithm in P2P VoD”,
IEEE/ACM Transaction on Networking, Vol.21, No.1. (2013)
[5] T.Suresh and K. venkatachalapathy, “Popularity Aware Limited Caching for Reliable On Demand
P2P Video Streaming”,International Journal of Computer Application, Vol.58, No.13. (2012)
[6] Soraya Ait Chellouche, Daniel Negru, Yiping Chen,“Home Box-assisted Content Distribution
Network for Internet Video On Demand Services”, ISCC/ ICT-ALICANTE. (2012)
[7] Weijie Wu, John C.S, Lui, “Exploring the Optimal Replication Strategy in P2P-VoD Systems:
Characterization and Evaluation”, IEEE/ACM Transaction on Parallel and Distributed System,
Vol.23, No.8. (2012)
[8] Erika Rosas, Nicolas Hidalgo and Mauricio Marin, “Two Level Results Caching for Web Search
Quaries On Structured P2P Networks”, IEEE International Conference on Parallel and Distributed
System. (2012)
[9] Weijie Wu, John C.S. Lui, Richard T.B, “On Incentiving Caching for P2P VoD Systems”, IEEE-Part
of INFOCOM, 978-1-4673-1017-8/12. ,(2012)
[10] Saurabh Tewari and Leonard Kleinrock “Proportional Replication in Peer to Peer Networks”, Journal
Of High Performance Computing And Networking. ,(2012)
0
10
20
2 3 4 5 6 7 8 9 10
AverageLookup
Latency
Peers
Latency Comparison
Hybrid p2p-LRU
Hybrid p2p-TC
Hybrid p2p-LFR

32
[11] Nikhil Chaturvedi and Dinesh Chandra Jain “Analysis of Replication and Replication Algorithms in
Distributed System”, International Journal of Advanced Research in Computer Science and Software
Engineering. Vol. 2, No.5. (2012)
[12] Shabnam Ataee, Benoit Garbinato, Fernando Pedonet ,“RESTREAM- A Replication Algorithm for
Reliable and Scalable Multimedia Streaming”, Parallel and Distribution System. (2013)
[13] E.Kalaivani, J. Selva Kumar, “An Efficient Caching Scheme and Consistency Maintenance in Hybrid
P2P System”, International Journal of Computer Networks and Wireless Communication, Vol.2.
No.1. (2012)
[14] S. Ayyasamy and S.N. Sivanandam, “ A QOS-Aware Intelligent Replica Management Architecture
for Content Distribution in Peer-to-Peer Overlay Networks “, International Journal on Computer
Science and Engineering, Vol.1, No.2. (2009)
[15] Sharrukh Zaman and Daniel Grosu, “A Distributed Algorithm for the Replica Placement Problem”,
IEEE Transaction on Parallel and Distributed Systems. Vol. 22, No.9. (2011)
[16] Bin Cheng, Lex Stein, Hai Jin,“A Framework for Lazy Replication in P2P VoD”, ACM 978-1-60588-
157-6/05(2008)
[17] Dr.Anna Saro Vijendran1 and S.Thavamani, “Analysis Study on Caching and replica placement
Algorithm for Content Distribution in Distributed Computing Networks “, International Journal of
peer to peer network, Vol.3, No. 6. (2012)

International Journal of Computer Science, Engineering and Information Technology (IJCSEIT)

More Related Content

What's hot (18)

Similar to International Journal of Computer Science, Engineering and Information Technology (IJCSEIT) (20)

Recently uploaded (20)

International Journal of Computer Science, Engineering and Information Technology (IJCSEIT)