Mphil Computer Science Cloud Computing

1. DISSERTATION PROPOSAL : APREFETCHING
TECHNIQUE USING HMM FORWARD CHAINING
FOR THE DFS IN CLOUD
Researcher
Thilaganga.V
Reg No:16255106
Under the Guidance of
Dr.M.KARTHIKA, M.C.S.A.,
M.Phil.,Ph.D.,

2. Research Overview
• Abstract
• Introduction
• Cloud Computing
• Distributed File System (DFS)
• Hidden Markov Model (HMM)
• Forward Algorithm
• DFS with Forward Chain
• Client Server Transaction
• System Architecture
• Implementation
• Conclusion and Future Enhancement

3. Abstract
 In recent applications of Distributed File System for cloud have
managed in providing high security. The Hidden Markov Model
(HMM) has played a vital role in prediction analysis in various areas.
 Hidden Markov Model and Distributed File System (DFS) applications
have been in a variety of problems in cloud. The HMM includes
Evaluation, Decoding and Learning.
 The HMM problems can be used to solve various sequences analysis
problems, such as forward, backward and Viterbi algorithms.
 Cloud computing synchronizes the operation of all systems, with
techniques such as data center networking, the map-reduce framework
with supports data intensive computing applications in parallel and
distributed systems that provide lively resource allocation, permitting
various operating system to exists on the same physical server.
 So Distributed File System provides more reliable services.

4.  The paper represents client and server process, which is a storage
server, can directly pre-fetch the data from client machines. The
prefetching techniques, uses Hidden Markov Model Forward chain
and work Distributed File System for cloud.
 The prefetching techniques can be implemented to analyze the I/O
from the client machine and then send the prefetching data to the
relevant machine proactively.
 Finally, the storage server can forward the prefetching data, in prior
to the request from the client machine.
 Keywords: Distributed File System, Hidden Markov Model,
Storage Server.
Abstract Cont.…

5. Introduction
In this part discuss about an introduction of following as :
 Cloud Computing
 Distributed File System (DFS)
 Hidden Markov Model (HMM)

6. Cloud Computing

7. Cloud Computing Cont.…
cloud computing is the delivery of computing services servers,
storage, databases, networking, software, analytics and more over the
Internet.
Uses of cloud computing
 Create new apps and services
 Store, back up and recover data
 Host websites and blogs
 Stream audio and video
 Deliver software on demand
 Analyze data for patterns and make predictions

8. Distributed File System (DFS)
 In cloud computing application, distributed file system is very core
technology.
 A distributed files system for cloud is a file system that allows many
clients to have access to data and supports operations (create, delete,
modify, read, write) on that data.
 Each data file may be partitioned into several parts called chunks. Each
chunk may be stored on different remote machines, facilitating the
parallel execution of applications.
 Typically, data is stored in files in a hierarchical tree, where the nodes
represent directories.

9. Hidden Markov Model (HMM)
The Hidden Markov Model is a restricted set of states, each one of which
is associated with a (commonly multidimensional) possibility
distribution.
A hidden Markov model can be considered a broad view of a mixture
model someplace the hidden variables or dormant variables, which
control the mixture constituent to be designated for each observation, are
related over a Markov process somewhat than self-determining of each
other.
Types of Algorithms in HMM
 The Viterbi Algorithm
 The Forward Algorithm
 The Backward Algorithm

10. Forward Algorithm
 In this Algorithms are used to training for the data that are
accessed and processed with frequently. So we get trained
data that is used for finding the shortest path with values.
 In this Forward Algorithms are calculating and finding best
path value is equal 0.0315. Which means all the HMM
algorithms brings the same value using matrix format.
 The Forward Algorithm is a recursive algorithm for
calculating the observation sequence of increasing length.

11. Example
X —states
y —possible observations
a —state transition probabilities
b — output probabilities
 In a room that is not visible to an observer there is a genie. The room
contains urns X1, X2, X3, each of which contains a known mix of balls,
each ball labeled y1, y2, y3.
 In a room that is not visible to an observer there is a genie. The room
contains urns X1, X2, X3, each of which contains a known mix of balls,
each ball labeled y1, y2, y3.

12.  In HMM chaining algorithms are finding best solution path with values.
It is Combined and worked with Distributed File System for Cloud and
bring a best solution for the clients and servers.
 Distributed file system allows to multiple clients to access the data and
support operations (create, delete, modify, read, write) on that data. Each
data file may be partitioned into several parts called chunks.
 Each chunk may be stored on different remote machines, facilitating the
parallel execution of applications.
DFS with Forward Chain

13. Master Server
Chunks Handled and Replicas Locations
Client Server Transaction

14. Server Caching
 Server caching is automatic at the server in that the same buffer cache is
used as for all other files on the server.
 To avoid unexpected data loss if the server dies.
Client Caching
 The goal of client caching is to reduce the amount of remote operations.
Transfers of data are done in large chunks.
 As soon as a chunk is received, the client immediately requests the next
8K-byte chunk.
 This is known as read-ahead. This way, by the time we do, it will either be
there or we don’t have to wait too long for it since it is on its way.
Client Server Transaction Cont.…

15. System Architecture

16. Implementation
Client File System:
 Client file system is responsible for collecting the information about the
client file system and applications, as well as piggybacking it onto real I/O
requests.
Storage Server:
 It handles file management and provides the actual file I/O service for
client file systems and performs initiative prefetching. It records the block
access events with the piggybacked information sent by the client file
systems.
Training Data Input:
 The input test data is a one dimensional continuous series which cannot be
used as such unless it is converted into Usable/Classical dataset where
shape based clustering can be done. This series though one dimensional is
dependent on past values and the series is modeled in terms of predictor
and dependent variables.

17. Conclusion
 The proposed, implemented and evaluated in data prefetching Distributed
File System for Cloud, which the client machines can receive relevant data
proactively through by the storage server in a cloud environment.
 The storage servers are capable to analyze and predict the client I/O events
and then they proactively push data to the relevant client machines for
satisfying client’s future applications requests.
 The purpose of forward I/O events and regarding client machine
information’s are piggybacked and then transferred to corresponding
storage server from the client nodes.
 The Hidden Markov Model Froward Algorithm computes probability
much more efficiently than the naïve approach, which very quickly ends
up in combinational explosion. It can provide the probability of a given
emission or observation at each position in the sequence of observations.

18.  The current implementation of proposed data prefetching
process in Distributed File System for cloud using Hidden
Markov Model Froward prediction algorithms for proactively
fetch disk I/O events from the client nodes.
 Using forward chain the data is trained and increases the
learning process.
 The current implementation of proposed initiative prefetching
scheme can classify only two access patterns and support two
corresponding prediction algorithms for predicting future disk
I/O access.
Conclusion Cont.…

19. Future Enhancement
 In this work of implementation is using Hidden Markov Model Forward
Chain for predict the data during process time over the networks.
 In Future enhancement of prediction algorithm using the Viterbi
algorithm is a dynamic programming algorithm for finding the
most likely sequence of hidden states called the Viterbi path that results in
a sequence of observed events, especially in the context of Markov
information sources and hidden Markov models.
 Using Hidden Markov Model Viterbi Chain for predict the data during
process or run time over the network connections.
 Using this Viterbi algorithm should be trained data when it predict or
analyze the data retrieve processing.

20. Screen Shots
Input

21. Output