Escolar Documentos
Profissional Documentos
Cultura Documentos
[a]
Asso. Prof., VIT University, Vellore, India
[b]
UG Student, VIT University, Vellore, India
arranged Architecture (SOA) that can enable the client to stable stockpiling. In this model, calculation lost because of a
break these issues into administrations that can be coordinated disappointment can be dealt with similarly as a calculation
to give an answer. Cloud Computing gives the greater part of moved back because of a straggler.In an optimistic scheme, a
its assets as administrations, and makes utilization of the process may fail without logging any of its received messages
settled models and best practices picked up in the area of SOA since its last checkpoint. This implies that, to reduce the cost
to enable worldwide and simple access to cloud benefits in a of accessing stable storage, messages can be logged only when
standardized manner. checkpoints are being written to stable storage. This makes
optimistic schemes well suited for distributed simulation,
B. FAULT TOLERANCE IN CLOUD COMPUTING where message activity is high.
In a distributed system (for eg. a cloud computing II. EXISTING APPROACHES
environment), a crash of a Logical Process makes the whole
process/calculation to stop. The likelihood that one of the A. HADOOP
logical processes will crash amid the simulation increases with
the number or processes participating in the simulation. One of the first mainstream attempts at handling big data
Simply restarting the failed process may leave the simulation was made by Apache’s Hadoop. It was created by Doug
in a conflicting state. Cutting and Mike Cafarella in 2005 [1]. Doug Cutting named
Up until this point, the main response in such a circumstance it after his son’s yellow elephant, hence the logo as well.
has been to restart the whole system. However simply Hadoop is an open source, Java-based programming
restarting the framework is unsatisfactory for simulations that structure that backings the preparing and capacity of to a great
keep running for quite a long time (hours or days together). As degree huge informational indexes in a Cloud Computing
a result, some type of fault tolerance is required to limit the condition. It is a piece of the Apache venture supported by the
erroneous calculation to a minimum[1]. Apache Software Foundation. [1]
A Logical Process may crash because of a bug in the
application code, test system code, or working framework B. HADOOP’S CORE IDEA
code. Even when the entire code is right, the code being keep
Using a single system or server to handle terabytes and
running with a distributed test system may have been written
for a sequential test system. In such cases, it is troublesome to petabytes of data was impractical. It would take a lot of time to
find and adjust the source of the crash. The client of an store, analyse, and retrieve the data. So, what Hadoop, and
application may not be the designer of the code. So, the client Google before it, came up with was to split the big data into
might be not able (or unwilling) to troubleshoot the application small chunks and store it on various individual storage
even when the bug lies completely in the application code. The locations. This would reduce the time needed to work upon the
circumstance will be hopeless, assuming each time the system data.This is similar to the clustering method described earlier
is restarted, a similar bug was to lead to a similar crash[1]. in Section 1.2
Fortunately, experiments with different software systems have
demonstrated that most of the bugs encountered in real
scenarios are transient[1].
At the point when the procedure is restarted, more memory
may be available, subsequently keeping the crash. Accidents
are especially anticipated that would be transient in the
optimistic recreation, where an other message requesting or an
other procedure planning achieves an alternate execution,
likely bypassing the bug that caused the crash in any case.
Hence, restarting the fizzled procedure is a handy decision, if
steps are taken to ensure that the resulting framework state is
predictable. An adaptation to internal failure methodology
Figure 1: Basic Working of Hadoop
ought to likewise have the capacity to endure equipment
disappointments. Equipment disappointments might be as
Figure 1 shows that all of the data being operated upon is
processor breaking down, control disappointment or somebody
stored in the Hadoop Distributed File System
stumbling over the interfacing wires.
orHDFS[Apache HDFS]. The data is first allocated space in
We accept that procedures bomb by basically slamming and the cluster, taken out for performing an operation on them and
they don't convey any incorrect messages or do some other are then returned to the cluster along with a suitable output.
mischief. A procedure loses all its unpredictable memory in a The operations on the data is carried out using MapReduce.
disappointment. To diminish the measure of squandered For all operations on the data, it is easier to access data from
calculation, it occasionally composes its checkpoints to stable the cluster of data storage devices where the data is stored. The
stockpiling. After a disappointment, it is restarted from its last cluster consists of servers that are used to house the data[1].
stable checkpoint. We demonstrate a disappointment as a For further studies, the technical contribution of authors in
straggler occasion with a timestamp equivalent to the the field of ant colony and networks are available in [2].
timestamp of the most noteworthy checkpoint saved money on
3
of the data
use. In the event that this is done, it would permit for latency VMs. A broker implements the policies for selecting a VM to
to be reduced whilst still allowing for smooth, synchronous run a Cloudlet and a Datacenter to run the submitted VMs.
access.
COST
Total Virtual Machine $0.50
Figure: Response times from the process
Cost
graphically Total Data Transfer Cost $0.28
Grand Total $0.78
REFERENCES
[1] Rimal, B. P., Choi, E., &Lumb, I. (2009). A Taxonomy and Survey of
Cloud Computing Systems. NCM, 9, 44-51.
[2] Zhao, W., Melliar-Smith, P. M., & Moser, L. E. (2010, July). Fault
tolerance middleware for cloud computing. In Cloud Computing
(CLOUD), 2010 IEEE 3rd International Conference on (pp. 67-74).
IEEE.
[3] Bala, A., & Chana, I. (2012). Fault tolerance-challenges, techniques and
implementation in cloud computing. IJCSI International Journal of
Computer Science Issues, 9(1), 1694-0814.
[4] Jhawar, R., Piuri, V., &Santambrogio, M. (2013). Fault tolerance
management in cloud computing: A system-level perspective. IEEE
Systems Journal, 7(2), 288-297.
[5] Gong, C., Liu, J., Zhang, Q., Chen, H., & Gong, Z. (2010, September).
The characteristics of cloud computing. In Parallel Processing
Workshops (ICPPW), 2010 39th International Conference on (pp. 275-
279). IEEE.
[6] Zhang, Q., Cheng, L., &Boutaba, R. (2010). Cloud computing: state-of-
the-art and research challenges. Journal of internet services and
applications, 1(1), 7-18.
[7] Jhawar, R., Piuri, V., &Santambrogio, M. (2012, March). A
comprehensive conceptual system-level approach to fault tolerance in
cloud computing. In Systems Conference (SysCon), 2012 IEEE
International (pp. 1-5). IEEE.
[8] Das, Pranesh (2013) Virtualization and Fault Tolerance in Cloud
Computing. MTech thesis.
[9] Malik, S., &Huet, F. (2011, July). Adaptive fault tolerance in real time
cloud computing. In Services (SERVICES), 2011 IEEE World Congress on
(pp. 280-287). IEEE.
[10] Zikopoulos, P., & Eaton, C. (2011). Understanding big data: Analytics for
enterprise class hadoop and streaming data. McGraw-Hill Osborne Media.