Article Preview
Top1. Introduction
Cloud Computing is a large-scale distributed computing epitome that is driven by economies of scale as per use basis, dynamically scalable, in which a pool of abstracted / scatted virtualized, manageable and colossal computing power, mammoth storage, stupendous type of platforms, and services are delivered on demand to external customers invoke guaranteed Quality of service over the Internet which could be accessed in a simple and a pervasive way. It has the capability to distich the internet and wide area network (WAN) to use the remotely available resources to furnish cost effectual solutions on the provision of pay per use basis. Cloud computing – a relatively recent term, embraces cyber infrastructure and is reinforced upon distributed computing, utility computing, virtualization, and more latterly networking, software, web and vane services. It insinuates a service oriented architecture (SOA), providing reliability and an aspect to reduce information technology overhead for the end-user, furnishing greater flexibility and scalability, facilities location independence, reduced total cost of ownership and on-demand services (Patra, et al. 2013a). Now in distributed and indiscriminately changing environment of cloud it is hard to implement and provide scalable facilities to every cloud related application (Vouk 2008). The dynamic environment of Cloud results in various unexpected faults and failure.
The capability of a system to respond graciously to an unexpected hardware or software failure is comprehend as fault tolerance (Mittal et al. 2011). The need of fault tolerance in cloud paradigm is a key factor for concern due to the hasty exponential maturation of cloud computing epitome. Fault tolerant cloud is in an important aspect for consideration, in order to achieve robustness and dependability. Failure should be assessed and handled effectively to do so (Bala et al. 2012). Fault tolerance intend to carry out dependability and robustness in any system. On the basis of policies and techniques we can classify fault tolerance into 2 types: 1. Proactive fault tolerance (Bala et al. 2012) and 2. Reactive fault tolerance. The foremost one is to evade recovery from fault, errors and failure or we can say how to anticipate failures. This is done by predicting them (fault, errors and failure) and proactively substitute the suspected component, means detects the problem before it rattling come. There are several challenges we are on the way for implementing the proactive fault tolerance technique. They are a) Fault prediction b) Prediction accuracy and c) application manipulation like migration, pause/un-pause state etc. The reactive fault tolerance procedure reduces the endeavor of failures, when the failure effectively occurs or we can say how to react to a failure. Drilling down, these can be aid classified into two sub-techniques, a) Error processing and b) Fault treatment. Removing errors from the computational state is the primary goal of Error processing. Taking about Ingredients of Error processing, we have two constituent phases. The phases are 1. Effective error processing, 2. Latent error processing. “Effective error processing” which aimed to take the effective error back to a latent circumstance, before (if possible) occurrence of a failure and “Latent error processing” aimed at shielding that the error does not become active again (Lussier et al. 2005). Fault treatment aims at hindering faults from being re- spark off. Now in the case of Adaptive fault tolerance, all the procedure done automatically (it may proactively or reactively) according to the situation, providing learning and adapting (Lussier et al. 2005, Latchoumy and Khader 2011). The existing fault tolerance technique like proactive, reactive and adaptive fault tolerance in cloud computing or in gird rather can say in a distributed computing environment consider various parameters for clustering various architectures into various clusters eventually. These parameters like performance, throughput, response-time, reliability, scalability, availability, security (Patra et al. 2013b), usability, and associated over-head, which are formed by analyzing various architectures which are discussed in our previous paper (Patra et al. 2013a).