Improving Upon Current Fault Tolerance Techniques with Emphasis on Low-Latency Fault Tolerance in Cloud Systems

摘要

With the increase in efficiency of Cloud computing as a second-option to old-fashioned information processing systems, its reliability and stability in the face of faults or of faulty components has become an area with great scope for research and innovation. A typical cloud computing data centre consists of multiple servers that efficiently co-ordinate the storage, execution, and handing of users' tasks. One of the obvious bottlenecks here are the servers themselves. Each consists of hardware devices, like processors, memory registers, network cards, etc., all of which have a scope for failure. This hardware failure can be tackled in two ways, one being a software method and the other a hardware-based or network or a connection-based solution. In this paper, we have studied the various methods and algorithms being employed to combat the issue of fault tolerance as of today, and using the data that we have collected to improve upon a current method. One of the proposed models is LLFT, which contains a Low Latency Fault Tolerance middleware for providing fault-tolerance based on the frequency of usage. Similarly, there are many areas where certain models do not work well. Hence, through the course of this paper, we hope to incorporate a tried and tested file system management method in cloud computing systems such that an efficient and secure fault tolerant model can be developed, improving upon the current methods.