|dc.description.abstract||With the advent of Big Data technologies-the capacity to store and efficiently process large sets of data, doors of opportunities for developing business intelligence that was previously unknown, has opened. Each phase in the processing of this data requires specialized infrastructures. One such phase, the preservation and archiving of data, has proven its usefulness time and again. Data archives are processed using novel data mining methods to elicit vital data gathered over long periods of time and efficiently audit the growth of a business or an organization. Data preservation is also an important aspect of business processes which helps in avoiding loss of important information due to system failures, human errors and natural calamities.
This thesis investigates the need, discusses possibilities and presents a novel, highly cost-effective, unified, long- term storage solution for data. Some of the common processes followed in large-scale data warehousing systems are analyzed for overlooked, inordinate shortcomings and a profitably feasible solution is conceived for them. The gap between the general needs of 'efficient' long-term storage and common, current functionalities is analyzed. An attempt to bridge this gap is made through the use of a hybrid, hierarchical media based, performance enhancing middleware and a monolithic namespace filesystem in a new storage architecture, Tape Cloud.
The scope of studies carried out by us involves interpreting the effects of using heterogeneous storage media in terms of operational behavior, average latency of data transactions and power consumption. The results show the advantages of the new storage system by demonstrating the difference in operating costs, personnel costs and total cost of ownership from varied perspectives in a business model.||