Abstract:
Data deduplication is a unique data compression technique. It has been extensively adopted to save backup time in addition to storage area, especially in backup storage systems. Rapid increase in data and associated expenses has inspired the need to optimize storage and transfer of data records. Deduplication has proven a incredibly powerful technology in removing redundancy in backup data storage. With the explosive increase in data sizes, the I/O bottleneck has emerge as an more and more daunting task for big data analytics within the Cloud. Previous researches have shown that slight to high data redundancy truly exists in primary storage systems within the Cloud. Furthermore, at once making use of data deduplication to primary storage systems inside the Cloud will probably reason space contention in physical memory and information fragmentation on disks. Primarily based on these observations, we suggest a overall performance-orientated I/O deduplication, referred to as POD, in place of a capability-orientated I/O deduplication, exemplified by using iDedup, to enhance the I/O performance of primary storage structures in the Cloud without sacrificing capability savings of the latter. POD takes a -pronged method to improving the overall performance of primary storage structures and minimizing overall performance overhead of deduplication, specifically, a request-based selective deduplication method, known as select-Dedupe, to alleviate the data fragmentation and an adaptive memory control scheme, referred to as iCache, to ease the memory competition among the bursty read traffic and the bursty write visitors. Our evaluation consequences additionally display that POD achieves comparable or higher potential memory savings than iDedup.