Distributed Storage Using Similarity-Aware-Partitioning

Distributed Storage Using Similarly-Aware-Partitioning

Princeton Docket # 14-2948

Data volumes are increasing at an unprecedented rate due to the advent of large social networks, on-demand cloud-based video services, and file sharing applications. To store and access data efficiently, there has been an increased demand for distributed data stores. These storage services face an ever-increasing demand on their systems and must satisfy two important concerns: they should be space-efficient to manage the high volumes of data, and they should be access-efficient in terms of minimizing network accesses to read or write to these files

Researchers in the Department of Electrical Engineering at Princeton University have developed a distributed storage tool in Java based on Similarity-Aware-Partitioning (SAP) algorithms. This algorithmic technique is used to present solutions that are both space-efficient and access-efficient. This technique has considerable utility in terms of the cost of disk/memory required for storage while ensuring that customer deadlines are met due to the low access costs. Further, this novel technique presents fundamental insights in how to exploit similarity among data. This technology can be used in other areas that deal with data, such as data processing, analytics, and data communication. In addition, distributed storage systems form the backbone of many web-based services. This novel distributed storage tool based on the new algorithms can be used to perform deduplication in such storage systems.