Towards a Signature Based Compression Technique for Big Data Storage
Constantinos Costa, Panos K. Chrysanthis, Marios Costa, Efstathios Stavrakis, Nicolas Nicolaou · 2023
Although the volume of stored data doubles every year, storage capacity costs decline only at a rate of less than 1/5 per year. At the same time, data is stored in multiple physical locations and remotely retrieved from multiple sites. Thus, minimizing data storage costs while maintaining data fidelity and efficient retrieval is still a key challenge in database systems. In addition to the raw big data, its associated metadata and indexes equally demand tremendous storage that impacts the I/O footprint of data centers. In this vision paper, we propose a new signature-based compression (SIBACO) technique that is able to: (i) incrementally store big data in an efficient way; and (ii) improve the retrieval time for data-intensive applications. SIBACO achieves higher compression ratios by combining and compressing columns differently based on the type and distribution of data and can be easily integrated with column and hybrid stores. We evaluate our proposed tool using real datasets showing that SIBACO outperforms "monolithic" compression schemes in terms of storage cost.