Computational storage is a concept for computer-based data storage. The concept differs from currently used data processing in that the stored data is processed directly at the storage level, rather than being transferred to the computational level and processed there.
In traditional data processing and data analysis, the subset of data needed for data analysis is moved from a terabyte of memory to smaller memories at the computational level, where it is processed by the central processing unit. Corresponding working memories are RAMs and DRAMs, and their storage capacity is several gigabytes. The situation is different with computational storage, where computing services are provided directly at the storage level. This means that data no longer has to be moved, which benefits real-time computing. The computational storage concept works with several multi-core processors that perform various functions such as indexing the data during storage, searching for entries or supporting artificial intelligence programs. A key aspect is real-time storage and analytics.
The Storage Networking Industry Association( SNIA), through its Computational Storage Technical Working Group (TWG), has addressed this issue and defined two components for computational storage: The Computational Storage Drive( CSD) as a component that provides computational services in the storage system, and the Computational Storage Processor( CSP) that provides computational services in the storage system.