What Is Computational Storage?
Computational storage is a storage device architecture that adds compute to storage in ways that drive efficiencies and enable enhanced complementary functions. These architectures typically improve application performance and infrastructure efficiency.
In traditional storage applications, a request is made to move data from storage. In response to the request, the data is moved to compute and processing takes place on the server. Results are then moved back to storage. In most cases, this process must be repeated hundreds of times due to the mismatched size of the system memory and storage.
With computational storage, an operation is requested rather than the movement of data. Processing takes place on the storage device and only the desired results are sent to the server. Because a device uses controllers and additional memory to process data in the storage plane, the data doesn’t have to be transmitted back and forth between storage and computing planes, or moved between various locations for analysis.
Why Is Computational Storage Important?
The goal of computational storage is to allow the data collected by a device to be processed locally on the drive where it’s stored, reducing or eliminating the processing required on a remote server. This minimizes the need to transmit data over long distances, which reduces latency, bandwidth usage and energy requirements. This also aids in security and privacy as the data never leaves the drive.
When data is processed on the storage controller, the value of the data collected by IoT devices is maximized and opportunities in machine learning and other applications that rely on analytics are introduced. This also enables organizations to maximize the benefits of big data, with faster, easier access to information and opportunities for more scalable, flexible storage resources.
Adding computing capabilities to traditional storage systems and enabling processing directly on the storage drive, offers the ability to build more power and space-efficient systems at the edge to enable insights to be derived locally with the lowest time-to-value. IoT devices in smart cities, for example, can process data at the source, enabling real-time analysis that can provide fast, often critical information for the delivery of functions and services.