Abstract
Economics of back and forth data transfer with centralized data centers and time sensitivity of certain analytics and decision workloads have created a need for intermediate storage and data processing locations, called local edge data centers. Intelligent edge devices are required to store some data and process workloads at the edge and send the remaining to centralized data centers. The data centers have various storage performance tiers, from nearline HDDs to boot and edge devices to high performance NVMe SSDs, driven by economics of performance and QoS. The nearline high-capacity SMR HDDs can be used for sequential data, while high-performance NVMe SSDs can ingest data for machine learning applications. Similarly, edge device options include different SSDs, from SATA to NVMe, available in various form factors such as M.2 and 2.5-inch. However, today application platforms are not designed to take full advantage of the right storage devices- the right application fit. This presentation will focus on application system design and various storage device choices as per the application platform requirements from edge to datacenter systems.