Edge computing architectures typically consists of an array of IoT sensors that are collecting data in realtime which then send data to an edge data center for additional processing. Eventually, much of the collected data ends up in a public cloud for further, more detailed processing. Storage, throughout the edge architecture, needs to provide fast and consistent performance.
Designing SSDs for Edge Computing
Each tier of the edge architecture needs SSDs, but each layer requires different capabilities from the SSD. At the top tier are the IoT sensors. Each sensor is typically capturing data in realtime. The sensors need to store the data they are obtaining at the lowest latency possible so as not to interrupt subsequent captures. A busy sensor can send an almost constant stream of data to its SSD. In most cases, this steam does not require a lot of bandwidth, but it does need consistent latency. SSD vendors need to design the SSD so that latency inducing background functions, like garbage collection, don’t negatively impact the write stream.
The second tier is the edge data center. These smaller data centers typically act as an offload point for the sensor data as well as performing some processing of data. These devices also need excellent write performance. The IO of a single sensor may not overwhelm them but the aggregate IO of thousands of sensors will. These edge data centers also need to be equipped with SSDs that can handle the high write IO rates of thousands of sensors while again being very consistent. Since the edge data center also processes the data that the IoT devices send them, they also need excellent random IO performance.
The third tier is either a private or public cloud data center. This data center will also process data as well as store it long-term. This data center needs SSD than can meet the IO demands of further analysis of data initially collected by the IoT devices. It can also benefit from less expensive SSDs designed for long-term data retention.
The challenge for edge data architecture is finding SSDs that can meet the various demands of each tier. Consistency is a need throughout the architecture, but some components need higher performance than others, while some need low-cost retention capabilities. Something that the organization needs to try to find is a single vendor who can adjust their SSDs for each specific use case.
To learn more about using SSDs as part of edge architecture watch our Lightboard Video.