Embedded systems that were once updated only occasionally are now expected to evolve throughout operational lifetimes that can exceed 15 years. That shift is exposing a weakness in traditional storage planning: partition sizes are fixed early, while application growth, logging, configuration data, and future update needs continue to change after deployment.
In the classic model, engineers reserve space for the operating system, applications, diagnostics, and update images long before the first device ships. That worked when software changed slowly and storage demand was predictable. In software-defined systems, however, one workload may remain nearly constant while another grows across several releases, leaving some flash exhausted and other partitions underused.
The article notes that simple overprovisioning is becoming harder to justify as NANDNot ANDA logic gate that outputs 0 when all inputs are 1, and outputs 1 in all other cases and NOR prices rise. Extra flash reserved “just in case” turns into a bill-of-materials issue, especially when OTA strategies require additional headroom for downloading, validating, and rolling back firmware. Traditional A/B partitioning remains robust because it keeps a previous image available, but it also permanently reserves a large amount of capacity for update scenarios.
As an alternative, the piece presents a dynamic data layer in which storage is managed as a shared resource instead of a set of rigid partitions. Applications can stay isolated through subvolumes, while quotas and reservations preserve guaranteed capacity for critical workloads. In parallel, snapshot-based update methods can track changes at block level, so only modified data needs to be written during an update rather than duplicating complete images.
The underlying constraint does not disappear: deterministic behavior, functional safety, reliability, and predictable recovery still matter. The change is architectural. Storage is no longer just background infrastructure once the partition map is set; it becomes part of how software can continue to grow, update, and remain maintainable over the full life of a connected embedded product.