1. The logic that had to remember everything

Murray Leinster’s “A Logic Named Joe” returns as the recalled work because the network in that story is valued not for the logics themselves but for what they can retrieve. Memory is the product. The desktop’s storage attachment is in the same business: keeping bits alive long enough to be useful, whether that means hours between downlink passes or years for a customer archive.

This entry reads about storage reliability and data management in orbit.

2. The sobering number

The MDPI review “Reliability and Risk in Space-Based Data Centers” reports that HPE Spaceborne Computer-2 saw a 45 % SSD failure rate during its mission. Not CPUs, not DRAM, not power supplies — solid-state storage. The review contrasts this with the Axiom/AWS Snowcone experiment, which reportedly survived without the same storage casualty rate. The difference appears to be less about orbit than about how each system handled radiation-induced effects in NAND flash and the firmware that manages it (MDPI).

That single number should change how the desktop thinks about storage. A drive that works on day one is not the same as a drive that works on day three hundred.

3. Why storage dies in orbit

Storage in orbit faces a stacked set of hazards:

  • Single-event effects: high-energy particles flip bits in NAND flash cells and can corrupt the controller’s SRAM or firmware state.
  • Total ionizing dose: accumulated radiation degrades oxides and shifts threshold voltages, reducing retention and increasing error rates.
  • Temperature cycling: repeated hot and cold passes stress solder joints, packages, and the flash cells themselves.
  • Controller complexity: SSDs hide bad blocks, wear leveling, garbage collection, and mapping tables behind firmware. A radiation strike in the controller can make the whole drive disappear.

A Seagate overview notes that commercial SSDs use low-density parity codes and complex mapping tables that were not designed for these conditions; space-ready storage needs error recovery and firmware behavior matched to the environment (Seagate).

4. Strategies for keeping data alive

The literature offers several approaches:

  • SLC NAND: single-level cells are more radiation-tolerant than multi-level cells because the margin between states is larger.
  • Radiation-tolerant controllers and firmware: ATP Electronics and others offer SSD controllers designed to detect and recover from radiation-induced errors without locking up (ATP Electronics).
  • Redundancy across drives: keep multiple copies, compare them, and rebuild when one copy diverges.
  • Application-level protection: checksums, erasure coding, and provenance so that corrupted data is detected before it is delivered.
  • New materials: Georgia Tech researchers reported a NAND-flash-related storage technology claimed to be up to 30 times more radiation-resilient than current data storage, though it is early-stage for deep-space missions (Georgia Tech).

5. Data management, not just drives

Reliable storage is only half the problem. The other half is managing what is stored:

  • Tiering: hot data on fast local SSD, warm data on redundant pooled storage, cold data queued for downlink.
  • Prioritization: not every byte is equally valuable. Telemetry, alarms, and paid payload products outrank logs.
  • Provenance: a delivered data product should be traceable to a sensor measurement, a processing step, and a time.
  • Retention policy: how long data stays on board before it is deleted or archived to ground.

6. What this changes

  • Storage is the most likely compute-related casualty in orbit; the desktop must design for drive failure as a normal event.
  • Drive-level reliability is necessary but not sufficient; data must be protected across drives and checked end-to-end.
  • The next entry will read about flight software architectures and how workloads are actually run on orbit.