Partitioned Data Storage Data Loss Error Model

SWITCH TO API

Description

The Partitioned Data Storage Data Loss Error Model simulates random data corruption and loss events in spacecraft storage systems. Storage devices in space can experience data loss due to various failure mechanisms including radiation-induced latch-up, wear-out effects in flash memory, or controller malfunctions. This model introduces stochastic data deletion events at configurable intervals, removing random segments of data from storage clusters. This enables analysis of data redundancy strategies, downlink prioritization, and overall storage system resilience.

---

Example Use Cases

• Storage Reliability Analysis: Evaluate the impact of random data loss on mission data return and science yield.
• Redundancy Validation: Test the effectiveness of data replication and backup strategies under realistic failure conditions.
• Downlink Prioritization: Analyze the risk of losing high-priority data before transmission to ground.
• Failure Mode Simulation: Model intermittent storage failures for fault detection and recovery algorithm testing.

---

Module Implementation

The Partitioned Data Storage Data Loss Error Model is a Universe Model that attaches to a Partitioned Data Storage component and introduces random data deletion events.

Deletion Timing

Data loss events occur at random intervals drawn from a Gaussian distribution. The time until the next deletion event is:

$$t_{\text{next}}=t_{\text{current}}+\Delta t$$

where $\Delta t$ is sampled from:

$$\Delta t\sim \mathcal{N}({\mu }_t,{\sigma }_t)$$

with ${\mu }_t$ being the TimeMean parameter and ${\sigma }_t$ being the TimeStd parameter.

Deletion Size

When a deletion event occurs, the number of bytes removed is also drawn from a Gaussian distribution:

$$n_{\text{bytes}}\sim \mathcal{N}({\mu }_d,{\sigma }_d)$$

where ${\mu }_d$ is the DeletionMean parameter and ${\sigma }_d$ is the DeletionStd parameter. The result is rounded to the nearest integer with a minimum of one byte.

Cluster Selection

At each deletion event, the model:

1. Selects a random storage cluster from the available units
2. Chooses a random starting byte position within the allocated data
3. Deletes the computed number of bytes, capped by the remaining data in the cluster

Statistical Properties

For default parameters, the expected deletion rate and data loss can be characterized as:

Parameter	Default Value	Description
Time Mean	100 s	Average interval between deletion events
Time Std	10 s	Variability in deletion timing
Deletion Mean	2 B	Average bytes deleted per event
Deletion Std	4 B	Variability in deletion size

The expected data loss rate over time is approximately:

$$\dot{D}\approx \frac{{\mu }_d}{{\mu }_t}$$

Seed Control

The SetSeed method allows explicit control of both the cluster selection and distribution sampling random number generators, ensuring reproducible deletion sequences for Monte Carlo analysis.

---

Assumptions/Limitations

• Deletion events only occur when storage clusters contain data; empty clusters are skipped.
• The Gaussian distributions may produce negative values which are handled by enforcing minimum bounds.
• All clusters have equal probability of experiencing a deletion event regardless of size or usage.
• Deleted data is permanently removed; no recovery mechanism is modeled.
• The model does not distinguish between different data types or priorities when selecting data for deletion.
• Deletion timing is independent of system state; correlated failures are not modeled.