The satellite industry is at a crossroads, and it’s not just about launching more spacecraft into low Earth orbit (LEO). What’s truly fascinating is the invisible bottleneck emerging behind the scenes: the cardinality wall. Personally, I think this is one of the most underappreciated challenges in modern space operations. It’s not about rockets or payloads; it’s about the data—specifically, the sheer volume and complexity of telemetry streaming from these constellations. If you take a step back and think about it, we’re talking about millions of telemetry streams in real-time, each carrying layers of metadata. This isn’t just a scaling issue; it’s a fundamental shift in how we think about space infrastructure.
One thing that immediately stands out is how unprepared traditional systems are for this deluge. Most telemetry pipelines were designed for a handful of satellites, not fleets of hundreds or thousands. What many people don’t realize is that the problem isn’t just about collecting data—it’s about preserving context and fidelity. For instance, a single satellite can generate tens of thousands of telemetry signals per second, from battery voltages to software events. Multiply that by hundreds of spacecraft, and you’re looking at a data tsunami. This high cardinality—the explosion of unique telemetry streams—is where traditional ground system databases start to crumble.
From my perspective, the root of the issue lies in how these systems were built. Relational databases and log analytics platforms, while powerful, weren’t designed for this kind of workload. They rely heavily on indexing, which becomes a liability when dimensions like spacecraft ID, subsystem, and component multiply exponentially. The system spends more time maintaining indexes than processing data. What this really suggests is that we’re trying to fit a square peg into a round hole.
A detail that I find especially interesting is how telemetry retention exacerbates the problem. Operators want to store data for years or decades to train predictive models or investigate anomalies. But here’s the catch: few databases can handle both real-time ingestion and large-scale historical analysis simultaneously. This forces operators into a dangerous tradeoff—simplify the data or risk system failure. Take Loft Orbital, for example. They were ingesting over 500 million telemetry measurements daily, and their relational database approach couldn’t keep up. By switching to a time series-oriented architecture, they regained control. This raises a deeper question: How many operators are still clinging to outdated systems, sacrificing context for stability?
What makes this particularly fascinating is the impact on anomaly detection and machine learning. Stripping metadata or downsampling signals might reduce load, but it also blinds engineers. Imagine trying to diagnose a thermal spike without knowing the orbital position or solar panel orientation. It’s like solving a puzzle with half the pieces missing. For machine learning models, this loss of context is catastrophic. Models trained on high-resolution telemetry lose their predictive power when the data is stripped down.
If you ask me, the cardinality wall isn’t just a technical challenge—it’s a wake-up call. Telemetry infrastructure is no longer a peripheral IT component; it’s mission-critical. The teams making progress are the ones decoupling their pipelines, isolating bottlenecks, and redesigning from the ground up. Incremental tuning won’t cut it anymore. The next generation of LEO infrastructure needs to be built with scale, distribution, and context preservation in mind from day one.
In my opinion, this is where the industry will separate the innovators from the laggards. Those who recognize telemetry as a distributed systems problem—not just a database issue—will thrive. The rest will find themselves drowning in data, unable to extract meaningful insights. As we push the boundaries of space exploration, let’s not forget that the real frontier might just be in how we handle the data we collect.
Key Takeaways:
- The cardinality wall is a hidden bottleneck in LEO constellations, driven by the explosion of telemetry streams and metadata.
- Traditional databases fail due to indexing inefficiencies, uncontrolled write workloads, and misalignment with time-based queries.
- Simplifying data to reduce load introduces blind spots, compromising anomaly detection and machine learning.
- Progress requires decoupling pipelines, preserving context, and treating telemetry as mission-critical infrastructure.
What this really suggests is that the future of space operations isn’t just about what we launch—it’s about how we manage the data that comes back down to Earth.
