Plan raw ingest, compression, retention, replication, and storage tiers for IoT telemetry
AnimationBeginner firstCapacity planning
IoT Storage Calculator
Follow one day of telemetry from device samples to hot, warm, and cold storage.
Change the fleet size, sample rate, compression, retention, and replication to see
why raw ingest is only the first part of an IoT storage estimate.
Daily ingest after reduction0 GB/day
Total durable footprint0 GB
Monthly tier cost$0/mo
1. SampleDevices produce timestamped readings.
2. EncodePayload plus headers creates raw bytes.
3. ReduceCompression and rollups cut volume.
4. RetainHot and warm windows keep data queryable.
5. PlaceEdge buffer and cloud tiers share storage.
6. GrowReplication and fleet growth set headroom.
Animated Storage Pipeline
Samples are leaving devices and becoming raw telemetry bytes.
Pipeline stage
Sample
Raw daily stream0 GB/day
Payload, timestamp, device id, and protocol overhead before reduction.
Retained logical data0 GB
Hot, warm, and archive windows before explicit replication.
Recommended patternHybrid edge + cloud
Short edge buffer, queryable hot tier, and cheaper archive for long history.
How To Read The Model
Start with samples
Count devices, readings per hour, payload bytes, and protocol or timestamp overhead. This gives raw ingest per day.
Normalize before sizing
Raw payloads usually become structured records with tags, timestamps, indexes, and metadata. Include that overhead.
Reduce deliberately
Compression, delta encoding, and downsampling reduce bytes. They also decide whether later analysis can still answer the question.
Retain by value
Keep recent data hot for dashboards, older data warm for audits, and long history cold unless frequent queries justify the cost.
Formula Trace
Calculating...
Reference Material
Use these quick references when checking another IoT design or adapting this calculator inside a chapter.
Storage Sizing Quick Reference
Bytes per sample
Include the measured value, timestamp, device identifier, units, status flags, and transport overhead when estimating raw ingest.
Sample rate
Use sustained sample rate for capacity. Burst rate belongs in buffer and throughput checks, not only long-term storage.
Compression
Time-series telemetry can compress well when values change slowly, but encrypted or image payloads may compress poorly.
Rollups
Downsample old raw readings into minute, hourly, or daily aggregates only when the learning or compliance need allows it.
Retention
Separate hot query windows from long archive windows. Treat "keep forever" as a cost and governance decision.
Replication
Explicit replicas increase durable footprint. Some managed cloud prices already include durability, so avoid double counting.
IoT Storage Pattern Guide
Time-series database
Best for high write rates, time-window queries, device dashboards, retention policies, and continuous aggregates.
Object storage
Best for cheap long-term storage, logs, images, model training data, and parquet or columnar files.
Edge buffer
Keeps data during network outages and reduces cloud dependency. Size it from daily reduced ingest and outage duration.
Hybrid pipeline
Common IoT design: short edge buffer, hot database for operations, object storage for history, and summarized views for analytics.
Cold archive
Useful when data must be retained but is rarely queried. Retrieval delay and egress cost can matter during incident review.
Schema discipline
Stable tags, units, and timestamps make storage useful. Poor schemas can make small data expensive to query.
Technical Accuracy Notes
GB vs GiB
This teaching model reports binary sizes using 1024-based units while pricing approximates GB-month behavior. Real bills use provider definitions.
Provider durability
Managed services may include erasure coding and multiple copies. The replication control shows the concept, not a vendor bill.
Index overhead
Indexes, metadata, tags, and query acceleration structures can be significant. They are represented as part of reduced logical data.
Compression limits
Compression factor depends on value repetition, encoding, encryption, image/video codecs, and database storage format.
Retention policy
Deletion, compaction, and rollup policies must be tested. A policy written down but not automated will not control storage growth.
Cost model
Cost excludes requests, compute, network egress, backups, snapshots, and staff time. Use it as a sizing conversation starter.