01
Object-storage economics
Use S3, GCS, Azure, or MinIO as the durable telemetry layer.
Solution · Telemetry cost control
Control the storage model before cutting visibility.
Keep telemetry on object storage you choose, use columnar pruning and caches to reduce reads, and evaluate cost with visible assumptions.
- object-store choice
- S3 / GCS / Azure
- columnar format
- Parquet
- metadata and result caches
- 3 levels
- default disk-cache cap
- 10 GB
Cost control in the architecture, not a billing toggle.
molesignal moves the economic boundary to infrastructure you can inspect and tune.
02
Pruned columnar reads
Use time, Parquet metadata, Tantivy indexes, and file-level planning to avoid broad scans.
03
Three cache levels
Cache file metadata, Parquet metadata, query results, and hot Parquet blocks on disk.
04
Scale roles independently
Split ingest, query, compaction, alerting, and connector work only when the workload requires it.
workflow
A more inspectable cost model.
The evaluation starts with your volume and retention rather than a marketing percentage.
- 101 · Measure
Use actual daily ingest and retention
Start with the volume the team wants to keep, not a pre-filtered subset.
- 202 · Model
Separate storage, compute, and operations
Compare infrastructure assumptions with the current commercial pricing model.
- 303 · Test
Run representative queries
Validate pruning and latency against your fields and time ranges before changing retention.
- 404 · Tune
Change caches and runtime roles
Scale the expensive part of the workload instead of scaling every component together.
proof / shipped
The calculator keeps its assumptions visible.
The existing cost page exposes ingest, retention, pricing snapshot, and infrastructure baseline rather than promising a universal saving.
- Interactive ingest and retention inputs
- Public pricing snapshot date
- Object storage plus compute baseline
- No invented production benchmark numbers
Solution · Telemetry cost control
Control the storage model before cutting visibility.
Keep telemetry on object storage you choose, use columnar pruning and caches to reduce reads, and evaluate cost with visible assumptions.