Downsampling
Downsampling is the practice of reducing the resolution of time-series data by aggregating many fine-grained points into fewer, coarser ones. For example, turning per-second readings into per-minute averages. It saves storage but loses detail.
The tradeoff downsampling makes
Downsampling is a common way to control the cost of keeping time-series history. Recent data stays at full resolution, but older data is rolled up into larger buckets: minute averages, then hourly, then daily. The old detail is discarded.
It is a reasonable strategy when you genuinely do not need old detail. The danger is downsampling because storage is expensive, not because you stopped caring about the detail. When an incident or audit sends you back to last quarter's data, downsampled history may have thrown away exactly the spike or anomaly you needed to see.
The better position is to downsample by choice, not because your database forces it on you for cost reasons.
How Arc handles Downsampling
Because Arc keeps data as cheap compressed Parquet on object storage, the cost pressure to downsample is much lower. Many teams keep full-resolution history far longer with Arc than they could afford to on a traditional time-series database.
Arc is a high-performance columnar database. Open Parquet on storage you own, single Go binary, production-ready in 30 seconds.