Data management policy
The way Rely.io manages and stores the data that comes from your various data source integrations must be understood so that one can clearly interpert the calculations, values and charts displayed throughout the platform. Since at any given time Rely.io might be ingesting very large amounts of datapoints, storing all this data indefinitely is not viable. As such, Rely.io applies a downsampling policy to historical time series data so that we can scale alongside our customers.
We can generically categorize user data in two ways:
- Raw data: data that is being ingested directly from a user's data sources.
- SLO data: data that is generated as a consequence of monitoring an SLO such as SLI time series data, error budget and burn rate data, etc...
Rely.io only stores raw data if it's being actively used in the calculation of a given SLI. This type of data is subject to down-sampling. Otherwise, Rely.io requests this data directly to the user's data source integrations when it's needed and does not store it. An important exception are push integrations such as Prometheus Remote Write for which Rely.io handles raw data a bit differently.
SLO data is always stored by Rely.io and is not subject to down-sampling.
Both raw data and SLO data is stored by Rely.io up to 1 year. If you require a larger data retention period, you can contact a Rely.io representative directly to work on a custom data retention plan.
Raw data sent to Rely.io through push integrations such as Prometheus Remote Write has a TTL of 14 days and is down-sampled to 1 hour intervals after 1 day has elapsed.
Down-sampling is the mechanism through which high granularity time series data sets are transformed into a more coarse and manageable set of data. For example, aggregating a time series with 1 minute intervals into a lower frequency time series of 1 hour intervals.
As time moves forward, historical raw data is down-sampled according to the following table:
Maximum Time Series Interval
<= 1 day
<= 3 days
<= 7 days
<= 2 weeks
<= 1 month (30 days)
> 1 month