Limits and parameters
All the default limits and parameters of the Glass Data platform are listed below. If you reach any of them, get in touch: Glass Data evaluates exceptions on a case-by-case basis.
Retention
Original data
The platform retains the events received for 5 years, with every field sent and those filled in by rules. The Explore screen queries this data.
Retention applies per event type and does not depend on volume: a payment event type and an authentication one are retained for the same period. The exception is the failure event type, _failure, retained for 7 days.
Metrics and alerts
The retention of a metric and its alerts depends on the aggregation interval:
| Interval | Retention |
|---|---|
| 1 minute | 7 days |
| 5 minutes | 30 days |
| 1 hour | 6 months |
| 1 day | 1 year |
| 7 days | 5 years |
Monitoring
Anomaly detection trains the model with the recent history of the metric, measured in intervals of the monitor itself. The default window depends on the interval and on whether the seasonality adjustment is enabled.
| Interval | Default analysis window | With seasonality adjustment |
|---|---|---|
| 1 minute | 40 intervals | not applicable |
| 5 minutes | 40 intervals | not applicable |
| 1 hour | 40 intervals | 168 intervals (7 days) |
| 1 day | 40 intervals | 49 intervals (7 weeks) |
| 7 days | 40 intervals | not applicable |
The analysis runs only when the window has at least 3 points. The platform discards the points still within a suspended learning period before that check.
The window can be adjusted per monitor, along with the sensitivity and the other options described in advanced settings.
Alerts
The platform automatically closes an open alert by normalisation or by expiration. A new alert also temporarily interrupts machine learning, so that the anomaly does not enter the predictive models too early.
| Interval | Normalisation | Expiration | Learning pause |
|---|---|---|---|
| 1 minute | 3 measurements | 10 minutes | 2.5 hours |
| 5 minutes | 3 measurements | 50 minutes | 6 hours |
| 1 hour | 2 measurements | 10 hours | 12 hours |
| 1 day | 1 measurement | 10 days | 3 days |
| 7 days | 1 measurement | 70 days | 3 weeks |
Normalisation is the number of consecutive normal measurements required to close the alert. Normalisation can be adjusted per monitor. Expiration equals 10 intervals of the monitor.
Metrics and monitors
| Limit | Maximum quantity |
|---|---|
| Metrics per event type | 20 |
| Monitors per event type | 20 |
| Dimensions per metric | 10 |
| Values per filter | 10 |
| Aggregations queried in the chart at once | 5 |
| Series expanded from a list in the chart | 10 |
| Trackable fields per event type | 2 |
| Tracking retention | 90 days |
Rules and lists
| Limit | Maximum quantity |
|---|---|
| Rules per account | 30 |
| Conditions per rule | 30 |
| Lists used per condition | 3 |
| Lists used per rule across all conditions | 30 |
| Characters per expression | 1024 |
| Characters per regular expression | 256 |
| Open drafts per rule | 3 |
| Items per list | 10,000 |
Analyses
| Limit | Maximum quantity |
|---|---|
| Window of a cell | 90 days |
| Fields combined in a cell’s dimension | 5 |
| Values per filter | 20 |
| Observations evaluated on percentage measures | 100 |
See analyses for what each of these fields means.
Export
Exporting records on the Explore screen obeys the following limits:
| Limit | Maximum quantity |
|---|---|
| Records per export | 10,000,000 |
| Records exported per day across all exports | 10,000,000 |
The admin panel keeps each export available for download for 7 days.
Metadata
| Limit | Maximum quantity |
|---|---|
| Metadata per entity | 10 |
| Cardinality per metafield | 10,000 |
Cardinality is the number of different values in a metafield. The platform does not prevent registering values above the maximum cardinality; metrics can ignore the less frequent dimensions.