Methodology
Behind every number,
a definition.
Measurements are useful when their scope is clear. Understand the event, the reporting window and the limits of what a metric can prove.
01Define the evidence
Reliability reporting needs clear definitions, a reporting window and visible limitations. Read measurements together with the context that explains how they were collected.
02System State
Operational describes expected service behavior. Degraded means service remains available with reduced performance or reliability. Outage means a service is unavailable. These are definitions, not a current assessment of Driyu.
Missing status data is unknown; it does not establish either an outage or normal operation.
03Target Availability
The reliability level the system is aiming for. A target describes an objective; it does not prove that the service achieved it or guarantee future availability.
For example, a 99.99% annual target allows about 52 minutes and 34 seconds of downtime in a 365-day year. It is an objective, not measured uptime or a guarantee. Compare it with the actual reporting window and coverage.
04STS & the notification process
The median is the middle value in the recorded timing samples. This metric measures the interval that starts when the report request begins and ends when the owner-email sending step completes.
It is not the time from opening a camera, proof that a message was read, or the time until a pet returned home.
0595th Percentile STS
P95 is the value at or below which 95% of the recorded timing samples fall. It helps explain the slower end of a distribution alongside the median.
It is not a guaranteed maximum. The slowest cases can exceed it.
06Edge Routing
This label reflects the routing-region information included in the published metric data. It describes infrastructure, not a network of nearby people responding to a lost pet.
The guide explains the label; use the Trust Center for available routing data.
07The rolling 90-day record
Historical metrics describe their reporting windows. Incident entries can add impact and remediation context. A partial window or a missing value needs an explanation rather than a made-up result.
Green indicates stronger uptime, amber reduced performance, and gray insufficient data. Read the values and coverage as well as the colors; an incomplete day is not a confirmed outage.
08Incident history
Published incidents provide timing, impact and remediation context when available. Resolved labels a completed response; Archive identifies older records. Read incident context alongside changes in the metrics.
This guide does not load an incident feed. That is different from saying there were no incidents.
09Latency-triggered rollback
The published approach allows a change to be paused or reverted when performance worsens beyond its review criteria. Stable operation takes priority over expanding a rollout.
A process description does not show that a rollback occurred or prove current service health. Look for the relevant incident record and measurements.
PUBLISHED BOUNDARIES
The words
matter, too.
- We do not promise guaranteed recovery.
- We do not publish recovery-rate claims unless they can be independently verified.
- We are not GPS.
- We do not claim a nationwide alert network.
- We do not claim a 24/7 recovery team.
KEEP READING