Performance Metrics
The Performance tab on the device detail page charts a device’s raw resource utilization over time — CPU, memory, disk, and network — as interactive time-series graphs. This is the live-and-historical telemetry view, distinct from the computed reliability score: reliability rolls crashes, hangs, and uptime into a single 0–100 number, while Performance shows the underlying utilization values themselves so you can spot spikes, saturation, and trends.
Open a device from the Devices list and select the Performance tab (the URL hash is #performance). A condensed copy of the same graphs also appears on the device Overview tab.
What the graphs show
Section titled “What the graphs show”The tab renders up to three stacked charts. The network and disk-activity charts appear only when the device has reported data for them.
- CPU / Memory / Disk usage — a line chart with three series (CPU blue, Memory green, Disk purple) plotted as a percentage on a fixed 0–100% axis. This chart is always shown.
- Network Bandwidth — an area chart of download (inbound) and upload (outbound) throughput. Appears only when the device has reported non-zero bandwidth.
- Disk Activity — an area chart of read and write throughput (bytes/sec). Appears only when the device reports disk activity data.
Below each chart, latest stat cards summarize the most recent sample: current CPU / Memory / Disk percentages, download and upload rates, and read/write throughput plus read/write operations per second.
Time ranges
Section titled “Time ranges”Three range buttons in the chart header switch the window. Each range automatically selects an aggregation interval so the graph stays readable:
| Range | Aggregation interval |
|---|---|
| 24h | 5-minute buckets |
| 7d | 1-hour buckets |
| 30d | 1-day buckets |
Recent, fine-grained ranges are drawn from raw per-minute samples; longer ranges are served from pre-computed hourly and daily rollups, so the 30-day view loads quickly without scanning every raw sample.
How collection cadence affects the graphs
Section titled “How collection cadence affects the graphs”The graphs are only as detailed as the samples the agent reports. Each point is an average over its bucket, so:
- A device that has been offline or newly enrolled shows gaps or a short history — there is simply no data for those buckets.
- Because points are bucket averages, a brief spike between samples can be smoothed out. Narrow the range to 24h for the finest resolution.
- Percentages are clamped to the 0–100% axis; bandwidth and disk-activity axes auto-scale to the observed maximum.
Process drill-down
Section titled “Process drill-down”Click any point on any chart to open the process drill-down panel. It lists the top processes by CPU and memory at (or just before) the clicked timestamp, so you can attribute a spike to a specific process. Toggle Live in the panel to pull the device’s current running processes on demand instead of the historical sample. If a device has never recorded a process sample, or the click predates its first sample, the panel explains why it is empty rather than showing nothing.
Performance metrics vs. reliability score
Section titled “Performance metrics vs. reliability score”These two device views answer different questions:
- Performance metrics (this tab) — how hard is the device working right now, and over the last day/week/month? Raw CPU, memory, disk, and network utilization as time series.
- Reliability — how stable is the device? A weighted 0–100 score computed from crash events, hangs, service failures, hardware errors, and uptime.
Use Performance to investigate a slowness or saturation complaint; use Reliability to judge whether a device is fundamentally unstable and needs attention.
API reference
Section titled “API reference”The tab is backed by a single endpoint:
GET /api/v1/devices/:id/metrics?range=24h| Parameter | Description |
|---|---|
range |
Convenience window: 1h, 6h, 24h, 7d, 30d. Selects a matching interval automatically. |
startDate / endDate |
Explicit ISO 8601 window (overrides range). |
interval |
Aggregation bucket: 1m, 5m, 1h, 1d. Defaults to 5m. |
The response returns an array of points, each with timestamp, cpu, ram, disk (percentages), bandwidth in/out (bps), and disk read/write throughput and ops. The endpoint requires the devices.read permission and an organization, partner, or system scope.
Related
Section titled “Related”- Device Reliability — the computed stability score, complementary to these raw graphs.
- Boot & Startup Performance — boot-time timings and startup-item impact.
- Devices — the device list and detail page these graphs live on.