Read a monitor’s performance charts

The Performance tab charts every figure we can read for a monitor. Use it to find when a problem started and what changed.

Open the charts

1. Open the monitor from Monitoring.
2. Select the Performance tab.
3. Choose the time window at the top. See View a monitor for the presets and zoom.

Choose how values are summed up

• Statistic sets how each point of a chart is calculated. Average is the mean value for that slice of time. P95 shows the value 95% of readings stayed under. Peak is the highest value. P95 appears only for monitors that answer requests.
• Previous Period draws the same length of time just before the window as a second line, so you can compare. Charts that can’t compare say No Comparison for This Chart.
• The label above each chart says how points are grouped, for example 5-Minute Averages or Hourly Peaks.

The strip under the toolbar shows the main figure across a longer base window. Drag its handles to move the window. Hover over any chart to see the same moment on all of them, and drag across a chart to zoom.

The sections

Each monitor shows only the sections it can measure.

• Response Time charts how long the resource took to answer, in milliseconds. Response Time by Location splits it by checking place. Response Time Breakdown splits each answer into DNS Lookup, Connection, TLS Handshake, First Byte and Download. Lower is better. Watch for jumps above your usual level.
• Packet Loss charts the share of test packets lost, in percent. A healthy reading is 0%.
• CPU charts CPU Usage in percent, with Usage per Core, CPU Time Breakdown and Interrupts and Context Switches. Steady use near 100% means the server has no spare capacity.
• Memory charts Memory Usage and Swap Usage in percent or bytes. Regular swap use means the server is short of memory.
• Load charts Load Average, the number of processes waiting for the processor. Compare it with the number of CPU cores. A load above the core count means work is queuing.
• Disk charts Disk Usage and Usage by Partition in percent, Disk Throughput for reads and writes in bytes per second, and Disk Busy in percent. Choose one partition or All Partitions.
• Disk Fill Forecast estimates the number of days until a partition reaches a fill level, from recent growth. Used Now is today’s fill. Not Growing means usage is flat or falling, and Reached means the level is already passed. A forecast is an estimate, not a promise.
• Processes charts Blocked Processes, the processes waiting on the disk or network.
• Patterns shows Busy Hours, the average CPU by hour of day, and Monthly Trend, the monthly availability, outages and alarms.

A line under the charts names anything we can’t read in Lighthouse yet, such as network traffic per interface.

If something goes wrong

• Some charts in this section couldn’t be read for this window. means part of the data is missing. Try a different window.
• Not enough data yet. means the monitor hasn’t collected enough history.
• If the tab says there is no history before a date, the monitor was linked that day.