CPU & Load
What Is CPU & Load?
CPU & Load is the view tab that gathers the processor state of the device on a single screen on the TR7 home page. The tab consists of five panels: the usage breakdown that shows on which job types the processor spends its time, the processor load as the rate of the time not spent idle, the 1, 5 and 15-minute load averages, the processor temperature and the usage per core.
While the other tabs of the home page read the totals of the vService, WAF, AGS or DNS service traffic, this tab reads the own hardware measurements of the device. For this reason there is no scope upper row such as All vServices or All DNS Services in the title of the panels and no vService Distribution table that offers a vService breakdown takes place in the panels. The processor usage of a single vService is found not in this tab but in the vService CPU Usage panel on the own dashboard of that vService.
All of the panels are bound to the Settings permission. The panel number of the tab changes according to the type of the device: the CPU Temperature panel is created only on a physical device, on a virtual machine the tab opens with four panels. The series number of the CPU Cores and CPU Temperature panels, which produce a series per core, is also as many as the core number of the device.
Access to the CPU & Load Tab
- The Home shortcut on the left of the Header Bar is clicked.
- CPU & Load is selected from the view tabs.
The tab does not have a subtab, so the second row of the tab bar is not displayed. For the Header Bar shortcuts, see Header Bar.
The CPU & Load Tab
The screen controls are common to all the tabs
The time range selector, the refresh, the settings, the template menu and the arrangement of the panels are the same in all the view tabs of the home page; for the detail, see Screen Controls.
Panels
There are 5 panels in the CPU & Load tab and their order is in the CPU Usage, CPU Load, System Load, CPU Temperature, CPU Cores form. The tab and all of the panels are bound to the Settings permission; in a user who does not have this permission the tab does not take place among the view tabs at all.
The only conditional panel in the tab is CPU Temperature: the panel is created only on a physical device, on a virtual machine the tab opens with four panels and CPU Cores is laid out directly after System Load. There is no other conditional panel; the remaining four panels are laid out on every device.
All of the panels are by default one card wide and one card high. The card width changes according to the screen width: as many columns as the number of cards that fit side by side are opened and the cards fill the column. The only exception is the CPU Cores panel; when it is taken into the Stat Table view, if the core rows do not fit, the card grows up to two cards high. When the panel title is hovered over, the short explanation of the definition is displayed as a tooltip.
The Add Visualization menu offers the same five panels on this tab; there is no other processor panel that can be added to the tab. A panel selected from the menu can be added a second time even if it is already present in the layout. On a virtual machine only four panels are listed in the menu as well.
The panels in this tab are percentage and load values; the Total summary that is present in the rate metrics is offered in none of them. The summaries that can be opened are Last value, Avg. value, Min. value and Max. value; the only exception is the Arc Gauge view: the gauge offers only Last value, Avg. value and Max. value and Min. value is not present in this view.
CPU Usage
Shows on which job type the processor of the device spends its time as five series. The values are the percentages calculated over all the cores.
| Series | Meaning |
|---|---|
| User | The processor time used by the application processes. |
| System | The time the operating system uses for its own internal jobs. |
| I/O Wait | The time the processor waits for an input/output operation such as the disk to be completed. |
| Soft IRQ | The time that goes to the processing of the software interrupts; it rises in the packet processing intensity. |
| Steal | In the virtualized setups, the time the device demands but the infrastructure gives to another guest. It stays zero on a physical device. |
In the default Pie Chart view the five series are shown as ring slices; in the middle of the ring the total of the slices and below it the name of the summary that is open are printed. The slices whose share stays below 2 percent are left unlabeled, when the card shrinks this threshold rises to 5 percent. As a summary, only Avg. value comes open; Last value, Min. value and Max. value are opened from the strip under the panel.
In the panel, the Line Chart, Stat Table and Pie Chart views are offered. In the Line Chart view the five series are shown as stacked areas, for this reason the MIN/MAX band is not offered in this panel. In the cluster setups the Pie Chart and Line Chart views average the value of the two devices; the Stat Table view lists every device in a separate row.
The total of the slices can be smaller than the CPU Load value
CPU Load gathers all the time not spent idle in a single percentage; this panel, on the other hand, separates only the five components above. Since the time spent in the low-priority tasks and in the hardware interrupts is not written into any slice, the values of the two panels do not have to be exactly equal.
CPU Load
Shows the processor usage of the device as a single series (the Load series): the rate of the time the processor does not spend idle over all the cores. The Y axis is fixed between 0 percent and 100 percent and the values are printed with one decimal place.
As a summary, only Avg. value comes open; Last value, Min. value and Max. value are opened from the strip under the panel. The MIN/MAX badge to the left of the legend strip adds behind the averaged line the band that consists of the highest and the lowest values of the same period; thus the short-lived jumps stay visible in the wide time ranges.
In the panel, the Line Chart, Mini Chart, Stat Table and Arc Gauge views are offered. The numerical values are colored according to the thresholds: green up to 70 percent, yellow between 70 percent and 90 percent, red at 90 percent and above; in the values between the thresholds the color makes a gradual transition. The coloring is applied to the value in the Stat Table, Mini Chart and Arc Gauge views, it is not applied to the line in the Line Chart view.
The same panel takes place in the Summary tab with the Arc Gauge view; see CPU Load. In the cluster setups the value of every device is shown as a separate series, it is not summed and not averaged.
System Load
Lists the load average of the system as three series: 1 min, 5 min and 15 min. The load average is not a percentage but a measure based on the number of the tasks that queue for a job on the processor and that are running; for this reason the panel has no unit and the value can rise above both 1 and the core number. To read it roughly, the value is compared with the core number of the device: an average that approaches the core number shows that the processor is saturated, and an average that rises above it that the tasks have started to wait.
| Series | Meaning |
|---|---|
| 1 min | The load average of the last one minute; it is the series that reflects the sudden changes the fastest. |
| 5 min | The load average of the last five minutes. |
| 15 min | The load average of the last fifteen minutes; it shows the permanent load. |
In the default Stat Table view all four of the summary columns come open: Last value, Avg. value, Min. value and Max. value. The header of the first column of the table is Metric and the rows preserve the series order. To the left of every value the moment at which that value was read is written relatively (such as Just now, ~3h ago); the values are printed with at most two decimal places. Threshold coloring is not applied in the panel.
In the panel, the Digital, Line Chart, Stat Table and Arc Gauge views are offered. In the cluster setups the Stat Table, Digital and Arc Gauge views show every device separately; the Line Chart view averages the value of the two devices.
CPU Temperature
Shows the temperature of the processor as Celsius. The series of the panel are Package for the whole of the processor and the core series named in the C1, C2, C3 form for every core. The Line Chart, Stat Table and Arc Gauge views show only the Package series; the core series take place only in the Bar Chart view and in that view the temperature difference between the cores is read.
In the panel, the Line Chart, Stat Table, Arc Gauge and Bar Chart views are offered. The Y axis is fixed between 0 and 100 °C and the values are printed with one decimal place in the 57.0°C form. The numerical values are colored according to the thresholds: up to 60 °C they take the color of the normal band, between 60 °C and 80 °C of the warm band, and at 80 °C and above of the hot band. As a summary, in the Stat Table view all four of the summaries are open. The Line Chart view does not offer a summary box because the series number of the panel exceeds six together with the core number; the summary values are read from the Stat Table or the Arc Gauge view.
In the cluster setups the temperature of every device is shown as a separate series. The list of the core series is produced from the core number of the device on which the panel is opened.
The panel is displayed only on a physical device
In the setups that run as a virtual machine the CPU Temperature panel is not created at all: the tab opens with four panels and the panel is not listed in the Add Visualization menu either. Since the temperature measurement is read from the hardware, it has no equivalent in a virtualized setup.
CPU Cores
Shows the processor usage per core. Every core is a separate series in the C1, C2, C3 form and the value of the series is the percentage of the time that core does not spend idle. The series number is as many as the core number of the device. The panel is used to read whether the load is distributed evenly to the cores: a single bar staying distinctly higher than the others shows that the load is gathered in a single core.
In the default Bar Chart view every core is a bar and the y axis is fixed between 0 percent and 100 percent. As a summary, only Avg. value comes open; Last value, Min. value and Max. value are opened from the strip under the panel. Threshold coloring is not applied in the panel.
In the panel, the Line Chart, Stat Table and Bar Chart views are offered. In the Stat Table view every core is a row and when the rows do not fit, the card grows up to two cards high. In the cluster setups the Bar Chart and Stat Table views show the cores of every device separately; the Line Chart view averages the value of the two devices.
Scope and Limits
INFO
- The tab is displayed only with the Settings permission. All of the five panels are bound to the same permission; in a user who does not have this permission the tab does not take place among the view tabs at all. The visibility of the tab does not look at the device type, the license or the number of defined vServices.
- The panels measure the device, not the vService. No value in this tab is separated per vService. For the processor usage of a single vService the vService CPU Usage panel on the own dashboard of that vService is used, and for the distribution table per vService the Add Visualization menu in the Traffic subtab of the vServices tab.
- CPU Temperature is conditional. The panel is not created on a virtual machine; the tab opens with four panels and the panel is not listed in the Add Visualization menu either.
- Add Visualization does not offer a new panel on this tab. There is no processor panel other than the own panels of the tab in the menu; for the memory, disk and network panels the related tabs are used.
- The combining in a cluster changes according to the panel and the view. In the CPU Load and CPU Temperature panels the value of every device is shown as a separate series in all the views. In the CPU Usage, System Load and CPU Cores panels the combining depends on the view: while Pie Chart and Line Chart average the value of the two devices, Digital, Bar Chart, Stat Table and Arc Gauge show every device separately. To see the values of a single device, the source selection in the settings is used; the options are the names of the two devices and
Both devices. - The counters are read in two-second rounds. The measurements that feed this tab are taken from the device once every two seconds; CPU Usage, CPU Load and CPU Cores are written into the per-second archive, and System Load and CPU Temperature into the fifteen-second archive.
- The fine-resolution history is limited. CPU Usage, CPU Load and CPU Cores are kept for the last one minute at the per-second resolution and for the last one day at the five-second resolution. In all of the five panels the fifteen-second resolution goes three days back, the per-minute resolution 14 days, the five-minute resolution 60 days, the fifteen-minute resolution 180 days and the hourly resolution two years; the older ranges are stored with a coarsening resolution for up to ten years.
- The appearance of an empty panel changes according to the card type. If there is no value in the selected time range,
No-datais displayed in the Pie Chart, Bar Chart and Arc Gauge views andNo datain the Digital view; if a measurement has never been recorded, the textNo data recorded for the selected time range.is printed inside the card. - The Header Bar widgets read the same definitions. The CPU, Load and CPU Temp widgets in the Header Bar are fed from the definitions of the CPU Load, System Load and CPU Temperature panels respectively; for the detail, see Header Bar.
- There is no passing to another screen from the cards. Clicking a core bar or a series name only hides or shows that series; no link to another screen is offered for detailed processor measurements.
