vHost performance details
Home > Select Project > Virtual environment > vHost performance details
Once you have found a host that looks abnormal in the vHost performance summary, the next question is "which resource on that host, and because of which VM?" vHost performance details lets you choose one ESXi host and view the CPU, memory, disk, network, and datastore trends as time series, and check the contribution and configuration information of each GuestVM on the same screen. Each chart comes with an interpretation guide that explains how to read the values. It is intended for operators who need to narrow down the cause of performance degradation to the host level in a virtualized environment.
Screen layout
The screen is divided into three areas.
| Area | Description |
|---|---|
| Main charts | Shows five resource types of the vHost as time series. Click a card to open the detail charts |
| GuestVM table | Metrics per guest VM running on this host |
| vHost information tabs | Three tabs: vHost information summary, vHost inventory, and vmGuest inventory |
The five main charts are as follows.
| Chart | Description |
|---|---|
vHost CPU Usage | CPU usage of the whole host |
vHost Memory Usage | Memory usage of the whole host |
vHost Disk Usage | Stacked chart of disk reads and writes |
vHost Network Usage | Stacked chart of network receive and send |
vHost Datastore Usage | Usage per datastore. One line per target |
Why the total is not drawn separately in stacked charts
The disk and network charts stack reads and writes (receive and send). Drawing the total value sent by the agent here as well would count the same amount twice, so it is not included. Check the total from the height of the stack and the Total value in the tooltip.
Opening the detail charts
Clicking a main chart card opens a drawer with nine detail charts for that resource. For example, clicking the CPU card lets you view CPU Usage, CPU Demand, GuestVM CPU Usage, CPU Latency, CPU Ready, CPU Co-Stop, CPU Active vCPU, CPU Active Overcommit Core, and CPU Allocated Overcommit Core together.
While the drawer is open, real-time refresh of the main charts behind it stops. This keeps two screens from querying the same range at the same time.
Reading the interpretation guide
The drawer also contains an interpretation guide that explains how to read the metrics. The guide is grouped by the following topics for each resource.
| Resource | Guide sections |
|---|---|
| CPU | Basic load judgment · Contribution per VM · Contention · Scheduling · Capacity · Overcommit · Overall judgment · Observation tips |
| Memory | Basic usage judgment · Memory reclamation stages · Contribution per VM · Overcommit · Capacity · Overall judgment · Observation tips |
| Disk | Basic load judgment · Where the latency comes from · Saturation · Confirming the bottleneck · Contribution per VM · Overall judgment · Observation tips |
| Network | Basic load judgment · Errors · Drop diagnosis · Contribution per VM · Overall judgment · Observation tips |
| Datastore | Basic load judgment · Diagnosis by latency layer · Capacity · Thin Provisioning · Overall judgment · Observation tips |
The guide also states the caution and critical criteria for each metric. For example, it explains that CPU Ready is considered caution between 5% and 10%, and critical above 10%. Treat the guide text on the product screen as the authoritative source for threshold values.
Overcommit metrics are percentages
Metrics that express an "allocation multiple against physical capacity", such as CPU Active Overcommit Core, are displayed as percentages. Physical capacity is treated as 100%, so the value can exceed 100%.
Query range and refresh interval
- When you first enter the screen, the most recent 30 minutes are queried.
- The real-time refresh interval is 30 seconds.
- Drag a specific area of a chart to zoom into that range. The smallest range you can zoom into is 60 seconds.
The resolution changes beyond three hours
Ranges of three hours or less are queried at raw resolution, and wider ranges come down as 5-minute aggregates. To check a brief spike, narrow the range to within three hours.
vHost information tabs
The vHost information summary tab shows the main configuration information of the selected host: host name, ESXi management IP, vCenter address, datacenter, cluster, model, ESXi version, overall status, total memory, number of guest VMs, number of datastores, and uptime, in that order.
The vHost inventory and vmGuest inventory tabs show the same information in inventory table form. Clicking the View in VM inventory button on each tab takes you to the corresponding tab of the VM inventory screen.
Uptime is shown in days
Uptime is the value recorded when the inventory was collected, so the number does not move immediately even though the screen refreshes every 30 seconds. Showing hours and minutes would expose this gap, so it is shown in days only.