The ESXi Host details page display the problem and event history of the selected host. To help you assess the health of the selected host, the following performance metrics are captured and displayed:
- Disk (storage I/O)
- NIC (network)
Dynatrace displays the percentage of actively used CPU of the ESXi host. Active CPU is approximately equal to the ratio of used CPU to available CPU.
A CPU measurement of 100% means that all CPUs on the host are utilized. So if a four-CPU host were running a virtual machine that has two CPUs, and active CPU usage is 50%, then you would know that the host is utilizing 100% of the available resources of two physical CPUs.
Memory health includes:
Memory usage: Percentage of total RAM used by processes.
Memory compression rate: Memory compression saves memory but requires additional CPU cycles.
Memory swapping: Rate at which memory is swapped between disk and active memory.
Disk health includes:
Throughput: Total number of bytes read and written to the disk per second.
IOPS: I/O (read/write) operations per second.
Read/Write latency: Time from I/O request submission to I/O request completion. This is the average delay of disk read/write operations in milliseconds.
Command queue latency: The average time spent in VMKernel queue per SCSI command.
Aborted commands: The number of aborted SCSI commands. This value serves as the basis for ESXi host storage overload incidents.
Number of datastores: Number of datastores connected to your vCenter. Includes metrics for free and used space.
NIC (network) health
NIC health includes:
Traffic: Average transmitted/received network traffic throughput.
Dropped packets: Number of received/transmitted packets dropped during collection interval. When packets are dropped by your network, they need to be retransmitted. Retransmission has a negative impact on network performance. Where it may be okay for a network to drop 0.1 percent of packets, it is not acceptable for an ESXi host network interface card to drop packets.