In VMware Aria Operations for Networks, you can get visibility into the latency, set threshold to get notified when the latency violates the configured threshold limits, and identify the root cause of the high latency.

In VMware Aria Operations for Networks, you can get details about the following latency-related metrics:

Types

Descriptions

Access the latency metrics

vNIC to vNIC latency

The time taken by the packet to travel from the source VM’s vNIC to the destination VM’s vNIC on the same ESXi host or across ESXi hosts.

See Latency Metrics.

To monitor these metrics, you can do the following:

- Create a custom dashboard to monitor these metrics

- Set a threshold limit to get alert when there is a threshold violation.

- Create search-based alert for the metrics that are not supported in threshold metrics.

vNIC to PNIC and PNIC to vNIC latency

The time taken by the packet to travel from the source VM’s vNIC to the PNIC of the host transport node of the source VM.

PNIC to PNIC latency (VTEP to VTEP latency)

The time taken by the packet to travel from the PNIC/VTEP of the host of the source VM to the PNIC/VTEP of the host of the destination VM over the underlay network connecting the two hosts.

TCP RTT & Retransmission counts for IPFIX Flows

NSX-T DFW IPFIX flow records include additional information about the TCP Round Trip time for establishing the flow and the TCP packet retransmission count to compensate for the high RTT or drops. These are in the form of retransmissionCount and averageLatency fields associated with the flow between a pair of IP addresses. vRNI enriches the flow information with VM, Hosts, Applications, E-W/N-S nature of traffic etc. and aggregates them and presents the max and average values for these metrics in association with the flows.

Troubleshoot Latency

When you gain visibility on the latency metrics, you find the root cause of the problem and solve the latency problem.

Let's explore how you can troubleshoot latency using VMware Aria Operations for Networks:

  1. Run the following search query to get the top 10 vNIC-vNIC pairs with high latency: netopa.interfacePairLatency.absolute.maximum.microsecond, netopa.interfacePairLatency.absolute.average.microsecond,vm1, vm2, virtualInterface1, virtualInterface2 of interface pair info where host is set and virtualInterface1Type != VMKNIC and netopa.interfacePairLatency.absolute.maximum.microsecond > 0 order by netopa.interfacePairLatency.absolute.maximum.microsecond limit 10

    The search query lists the top 10 vNIC-vNIC pairs with high latency.

    The overall vNIC-vNIC Latency consists of the following components:

    • Source VM to source PNIC (vNIC - PNIC)

    • Source VTEP to destination VTEP (VTEP - VTEP)

    • Destination PNIC to destination VM (PNIC - vNIC).

  2. Identify which of these three components is responsible for the high vNIC-vNIC latency.

    This helps to pinpoint the component responsible and investigate further.

    To determine the latency between the above-mentioned components, we need to run the following queries:

    • Source VM to source PNIC (vNIC - PNIC): netopa.interfacePnicLatency.absolute.maximum.microsecond, netopa.interfacePnicLatency.absolute.average.microsecond, pnic, virtualInterface, vm of interface pnic pair info where host = '<source host name>' and vm = '<source VM name>'

      or

      vNIC PNIC Latency, max vNIC PNIC Latency, pnic, virtualInterface, vm of interface pnic pair info where host = '<source host name>' and vm = '<source VM name>'

      You get the latency of the first component.

    • Source VTEP to destination VTEP (VTEP - VTEP): netopa.interfacePairLatency.absolute.maximum.microsecond, netopa.interfacePairLatency.absolute.average.microsecond, virtualInterface1, virtualInterface2, host1, transportNode1, host2, transportNode2 of interface pair info where virtualInterface1Type = VMKNIC and host1 = '<source host name>'

      Tip:

      You get the source host and destination host names by running the initial query to get the vNIC-vNIC latency.

    • Destination PNIC to destination VM (PNIC - vNIC): netopa.pnicInterfaceLatency.absolute.maximum.microsecond, netopa.pnicInterfaceLatency.absolute.average.microsecond, pnic, virtualInterface, vm of interface pnic pair info where host = '<destination host name>' and vm = '<destinationvm name>'

      Or

      PNIC vNIC Latency, max PNIC vNIC Latency,pnic, virtualInterface, vm of interface pnic pair info where host = '<destination host name>' and vm = '<destinationvm name>'

      Based on the result of the above queries you can determine which component is responsible for high vNIC-vNIC latency.