This section covers the analytics that help you assess the virtual service performance metrics. The charts and metrics reflect the selected display time.

The virtual service analytics comprises the following sections:

End-to-End Timing

The end-to-end timing provides an overview of the quality of the end-user experience and the occurrence of any slowdowns. The chart breaks down the time required to complete a single transaction, such as an HTTP request. It breaks total transaction time into the following components:

Field

Description

Client RTT

Average network TCP latency between the client and NSX Advanced Load Balancer for all clients, both local (within the same datacenter) and remote (Internet). This metric indicates the duration it takes to establish connections and return acknowledgments. This number will often be higher than an ICMP ping.

Server RTT

Round-trip latency for SE-to-server traffic. An abnormally high server RTT indicates that the network is saturated or that a TCP stack cannot quickly establish new connections due to heavy load.

App Response

The time taken for the server to respond. It includes the time the server took to generate content, potentially fetch backend database queries, and begin transferring the response back to the NSX Advanced Load Balancer. This metric is only available for a L7 virtual service.

Data Transfer

Average time required for the server to transmit the requested file. It might vary greatly depending on the size of objects requested and the latency of the server network. The larger the file, the more TCP round trip times are required due to acknowledgments (ACKs), directly impacted by the client and server RTT. This metric is only used for a L7 virtual service.

Total Time

Request-response time.

This is the most critical end-to-end timing to watch, because it is the sum of the other four metrics. If the Total Time is consistently low, the application is probably serving traffic successfully.

Metrics Tiles

Metrics tiles continuously display time-averaged data. Clicking a particular metric tile will display the corresponding data over time in the chart. The following are the available metrics tiles:

Tile Name

Description

End-to-End Timing

Charts the end-to-end timing by default. It governs what appears in the chart display. It displays a color-coded plotted timing to indicate the client RTT, server RTT, data transfer, and app response times at any recorded instant.

Throughput

Total bandwidth passing through the virtual service in Mbps. Pointing over this graph displays the throughput in Mbps for the selected time. Throughput is measured as bytes transferred between the client and SE, not including data transferred between SE and the servers.

Open Connections

The number of TCP client connections or UDP transactions currently in an open state. UDP transactions are counted even though they are technically connection-less. The number of open connections does not necessarily translate into the number of unique clients, as protocols such as HTTP 1.1 typically open six connections per client browser. If the Maximum Concurrent Connections setting has been set for a virtual service, a horizontal red bar superimposed over the chart pane will show the maximum number of connections allowed. For more information, see Rate Shaping and Throttling Options.

Conns

The average rate of connections completed per second. The tile also summarizes the percentage of connections resulting in errors. Pointing over the tile shows the rates of good, lossy, and bad connections.

  • Lossy connections experience issues such as retransmissions, zero window size events, or high connection setup times.

  • Bad connections are those which terminate ungracefully. Connection closed abnormally.

  • Requests ended abnormally will result in bad connections.

Requests

Number of responses to requests per second. It breaks down successful requests versus errors (such as 4xx or 5xx errors).

Within the chart pane, you can toggle radio buttons to break down the errors by whether they are generated by the server or by the NSX Advanced Load Balancer. For instance, if no servers are available in the pool, NSX Advanced Load Balancer will send out an HTTP 500 status code. The Client Logs section will provide more details on specific errors. This metric is only available for HTTP virtual services.

HTTP vs HTTP2 Requests

Displayed as requests-per-second rates.

For more information on the chart and overlays pane, see Chart Pane and Overlays Pane.