Latency
The latency tab gives you a drilled down detail of various measured latencies in your performance test.
The third tab in the result analysis tab shows you various values of latency throughout the test duration. There are three main latencies which are displayed:-

Message Generation Latency

How much time it takes on average to run the message function? This chart displays the minimum, maximum and average value in form of an eye chart. Here is an example of a latency chart. Each step on the x axis is an iteration, and chart shows three value for each iteration.
From the chart, it is clear that while the latencies are well under control upto iteration 24, the Max latency suddenly increase from iteration 25 and remains elevated until the last iteration. Similarly, the average latency is increasing after iteration 35 and remains high until the last iteration. What could explain this behavior?
In this example, we were doing a POST call within the message function to an external server. It appears that server was taking a longer time to accept the POST request with time, therefor increasing the time it takes to complete the message function. This is an important observation, as a higher message generation latency will likely slow down the overall performance of the test.

Packet Sending Latency

This latency graph shows the average, minimum and maximum time it took to send a message to the back-end. This is a true measure of your server performance, as it doesn't count the time it took to generate the message. As we could see, the packet sending latency is fairly maintained consistent with time.
Packet sending latency is highly dependent upon the protocol and transport. E.g. for MQTT QoS 0, the packet sending latency will be lower, as the transmission of packet is marked completed almost immediately. If the QoS value used is 1 or 2, the sending latency will be much higher for the same test, as higher QoS values requires every sent message to be acked for deliver, therefore considerably increasing the latency.

Subscription/Response Handler Function latency

This chart shows the min/max/average value for the response handler function execution. It is good to keep the latency low for the response handler in general. If there was no invocation of the response handler, the chart will be empty.
Last modified 1yr ago