Document Upload Jmeter script is giving inconsistent response times when running in AWS servers

Question

We are trying to do a performance test for document upload scenario with a 50 MB file. When we are running the script in a local machine with 100Mbps internet speed, it is matching the response time of manual file upload. However, when we are trying to run the same script in AWS server by limiting the internet bandwidth of Jmeter to 100Mbps, the response time is not matching and is almost half of what we got in local machine.

For limiting internet bandwidth to 100Mbps, we are uncommenting below 2 lines in jmeter.properties file httpclient.socket.http.cps=13107200 httpclient.socket.https.cps=13107200

CPS calculation (100 MB = 102400 KB) So, CPS = 102400 * 128 = 13107200

The response time should be consistent in any machine.

Answer 1

1. Are you sure that your local test is hitting the limits of bandwidth? Consider measuring other operating system health metrics like CPU, RAM, swap, etc. usage. It can be done using JMeter PerfMon Plugin

2. It might be the case you use Java implementation of the HTTP Request sampler which doesn't support HTTP

3. Response time consists of:

   • connect time
   • latency (TTFB)
   • and time required for the remainder of the response to reach JMeter

   so it might be the case AWS instance establishes the initial connection/handshake faster and the document transfer itself happens at the same speed