-1

I will be performing a distributed load test using JMeter. I am using the JMeter extras plugin to output some nice graphs but all of these graphs have to do with response times, response latency, throughput, etc. I want to also measure CPU, memory used/free, disk usage/latency, and network utilization, maybe some others.

I will be testing a web application that is running on Ubuntu 14.04.

What tools or commands can I use to gather these stats at various points during the load test and either output the raw data or averages?

Thank you for any information you can provide.

Ian Herbert
  • 1,071
  • 2
  • 16
  • 35

2 Answers2

0

Free and great for high level KPIs. Works within JMeter:
http://jmeter-plugins.org/wiki/PerfMon/

Free / Paid and great for detailed low level analysis (stand alone tool):
http://newrelic.com

We use New Relic ourselves and are very satisfied!

Ophir Prusak
  • 1,437
  • 13
  • 20
0

I am using Cacti for that, it is relatively easy to install and configure (on Centos it can be installed with yum from the EPEL repository). It uses snmp to get network, CPU, memory, load,..from the various target servers. To monitor disk io's there is a great template (https://github.com/markround/Cacti-iostat-templates), if you follow step by step their instructions it will work (at least on centos/red-hat).

What I like with cacti is that you can also define your own datasources, for example you can ask cacti to execute a shell script on your server that would parse your access.log (or any other application log files) and returns metrics like throughput (nbr requests, nbr bytes) or processing time,.. etc then you can get this plotted side by side with the devices utilizations metrics.

To set-up the whole think you will probably one day, it is not very intuitive how to define your own data sources for example. Also you have to enable snmp on the box, which is easy if you remove the whole /etc/snmp.conf and use the bare minimum. It is a great tool for capacity management.

Saule
  • 579
  • 5
  • 6