Historical Data Related To CPU,IO And Memory

Home » CentOS » Historical Data Related To CPU,IO And Memory
CentOS 9 Comments


Are there any utilities or tools to look at historical data of Memory, CPU
Utilization or IO activity on CentOS 6.4 or 5.9 Version? For example the how much the memory was consumed for the period of last six months.



9 thoughts on - Historical Data Related To CPU,IO And Memory

  • It might be overkill for a single box, but tools like OpenNMS will collect this info from any number of targets via snmp and let you graph the history up to a year back.

  • You could take a look at sysstat and the graph with whatever tool you wish. I use kSar to plot.

    By default, the sysstat output is kept for a week, but you could write scripts to keep the information for as long as you wished.

    sysstat rpm on CentOS 6.4: sysstat-9.0.4-20.el6.x86_64
    kSar web site: http://sourceforge.net/projects/ksar/

    . . . just my two cents

  • Back in March we had a thread on monitoring software [0]. You might find a solution in that thread that better suits your environment.

    [0] http://lists.CentOS.org/pipermail/CentOS/2013-March/133381.html

    And there are other trend graphing applications like Cacti that will give you the same sort of historical data. Data points are also collected via SNMP, but if you write custom scripts Cacti can use you’re not limited to just things exported via SNMP.

    [But extensibility via scripts is common to most open source software and exporting data points via SNMP is a best practice.]

  • From: Kaushal Shriyan

    The most lightweight, if you do not mind scripting a little bit, would be a “simple” cron script calling rrdtool. You can find many rrdtool examples online.


  • Opennms is very extensible too – and includes support for collecting metrics via wmi, xml, jmx, etc. But it is a big framework and fairly complicated to extend (but a lot works out of the box after a simple yum install). Something like collectd would be more lightweight.

  • you don’t even need to write scripts, the sar data collector, sadc, can be configured to collect as little/much data as you want, for whatever interval and period you wish simply by tweaking a few configure file settings. See /etc/sysconfig/sysstat

    but, if you really want historical data going back months, I do think I’d look into something like cactii, which is built on RRDtool, which replaced the legacy mrtg. This will maintain a set of web pages with hourly/daily/weekly/monthly stats of as many parameters as you wish to track going back as far as you like.

  • For anyone who doesn’t know about rrdtool, it is the ’round robin database’ that holds time series data with long histories in a small file fixed-size by having a certain number of data points representing ‘recent’ times where it holds the full-resolution sample values (normally converted to a rate), and as the samples age they are averaged into larger timespan samples so you lose granularity to save space since you usually are more interested in the nearby times. And it includes tools to draw graphs out of the stored data. OpenNMS can either use the standard native-code rrdtool library or a pure-java reimplementation called jrobin.