Announcement

Collapse
No announcement yet.

Jmeter baseline - using the JMeter OX templates

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Jmeter baseline - using the JMeter OX templates

    Hi There

    We are implementing Open-Xchange in our envrionment (University). We are using Jmeter tot test our setup before putting it into production. The problem with testing is that there is no baseline available from OX concerning the testsresults. We tried a basic test (the sample test from OX) of 250 concurrent users, importing the standard vcal info and contactinfo in a "rampup" of 30 seconds. What we experienced was that our backend MySQL service did not perform as we expected it should. It did not crash but became unfunctional. After tweaking MySQL end OX some performance improvements were gained, but after 140 connections MySQL was not able to perform the actions which were necessary to complete te tests....

    The problem we have is the fact that there is no baseline data available from OX to use in the JMeter tests. For instance is a "ramp-up" of 30 seconds with 250 users a reasonable test (looking at our specs) or is it a "kill" scenario ? To give an indication of our setup, the following specifications are used :

    1. Our "frontend" servers consist of 2 HP-blades with 2x4core cpu's and 24GB of ram each, running the OX services. We only use the email service, no other groupware functions are active.
    2. Our "backend" servers consist of 2xHP Proliant servers with 1x4core cpu and 12GB of ram each. These 2 servers deliver the MySQl database service in a Master-Master setup using a HP Storageworks fiber cabinet.
    3. The Mailstore is hosted on shared storage using NFS.
    4. The setup will be used to host 50.000 mailboxes with a concurrent use of 300-400 connections during working hours.

    Hopefully the Jmeter OX templates tests are used in the field and maybe someone can answer our question concerning a reliable and viable baseline in conjuction with our setup. Your help will be fully appreciated !!!!

    Regards,

    Robert Tuit

  • #2
    Hello Robert,

    We are currently developing tests one should be able to compare against (results not expected in the near future). The default profile, without wait times, is a "kill" scenario. I recommend a general wait time of at least 30000ms after each loop.

    If further assistance is needed, please directly get in contact with us:


    Best,
    Holger

    Comment

    Working...
    X