Longtime .hope u guys had a nice day.Well if you remember that scenario form Redhat,not bad for the Pros too, can really help on your production boxes immediately you try it.I had some tunings already on my production servers and really they do work give a try but with caution make sure you're on the console otherwise you might fly miles :-)
http://www.redhat.com/sundown/techquestion0517.html

--
***************************************************************************
 / ''We can't become what we need to be by remaining what we are''\
 \ ,,                                                           ,,/
***************************************************************************

Title: redhat.com | Home

IBM and Red Hat Solaris Migration Center

Tech Challenge May 17

Winner: Roger Seip

The Scenario

The sysadmin, the DBA, and the end user all want the same thing: a fast, secure, and reliable system. This week's question is going to address the "fast" component. Red Hat Enterprise Linux is a remarkably flexible operating system that lends itself to customization and performance tweaking. This flexibility assures the users that they get even more industry-leading performance out of applications like Oracle, Sybase, BEA, or IBM Websphere.

The Parameters

  1. No purchasing of additional hardware or software (you have used up your budget for the quarter/year/ or your boss's life expectancy).
  2. You are using Red Hat Enterprise Linux AS on x86 hardware.
  3. You have a fresh installation of Red Hat Enterprise Linux AS and a freshly installed Oracle, Sybase, BEA, or IBM Websphere application.
  4. You have already disabled or uninstalled the daemons and services that you deem unnecessary for this server.
  5. The four areas you are interested in tuning: kernel parameters, memory subsytem, file system, and networking.
  6. No recompiled kernels or anything that would void your Service Level Agreement with Red Hat Global Support Services. This does not include third-party drivers required for application functionality or hardware support.

The Question

  1. When running a top command, you notice that the iowait column seems exceedingly high. What does this mean and how can it be fixed?
  2. When tuning the memory subsystem, what are four relevant VM parameters that the administrator can adjust and monitor?
  3. A little bird told you to use ReiserFS or XFS as a faster filesystem than ext3. How can you increase the performance of an ext3 filesystem to meet (or sometimes beat) these two unsupported filesystems? Additionaly, what are the tradeoffs involved? Assuming that you already have high-quality backup power supplies.
  4. Your server gets a lot of incoming TCP connections from multiple sources. Many sockets seem to be stuck in TIME-WAIT. You wish to reuse these resources. What combination of valid sysctl commands allow the reuse and recycling of these resources?

The Answer

  1. Iowait means that the CPU waits for I/O block completion and it is a performance bottleneck.
  2. Administrator should play with
    1. /proc/sys/vm/ inactive_clean_percent
    2. /proc/sys/vm/ pagecache
    3. /proc/sys/vm/ max-readahead
    4. /proc/sys/vm/ overcommit_ratio
  3. journaling option data="" best performance should be with data="" but I never noticed a strong improvement
    Possibility to turn the no atime flag on the system. Tradeoff is journal FS that allows faster remounts for very little performance degradation
  4. Tune with sysctl:
    1. /proc/sys/net/ipv4/ tcp_tw_recycle
    2. /proc/sys/net/ipv4/ tcp_tw_reuse
    3. /proc/sys/net/ipv4/ tcp_max_syn_backlog
    4. /proc/sys/net/ipv4/ tcp_max_tw_buckets

>> return to the main page

_______________________________________________
LUG mailing list
[email protected]
http://kym.net/mailman/listinfo/lug
%LUG is generously hosted by INFOCOM http://www.infocom.co.ug/

Reply via email to