Re: What tuning can be done on disk writes?



Enable a 10046 trace before you run this script again. (Preferably on
both databases).

This will tell you whether your waits are for 'log file parallel write'
or 'log file sync' or some other I/O wait.

This will narrow your investigation down a bit.

Enable it with:

SQL> alter sesssion set event '10046 trace name context forver, level
8';

then run script..

Disable it with

SQL> alter sesssion set event '10046 trace name context off';

The trace file will be in the udump directory... run it through tkprof
with waits=y and then post your findings...

Matt

.



Relevant Pages

  • Re: Possible to change name of trace.axd in web.config file?
    ... > easier to simply write to your application's own log file for users to ... I'm writing an application that is writing trace information ... >> links are actually to the trace.axd webpage instead of using the name I ...
    (microsoft.public.dotnet.framework.aspnet)
  • Re: Possible to change name of trace.axd in web.config file?
    ... > easier to simply write to your application's own log file for users to ... I'm writing an application that is writing trace information ... >> links are actually to the trace.axd webpage instead of using the name I ...
    (microsoft.public.dotnet.framework.aspnet.security)
  • Re: ETW trace question
    ... SEH or vectored - and call flush from there. ... > If my user app crashes when logging to a ETW tracing session my log ... the log file is never written ... > until the trace session is stopped. ...
    (microsoft.public.win32.programmer.kernel)
  • Re: Possible to change name of trace.axd in web.config file?
    ... must make sure that the new extension maps to an extension that ASP.NET ... > easier to simply write to your application's own log file for users to ... I'm writing an application that is writing trace information ... >> links are actually to the trace.axd webpage instead of using the name I ...
    (microsoft.public.dotnet.framework.aspnet)
  • Re: LTT user input
    ... and to that end have taken the existing trace infrastructure ... script for that event (if there's a handler defined for the event ... This gives the script a chance to do whatever Perlish thing it ... syscall totals to individual syscall totals for each pid. ...
    (Linux-Kernel)