advise on future of glibc (mktime issue)

Jairo19@interhosting.us jairo19@interhosting.us
Tue Jun 7 14:36:00 GMT 2005


    Hola:

   This seems to be the right mailing list for this message, please let 
me know otherwise.

   We recently ran into some problems, and their root lies in glibc.

   I will ask my questions first, then I will elaborate on the problem.

   1.-  What is the relationship between RedHat and glibc (besides 
hosting the project)?,
   2.-  do RH patches make it to the general glibc releases ?, does this 
mean that eventually other Linux distributions will get those changes ? 
(I think the answer to the second part is almost an obvious YES, but I 
have to ask)
   3.- do other developers besides redhat's have the chance to 
review/approve RH's changes before the make it to the main tree ?


   We found our problem when migrating from RH7.3 to RHEL4 and FC3. Our 
application slowed down significantly, a process that takes 1 day in 
RH7.3 now takes 3-4 days in FC3 and RHEL4.

    After debugging I see the problem shows up because we call millions 
of times mktime (no way to avoid that for us). The patched glibc (I 
think by RH) has mktime checking for changes on the system's timezone, 
this is done by checking the file: /etc/localtime which is either a link 
or a copy of the local timezone file from /usr/share/zoneinfo, this is a 
call to the file system via stat64(), hence our program is accessing 
millions of times the file system and crawling in pain. The old glibc, 
used in RH7.3 and used in Suse Enterprise 9 does not check for changes 
to this file with every call to mktime.

  I was able to eliminate the checks by temporarily setting TZ to GMT, 
but as I understand that is no longer suggested/supported, and RH7.3 was 
still faster.

  I contacted RH and they say they will not change the behavior of their 
glibc because they prefer "correctness over performance" and they 
believe glibc should be checking for changes to the timezone. Their 
suggestions are: build our own glibc and loose RH support, rewrite 
mktime for our own use (I tried that but seemed a little complicated and 
looses the benefit of the library per se), and lastly they suggested to 
set TZ=/etc/localtime (which makes the program run a little faster, but 
not as fast as RH7.3).

  It is widely known that FC1,2,3 and RHEL* are slow systems, and I 
always figured it was for the better, then I question, how may of this 
little changes in glibc exists in RH's patches that by themselves may 
not be a big performance hit, but put them together and you get a slower 
system ?

   So our solution could be to migrate to another Linux platform, but if 
the changes that RH made are going to show up in the library anyway, we 
would be just delaying the problem.

  My little and simple program that shows the performance hit (I'm 
leaving all the includes that my main program uses):

#include <string>
#include <vector>
#include <iostream>

#include <cstdio>
#include <cstdlib>
#include <cmath>
#include <cctype>
#include <ctime>
#include <sys/time.h>

using namespace std;

int get_dayofweek(int date)
{
  // extract day/yr/mn
  int day = date % 100;
  int month = (date % 10000) / 100;
  int year = date / 10000;

  struct tm time;
  time.tm_sec = 0;  time.tm_min = 30;  time.tm_hour = 9;  time.tm_mday = 
day;  time.tm_mon = month - 1;
  time.tm_year = year - 1900;

  time_t time1 = mktime(&time);

  return(localtime(&(time1))->tm_wday); // Sun=0, Sat=6

} //get_dayofweek

/////////////////////////////////////////////////////
// main routine
int main(int argc, char *argv[])
{
  int today = get_dayofweek(20050115);
  for(int i=0; i<1000000;++i)
    today = get_dayofweek(20050115);

  return (1);
}

I recompiled it and time it in the same HW first running RH7.3 then 
running FC3, SUSE Ent9 and RHEL4.

Thanks in advance for your comments.



More information about the Libc-alpha mailing list