lkml.org 
[lkml]   [1999]   [Apr]   [19]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
Subjectkflushd and/or update: Performance Killer?
Hi,

as stated in the subject, i'm looking for some answers - explanations -
pointers to the source maybe concerning a mysterious performace loss i
experienced.

first i noticed some weird events, such as waiting several seconds for
characters typed to be echoed to the screen, extremly long load times
for normally fast applications like vi, ls etc. and at least repeated
error messages from fetchmail about timeouts (client server sync
failiure).

my best first bet was that i had hosed a self written programm when i
thought to improve it, because i'm running it 24 hours a day. but
checking it for hidden loops or memory leakage proved negative so i
started to investigate closer.

soon there were some common denominaters fixed that occured any time
i experianced that performance loss:

top and/or free showed the amount of memory used for buff seemingly
at the lowest possible values and stated extreme high idle-times
80% and more for the cpu, where i was used to see my self written
programm drawing 90% and more cpu-capacity.

next i had a look at the processes shown by top and noticed a steep
increase of usage by the processes kflushd and/or update. where i was
used to see accumulated time on them of fractions of seconds during a
whole day, now top showed 10 and more seconds runtime after one hour.

for now (getting the idea after reading the LDP book -tlk the linux
kernel) i have just found one workaround. whenever i experiance such
a situation i just start a seemingly senseless tar-command that will
then allocate more memory to the buff-space and kflushd/update are
called less frequently.

i dont know of others facing this phenomen, but i was able to reproduce
it on my computers (a 100 mHz Pentium and a 400 mHz PentiumII) with
different kernels 2.0.34, 2.0.36 and 2.2.5.

hope for some answers

Helmut

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:51    [W:0.042 / U:3.644 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site