[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Much Slower than 7.2?



Ronald W. Heiby wrote:

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello psyche-list,

I've got a large data compilation application. It takes in a very
large quantity of text data and produces a resulting binary file that
is about 1.3 GB. In the mean time, it holds on the order of 2 GB of
data in memory while operating on it to produce the output file.

Under Red Hat 7.2 with the 2.4.9-31 kernel, my data compiling machine
require about 6 hours and 50 minutes to do this operation.

Under Red Hat 8.0 with the 2.4.18-18.8.0 kernel, the same machine
requires just over 2 DAYS to perform the same operation.

According to /proc/meminfo, the machine has a total memory of
791080960 bytes. So, I am not surprised that I need to do a fair
amount of paging. In addition to a 2 GB swap partition, I created a
pair of 2 GB swap files, to handle any overload. But, as I recall,
this was the case under 7.2 as well.

I have not made any substantial changes to the code that performs the
data compile. I was wondering whether there had been some significant
change in the memory management system or something else that would
account for this.

Also, I just noticed that another RH 8.0 system in my office, set up
by someone else, has a 4 GB swap partition. The mkswap(8) man page
still says that the maximum useful size for the i386 architecture is
about 2 GB. Is the man page out-of-date, or are we wasting 2 GB of
disk space? The /proc/meminfo and /proc/swaps seem to be indicating
that all of it is there and available. ???



On the x86 arch you can't have a swap device more than 2G. You can have a goodly number of swap devices. (The max is defined in the kernel source.) Are you certain the issue is memory related? Have you tested things to see if your disk io is as fast as it use to be? ` hdparm -t /dev/$drive`

--
There is no such thing as obsolete hardware.
Merely hardware that other people don't want.
(The Second Rule of Hardware Acquisition)
Sam Flory  <sflory rackable com>







[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]