Hello guys. I think I've stumbled across something weird. My brother updated his machine today and since then he started to have horrible IO performance with kernel-2.6.13-1.1526_FC4. For example, transfering his stuff to my computer as a backup (we started to suspect the HDDs were going to stop working), the max speed with latest kernel was below 1MB/s. With 2.6.12-1.1456_FC4, transfer speed is over 3MB/s according to the ssh client I'm using right now. Not to mention the average responsiveness of the system (a ssh login with 1526 took about 10-20 seconds to show the login prompt, while with 1456 the prompt shows up immediately.
Any ideas on how I can test this to confirm if this is a bug or just faulty hardware? I'm going to run a full smart test on the discs tomorrow and a few RAM checks just in case, but the short test from smart passed without any problems and the smart reports shows no parameter outside the limits.
Thanks. Pedro Macedo
On Fri, Sep 30, 2005 at 11:43:51PM -0300, Pedro Fernandes Macedo wrote:
Hello guys. I think I've stumbled across something weird. My brother updated his machine today and since then he started to have horrible IO performance with kernel-2.6.13-1.1526_FC4. For example, transfering his stuff to my computer as a backup (we started to suspect the HDDs were going to stop working), the max speed with latest kernel was below 1MB/s. With 2.6.12-1.1456_FC4, transfer speed is over 3MB/s according to the ssh client I'm using right now. Not to mention the average responsiveness of the system (a ssh login with 1526 took about 10-20 seconds to show the login prompt, while with 1456 the prompt shows up immediately.
Any ideas on how I can test this to confirm if this is a bug or just faulty hardware? I'm going to run a full smart test on the discs tomorrow and a few RAM checks just in case, but the short test from smart passed without any problems and the smart reports shows no parameter outside the limits.
What kind of disk controller ? If IDE, is DMA enabled ?
It would be interesting to see the output of ..
<boot old kernel> dmesg -s 128000 >dmesg.old <boot new kernel> dmesg -s 128000 >dmesg.new diff -u dmesg.old dmesg.new
Dave
Dave, thanks for the quick response.. Looks like the testing is gonna need to be a little more extensive. After some time using the old kernel, the initial performance dropped to the same levels that 1526 was showing right after boot.
Dave Jones wrote:
What kind of disk controller ? If IDE, is DMA enabled ?
It's onboard IDE controller. DMA enabled on both disks (two seagate disks, using udma 5 , with 80 way cable). The controller is recognized by lspci as 00:11.1 IDE interface: VIA Technologies, Inc. VT82C586A/B/VT82C686/A/B/VT823x/A/C PIPC Bus Master IDE (rev 06)
It would be interesting to see the output of ..
I'll do some massive testing tomorrow to try to exclude the possibility of HW failure, since the fact that both kernels show the same results (even though 1526 shows it faster than 1456) makes me believe he may need to get either a new disk or a new motherboard.
[]'s Pedro Macedo