login
Search
I built two different releases of the 2.6 kernel using the same config file, a 2.6.6-rc2 and a 2.6.8.1. I realised that with 2.6.8.1 under moderate hardrive load, the system is laggy. Under heavy hardrive load, the system almost freezes. These symptoms are not present when using the 2.6.6-rc2 kernel. Could this be due to significant IDE changes between these kernels?
sort of OT =)
What about the performance of their 2.5/2.6 ide code ?
I remember getting ~45+ megs from hdparm and now i only get ~30-...
I have tweaked the udma mode (X69) and enabled dma (d1)... and played around with u1, c1.. m16... a16...a4096... you name it!
It's a disgrace ;-)
from 45+ to 30-..
This is performance; these are measurements that you can also get with dd when it reports the speed; it's been like this since some 2.5.x kernel.
Not even if i disable preemption does this goes away... not even if i pick different schedulers.. and i think it doenst change either when i go from io-apic/acpi to the normal xt-pic.
---------------------------------
1. HDParm; DD.. same result
2. Not fixed since 2.5.x
3. Many options tested... problem persists
4. I have an 80 wire cable; detected as such.
please ignore me =)
----------------------------------------
t00r@tuesday [06:03:19] [~] sudo hdparm -t /dev/hd[acegi] | grep iming
Timing buffered disk reads: 84 MB in 3.05 seconds = 27.53 MB/sec
Timing buffered disk reads: 86 MB in 3.03 seconds = 28.41 MB/sec
Timing buffered disk reads: 84 MB in 3.03 seconds = 27.75 MB/sec
Timing buffered disk reads: 74 MB in 3.04 seconds = 24.31 MB/sec
Timing buffered disk reads: 82 MB in 3.02 seconds = 27.16 MB/sec
THESE PEAKED AT ~45+ BEFORE -- UNDER 2.4 KERNELS.
Now they hit this 30 meg ceiling... makes me wanna go back to 2.4...
2.4 v. 2.6 speed testing with hdparm
I concur. I roll my own, and since switching to the 2.8 line from 2.4 I have seen reduced transfer rates reported by hdparm.
Tracking down
After taking each bk patch, applying it to the kernel, compiling, booting the kernel and testing performance, I have narrowed down the laggy system symptom to changes made in patch-2.6.7-bk6 which is available at http://www.kernel.org/pub/linux/kernel/v2.6/snapshots/old/patch-2.6.7-bk...
I am not an expert programmer and I am faced with over 100 thousand lines of code. So far, I am pretty sure that it's not due to Nick Piggin's new IO scheduler, as I've chosen other schedulers at boot, without any improvement. Any ideas/help would be greatly appreciated
Same problem
I had what sounds like the same problem when I first compiled 2.6.8.1-ck4 and chose high mem support (4G) and the page tables in high mem option. My system was useable until i got into X and started a compile. The system was lagging so badly it appeared to be frozen. I went back and disabled high mem completely when i noticed the new option for systems with only 1gb of ram. I have not experienced any further problems. Don't really know a whole lot about any of this stuff but I hope it narrows down the problem.
I can confirm that. The hardd
I can confirm that. The harddisk throughput of my laptop deteriorated from 28 mb/sec (2.6.7) to 22-25 mb/sec. It would be nice if this issue was fixed soon!! This should be top priority for the next release.
I've had x freeze, but that was early 2.6.7-bk's, later is ok
2.6.8/.1 has problems with cd-writing, so tried upgrading to 2.6.9-rc1-bk? which made for a more responsive system, but man app stopped working.
So, back to 2.6.7-bk21 and:
hdparm -tT /dev/hda
/dev/hda:
Timing buffer-cache reads: 2656 MB in 2.00 seconds = 1326.21 MB/sec
Timing buffered disk reads: 150 MB in 3.01 seconds = 49.87 MB/sec
/dev/hda:
Timing buffer-cache reads: 2624 MB in 2.00 seconds = 1310.23 MB/sec
Timing buffered disk reads: 150 MB in 3.00 seconds = 49.99 MB/sec
This on a A64-3200 with 1GB DDR400, ATA Western Caviar 200GB 8MB Cache, Asus K8vSE, slackware 10 with 2.6.7-bk21 and highmem = y, pagetable = y