2013-07-26 11:42:45

by Jingbai Ma

Subject: makedumpfile 1.5.4 + kernel 3.11-rc2+ 4TB tests

Hi,

I have run some tests with makedumpfile 1.5.4 and upstream kernel
3.11-rc2+ on a machine with 4TB memory, here is testing results:

Test environment:
Machine: HP ProLiant DL980 G7 with 4TB RAM.
CPU: Intel(R) Xeon(R) CPU E7- 2860 @ 2.27GHz (8 sockets, 10 cores)
(Only 1 CPU was enabled the 2nd kernel)
Kernel: 3.11.0-rc2+ (at patch b3a3a9c441e2c8f6b6760de9331023a7906a4ac6)
crashkernel=384MB
vmcore size: 4.0TB
Dump file size: 15GB
All measured time from debug message of makedumpfile.
As a comparison, I also have tested makedumpfile 1.5.3.

(all time in seconds)
Excluding pages Copy data Total
makedumpfile 1.5.3 468 1182 1650
makedumpfile 1.5.4 93 518 611


So it seems there is a great performance improvement by the mmap mechanism.

--
Thanks,
Jingbai Ma

2013-07-26 13:24:15

by Vivek Goyal

Subject: Re: makedumpfile 1.5.4 + kernel 3.11-rc2+ 4TB tests

On Fri, Jul 26, 2013 at 07:42:39PM +0800, Jingbai Ma wrote:
> Hi,
>
> I have run some tests with makedumpfile 1.5.4 and upstream kernel
> 3.11-rc2+ on a machine with 4TB memory, here is testing results:
>
> Test environment:
> Machine: HP ProLiant DL980 G7 with 4TB RAM.
> CPU: Intel(R) Xeon(R) CPU E7- 2860 @ 2.27GHz (8 sockets, 10 cores)
> (Only 1 CPU was enabled the 2nd kernel)
> Kernel: 3.11.0-rc2+ (at patch b3a3a9c441e2c8f6b6760de9331023a7906a4ac6)
> crashkernel=384MB
> vmcore size: 4.0TB
> Dump file size: 15GB
> All measured time from debug message of makedumpfile.
> As a comparison, I also have tested makedumpfile 1.5.3.
>
> (all time in seconds)
> Excluding pages Copy data Total
> makedumpfile 1.5.3 468 1182 1650
> makedumpfile 1.5.4 93 518 611
>
>
> So it seems there is a great performance improvement by the mmap mechanism.

[CC Andrew].

Thanks Jingbai for testing results. Good to see that vmcore mmap() patches
are helping.

Thanks
Vivek