[Lazarus] filesystem timing Linux vs Win

Sven Barth pascaldragon at googlemail.com
Sat Nov 27 19:04:49 CET 2010


On 27.11.2010 18:53, Henry Vermaak wrote:
> On 27 November 2010 17:22, Sven Barth<pascaldragon at googlemail.com>  wrote:
>> On 27.11.2010 18:16, J├╝rgen Hestermann wrote:
>>>
>>>> Calling the system to ask for the last-modification time that often
>>>> (even with all/most data cached by the OS) would take that long on
>>>> Windows, while under Linux it wouldn't even take a single second...
>>>
>>> But how does it come that there can be such a difference doing nearly
>>> the same things on Linux and Windows? I can't believe that Windows is
>>> *such* a bad design. They all cook with water I think.
>>
>> It would be interesting to see a comparison on the same filesystem. E.g.
>> fat32 or ext2 (using ext2ifs). NTFS is a bad example because it is
>> implemented on Linux using a user file system driver (fuse), which might
>> influence the performance test.
>
> This would be a useless comparison in so many ways.  ext2 drivers
> performs _way_ worse under windows, due to code quality.  there just
> aren't that many people interested in a high performance ext* driver
> for windows, understandably.  Also, the features of the filesystems
> are so different, you can't even compare them.  fat and ntfs are stuck
> in the dark ages compared to ext*.
>

That's why I suggested to do the test on the same filesystem. That 
should (ideally) reduce differences in the filesystem itself and show 
the latency from the get-last-modification-time call down to the driver 
and back.

While I agree that FAT today is no good any longer (except for small usb 
keys and such ^^), I'd like to know your reasons why you think NTFS is 
"stuck in the dark ages" as well.

> There is an in-kernel NTFS implementation, but it's no good for writing.

Hm... it might be enough for a test like this, because we only want to 
retrieve the last modification time, not change it. But here might apply 
the same as for the Windows ext2 driver: there might not be enough 
people interested in a high performance read only in-kernel NTFS driver 
while there is the more feature complete NTFS-3G one.

Regards,
Sven




More information about the Lazarus mailing list