Re: [ANNOUNCE] util-linux-ng v2.17 (stable)

[H. Peter Anvin]

On 01/08/2010 02:40 PM, Andreas Dilger wrote:
> On 2010-01-08, at 14:43, H. Peter Anvin wrote:
>> On 01/08/2010 01:33 AM, Karel Zak wrote:
>>> fdisk:
>>>   - the fdisk command aligns newly created partitions to  
>>> minimum_io_size
>>>     boundary ("minimum_io_size" is physical sector size or stripe  
>>> chunk
>>>     size on RAIDs).
>>>
>>>   - the fdisk command supports disks with alignment_offset now.
>>
>> I think we should align, by default, much more aggressively than  
>> that --
>> because frequently we just don't know what the real physical alignment
>> is (think of flash media, which uses large erase blocks underneath.)
>> Windows aligns partitions 1 MB boundaries by default now -- I think
>> that's probably a reasonably good idea, at least for any disk that's  
>> not
>> tiny, say 256 MB or less.
> 
> I agree whole heartedly.  We steer users very sharply away from using  
> partitions at all, because on h/w RAID devices the 512-byte offset  
> from fdisk completely kills RAID-5/6 performance.
> 
> Making the default minimum alignment for DOS/GPT partitions makes a  
> lot of sense, and LVM PEs should be on 1MB boundaries as well (I don't  
> think that is the case today either).
> 

As far as I can tell, there is absolutely no reason to not align all
partitions, all the time (for both GPT and MBR... GPT may need a "dummy
alignment partition" to fulfill the "no nonpartitioned space" dictum,
although it seems like an impossible requirement in practice -- I think
they main reason for it is to avoid abusers like Grub relying on putting
data in unpartitioned space.)

	-hpa
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/