Re: Calculating optimal disk partitions

[craig]

On Wednesday, May 8, 2013 16:46, "Chris Murphy" <address@hidden> said:

> On May 8, 2013, at 2:37 PM, address@hidden wrote:
> 
>> On Wednesday, May 8, 2013 13:17, "Chris Murphy" <address@hidden> said:
>>
>> Thanks for responding Chris!
>>
>>> On a 512byte physical and logical sector hard drive, the messages can be
>>> ignored.
>>> Alignment isn't an issue.
>>
>> Wouldn't the generation of an error message, when no error condition exists, 
>> be
>> considered a bug? At the very least I would think it would be undesirable.
> 
> It's an older version of parted. I'd try it with 3.1. In any case, it's 
> difficult
> because while 512B physical/logical sector drives should default to an 
> alignment
> value of 1 sector, this probably isn't good for most SSDs which claim a 512B
> physical/logical sector size. The way it works now by default is to align on 
> 1MB
> boundaries regardless of the physical sector size.

Thanks Chris, excellent point about the version. I stick with Debian's packages 
whenever I can, but I certainly don't mind trying the latest version of parted. 
I will let you know how that goes. In the meantime, do you mind if I post a 
pointer to our conversation here back on the Debian User list for anyone that 
might be following it there?

> 
> It's possible in your case the warning is about another partition which isn't
> aligned, rather than the new one you're trying to create. If that's the case, 
> the
> error message is misleading.
> 
>>
>> What about a 4096 physical sector drive?
> 
> Alignment on a 4Kn AF HDD is also a non-factor. There's nothing to align. But 
> I
> still think the default is 1MB alignment with new versions of parted. As a 
> data
> drive, they work. For boot drives, even recent firmware, let alone older 
> firmware,
> don't like them at all.
> 
> For 512e AF HDD, alignment is a considerable factor depending on the firmware
> implementation. It's best to align.

I would be more than happy to put together any helpful documentation if you or 
someone else would give me some guidelines on what to put in it.

>>
>>> For SSDs which effectively lie about their physical sector size, the
>>> consequences
>>> of miss alignment are variable the firmware.
>>
>> If mis-alignment can negatively impact performance, I would expect a 
>> documented
>> procedure on avoiding it. Or is my expectation out of line? I certainly 
>> wouldn't
>> mind helping document the procedure if I could get a clear understanding on 
>> how
>> what the procedure is.
> 
> It would be nice if it worked this way, but in reality solid state storage is 
> in
> flux on both a physical level, as well as a firmware (the thing most 
> responsible
> for the logical abstraction from the physical drive). Firmware makes a huge, 
> huge
> difference in SSD behavior.
> 
> 
> Chris Murphy

Most appreciatively,
Craig



Sent - Gtek Web Mail