Re: [OT] Hard Drive Energy Not Worth Conserving drives?



Stefan Monnier put forth on 1/11/2011 10:28 PM:

Isn't it rather than the kernel chooses to only use the logical
sector size? Where/when does the drive report 512B physical
sector sizes?

Read the ATA and SCSI specifications. Or ask on either mailing list. In short,
the drive presents its LBA addressing based on 512B sectors. The kernel can't
choose to ignore that--it's stuck with it. Since the drive is presenting LBA
based on 512B sectors, there is no way the kernel can address LBA based on 4K
sectors.

In any case, the issue is probably not really in the kernel but in the
filesystems and partitioning tools: all that's really needed to use the

The problem is none of those things. The problem is hybrid drives. IIRC, the
kernel, libata, libscsi, etc, are all ready for _native_ 4K sector drives. The
current "problem" with the hybrid drives is that the partitioning utilities
don't automatically align partitions on the underlying 4k sector boundaries.
So, what happens is, in essence, when a filesystem is laid down that uses 4K
blocks, it will lay across 4 _translated_ 512B sectors. Thus, you end up with a
4K filesystem block that lays across two physical 4K disk sectors. Thus, each
time that FS block is written, 8*512B sectors must be read and re-written.
Thus, _TWO_ physical 4K disk sectors must be read and re-written instead of just
one on an aligned partition, akin to the the RAID5/6 read/modify/write penalty.
This misalignment cuts performance by half or more because every FS block write
requires to reads and two writes instead of just one.

drive efficiently is for fdisk/parted and for mkfs to be told (and make
use of) the physical block size. Of course, maybe a good way to provide
this info is to teach the kernel about it so those tools don't need to
use side-band info via hdparm.

Maybe you should hit the LKML, GNU fdisk, and GNU parted list archives before
continuing this thread.

Indeed, and for that reason 4KB physical blocks wouldn't cause
additional disk space usage.

The space savings with 4KB sectors has nothing to do with file systems or user
data. It has to do with the per sector ECC information that all drives
calculate and store in between sectors to safeguard the data in the sector. The
number of bits of ECC required per 4KB sector is significantly less than that
occupied by the 4 ECC segments of four 512 byte sectors. This is the ONLY
reason these 4KB sector drives were developed: more actual end user space on
the drive.

--
Stan



--
To UNSUBSCRIBE, email to debian-user-REQUEST@xxxxxxxxxxxxxxxx
with a subject of "unsubscribe". Trouble? Contact listmaster@xxxxxxxxxxxxxxxx
Archive: http://lists.debian.org/4D2DEF0F.1080606@xxxxxxxxxxxxxxxxx



Relevant Pages

  • Re: Event ID 11 - please help
    ... a ton of drives. ... > two more SATA WD 120 GB drives connected to the Promise controller ... > Offset Length Sectors ... > Type Indicator Partition Partition ...
    (microsoft.public.windowsxp.hardware)
  • Re: max SATA disc size i can use
    ... PATA drive -- up to 281474976710656 sectors. ... Any disk you can buy ... was produced there were no drives larger than 2.1TB, ... the real limitation today is that the partitioning system used ...
    (uk.comp.homebuilt)
  • Re: Hard Disk Failure
    ... Does the HDD's OEM diagnostic tool see the full size of the HDD, ... Do you see anything in Disk Management that relates to the 160GB HDD, ... tool on the drives and this is the result. ... Sectors per Cluster: 8 ...
    (microsoft.public.windowsxp.hardware)
  • [opensuse] Disk Drive Technology and partition alignment [WAS: ]Re: 12.3 + ntfs + ext4 + USB3 + diff
    ... It is time for you to enter the 21st century in respects to disk ... drive is meaningless in modern rotating drives. ... I state that the S value for sectors per track is not ... Depends on what disk partition tool you use, but I don't know of any ...
    (SuSE)
  • Re: possible bug in ide-disk.c (2.6.18.2 but also older)
    ... your disk has an odd number of sectors? ... appear with a plain vanilla 2.6.18.2 kernel from www.kernel.org. ... The original reported number of sectors was an even number. ... and triggers turning off dma on some other drives and, ...
    (Linux-Kernel)