[mythtv-users] Storage directory management across SSD/HD

Stephen Worthington stephen_agent at jsw.gen.nz
Sat Mar 11 11:39:58 UTC 2023


On Sat, 11 Mar 2023 10:13:30 +0000, you wrote:

>On 11/03/2023 03:32, Stephen Worthington wrote:
>> 
>(snip)
>> 
>> I am not sure why you feel the need to have recordings go to your SSD.
>> As long as hard drives are spinning at the time, starting a recording
>> or playback is very fast and there is not much gained by using an SSD.
>
>One potential problem would be that most very large modern HHDs are shingled; I read that this makes 
>reads quick but writes much slower, as it has to rewrite a whole track every time anything on that 
>track changes. This might be a reason to use an SSD as a 'cache'.
>
>I have no numbers to hand to say whether the above is still true, or whether the inbuilt RAM cache 
>in the drive is large enough to soak up the required write rate.

You can never use a shingled drive for MythTV recording, as the Linux
and MythTV buffers combined are far too small to allow for writing to
the disk to stop for up to 60 seconds as can happen.  So for MythTV
use, you need to record to an ordinary CMR drive and afterwards you
can move the recording files to a shingled drive.  Four of my archive
drives are shingled and I wrote mythsgu as a way to automate my file
moves to the archive drives.  When I got my first shingled archive
drive, I had to update the kernel for it to work properly - older
kernels just timed out when a shingle rewrite happened.

But it is incorrect to say that most very large modern HDDs are
shingled.  If you are a data centre or massive corporation, the sales
people will offer you those options, but most retail sites do not even
list those sorts of drives as they require the operating system do the
shingle operations, rather than the drive doing them automatically.
There are consumer oriented shingled drives available, usually in the
Red/NAS type drives, where the drive does the shingled operations
automatically, and all you see from the operating system is that a
write operation stalls for a very long time.  Operating systems now
cope with that stall situation rather than timing out and saying the
write failed.  My shingled archive drives are of this latter type.  It
used to be that the largest drive I could buy was a shingled one, but
quite some margin, but that has changed over the last few years and
now you can buy monster size enterprise class drives instead.  I now
have two 18 Tbyte WD enterprise drives and a 20 Tbyte Seagate one, and
sundry smaller enterprise class drives (10-16 Tbytes).  The Seagate
EXOS line usually are cheaper than the Red/NAS drives of the same size
from both Seagate and WD and perform better.  They are not cheap, but
the performance is great, they have 5 year warranties and are huge.
And not shingled.  It is fascinating watching a file copy between
these drives - they really do do 250 Mbytes/s, so about half what a
good SATA SSD will do.  There are now 22 Tbyte enterprise drives
available and 24 Tbyte coming later this year.


More information about the mythtv-users mailing list