[CALUG] Looking for advice

[Howard Bampton]

[Looks like an unintended reply to failure. Redirecting and commenting
below.]

On Thu, Mar 7, 2019 at 4:47 PM Bryan Smith <b.j.smith at ieee.org> wrote:

> Sorry, forgot a note here...
>
> On Thu, Mar 7, 2019, 15:41 Bryan Smith <b.j.smith at ieee.org wrote:
>
>> On Wed, Mar 6, 2019, 22:20 Howard Bampton <howard.bampton at gmail.com
>> wrote:
>>
>>> 95% of the time you can share $HOME.
>>>
>>
>> Yep, I use the same LVM Volume Group (VG) for all distributions, and then
>> create 'f{root, var, tmp**}', 'u{root,var,tmp**}' LVs, along with a shared
>> 'elocal' LV (more on that in a bit).
>>
>
> **NOTE:  I don't create /tmp on systems with only a NAND storage device,
> but use tmpfs. I also avoid creating swap as well, or I create a swap LV,
> but turn vm.swappiness=0 (disabled). If the system has a spindle (or hybrid
> NAND+platter), then I put /tmp and swap on that device (along with /var,
> but /var must exist regardless) .
>
> The performance increase using NAND is not worth the number of blocks
> written.  That's the problem with most 'lifespan' tests of NAND.  They
> assume complete data block writes, not massive inode and small file
> creation/deletion that wears out the NAND much faster - - literally 32
> bytes here, 256 bytes there, that are equivalent to 32KiB NAND cell
> commits, even when coalesced.
>
> E.g., Windows Update is the worst.  It wears cells 1,000x more than it
> actually writes (which is why it's so slow on platter) .  Linux is not
> remotely as bad, and /var is buffered, but one should still eliminate /tmp
> and swap, and mitigate /var from NAND devices.
>
> Also disable browser disk caching if your user's home directory is on
> NAND.  Those temporary files are rather killer too.  Lots and lots of inode
> creation/deletion for little things.
>
>
Note that I go the other way for storage.

My 2007 vintage desktop ran a lot quicker when I added an MLC SSD for swap
(return from hybernate used to be slow enough that I'd go brush my teeth
after turning it back on and it often would still be waking up when I was
done). That drive is still in use (~5 years as a swap device, ~2 in an
external enclosure as extra storage with mostly WORM use patterns.)

My lightly used 2010 laptop (IBM SLC SSD) has no drive issues after 8 years
(OK, I ran into problems with thrashing because the swap space was too
small for the number of tabs I tend to have open in the web browser).
Mostly hibernate state when not in use up until last summer when Ubuntu
made it easier to use suspend. The battery is about gone, but the SSD seems
to be fine, even if it is a bit small by today's standards. Yes, part of
the reason this drive is still fine is probably because it is an SLC, not
an MLC drive, but a half dozen 30 minute thrashing sessions probably did
stress test it.

My current laptop (daily use, 2 years of heavy use) is fine so far (again,
MLC SSD with swap and hibernate or suspend when not in use).

I should note that I hit swap a fair amount because I do too many things at
once- all three computers had maximum RAM (4G for the desktop, 2G for the
2010 laptop, 8G for this one), so the problem really is me leaving too many
tabs open in my browser (some of which have ill mannered ads that
eventually crash the tab in question), leaving Libre' Office up for months
(it seems to have a slow memory leak), or running java based games that
have memory leaks.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.unknownlamer.org/pipermail/calug/attachments/20190307/6bbfffbc/attachment-0001.html>