From: adam@address.invalid   
      
   Bit Twister wrote:   
   > On Sat, 16 Jun 2012 14:46:29 -0400, Adam wrote:   
      
   >> Would you recommend putting swap on the HD or on /dev/shm?   
   >   
   > As I misunderstand it, shm has a finite number of write cycles. With   
   > my luck, it could burn out and system would not boot.   
      
   I'll have to look it up... I thought /dev/shm was similar to a tempfs,   
   not an SSD.   
      
   >> The new box has 8 GB RAM.   
   >   
   > I would go with disk swap by guessing you would hardly ever get into   
   > it with that much ram. Unless you are setting system to be able to   
   > suspend/resume operation. If so, you need swap to be size of ram + 1gig.   
      
   Makes sense.  I can always change things later too.  Although it's a   
   desktop system, my UPS configuration on my current (old) box is set to   
   automatically go into hibernation when there's a longer outage, and   
   automatically thaw when AC returns, so for that I would need swap to be   
   maximum RAM + 1 GB = 33 GB.  (This box shipped with 8 GB, but I intend   
   to max it out.)  A 33 GB swap partition?  Geez, I can remember when   
   entire HDs were smaller than that.   
      
   >> This (old) one only has 1 GB, and within an hour after   
   >> booting it's usually using a little of swap.  That's usually only for   
   >> short durations,   
   >   
   > That sounds about right for a 1 gig system.   
      
   This one invariably uses swap heavily if I run a VM or use GIMP.   
      
   > That is with MSEC hourly jobs   
      
   Are any of those security-related?  What do those do?   
      
   >>> I have a /hotbu partition for a hot backup of the "Production" install   
   >>> for just in case I screw up as root.   
   >>   
   >> Sounds like a good idea, but how often do you do that, and when?   
   >   
   > Before I install some updates which might clobber the system like new   
   > video drivers. Rest of the time is after 20 or so updates.   
      
   IOW it's not automated, and only happens when you request it.  That   
   makes sense; I think I'll implement that too.   
      
   >> I come from an era when minimizing seek time   
   >> and saving bytes mattered. :-)   
   >   
   > I hear that. When was the last time you ran [SpinRight].  :)   
      
   A while ago.  Back in my TRS-80 days, I used to manually optimize the   
   placement of files on a disk, and once experimented with changing the   
   low-level HD formatter to try different interleaves.  The results were   
   interesting -- the interleave that was fastest for executable files   
   wasn't the same one that was fastest for data files.   
      
   > You would have to know the physical setup on your drive. How many   
   > heads/platters. Personally I placed my big partitions on the end/back   
   > of the drive and OSs in the front and left unsed space in the   
   > middle. Only reason for the above was old boot loader could not reach   
   > OS for booting unless towards front of drive address space.   
      
   I gather the 1024-cylinder limit for /boot is long gone.  I'm planning   
   things so /accounts isn't _too_ far from any root partition I'd use   
   regularly.   
      
   > Trying to get that last bit of performance is not worth it for me.   
      
   Me neither.  My motto for computing is "good enough" -- if I tried for   
   perfection I'd never get anything finished.   
      
   > Keep in mind, I always do clean installs and I am cycling through   
   > releases with last install, Production, new install partitions.   
   > Once I am satisfied new install is ok, it becomes Production, and last   
   > install partition will become the next new install partition.   
      
   That's what I'm doing.  You've given me a lot of very useful suggestions   
   over the past few years, for which I thank you.   
      
   Adam   
   --   
   Registered Linux User #536473   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)