SCIENTIFIC-LINUX-USERS Archives

June 2013

SCIENTIFIC-LINUX-USERS@LISTSERV.FNAL.GOV

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Chuck Munro <[log in to unmask]>
Reply To:
Chuck Munro <[log in to unmask]>
Date:
Tue, 11 Jun 2013 08:14:16 -0700
Content-Type:
text/plain
Parts/Attachments:
text/plain (26 lines)
My thanks to all the responders.  I'm encouraged by Vladimir's advice 
regarding writes.  I've been thinking about wear leveling, and based on 
how much disk space I really need to run the relatively static firewall 
VM images, there will be a *lot* of unassigned spare cells in the flash, 
so writes shouldn't matter much after all.

I'll mount the filesystems with the 'noatime' option since that info 
isn't needed, and put /tmp and /var/tmp in tmpfs.  There are also other 
tricks mounting ext4 partitions that significantly reduce writes.  So, 
it would appear this is a no-brainer.

I was also made aware of the 'vm.swappiness' kernel parameter, and since 
swapping is very unlikely to ever happen, I can safely set this to zero.

On a side note, I'm really surprised that constant same-region reads can 
cause wear.  That's something I would never have thought of.

The plan is to use Intel 520-series SSDs in software RAID-1.  A 
high-quality power supply and a motherboard with all-solid capacitors 
would also raise reliability.  I usually underclock CPU chips a bit, and 
use really good coolers (such as Noctua), so it all adds up.  With any 
luck the box would run pretty well forever :-)

Thanks all!
Chuck

ATOM RSS1 RSS2