I’m using a opensuse 11.1 box 24 hours per day as a DSL gateway, with a mobile processor main board, just to save energy. I bought a 32 GB solid state SATA-disk intended for grabbing all the system traffic, that makes the HD never go down, in power save mode, though there are no valuable user tasks running or requested either, for longer periods.
First I thought, simply to mount the /var directory onto the solid state disk but now I’m not sure, wether that might be sufficient to catch all the intended hi frequent system traffic. Is anyone out there, who’s got a fairly simple solution for this sort of problem?
Placing the swap partition on it (if you have one) would probably be a good idea.
Got no other advice to give as my server runs off the disk it is installed on and it if I understand you correctly you want the operating system disk to idle, putting all disk activaties after booting on the ssd?
Maybe reinstalling the whole thing on the SSD would be a better idea (if the 32GB is big enough for you) as even an idle disk of course eats more power then no disk.
Sure, I don’t see any harm in it. Try it and see what happens. But maybe you might want to make var a subdirectory on the SSD in case you find other sources of frequent traffic. I.e.
/var -> /ssd/var
and maybe later you find /tmp gets temp files frequently,
Thank´s for the tips. Because I was busy for a couple of days now, I didn’t react yet.
I forgot to mention of course, that I already put the swap partition onto the ssd.
To make all the installation on the ssd and route the dedicated user paths only, via symlinks for example, to a mount on a normal hd, seems to me a good choice indeed but I’d fairly prefer a bit mor transparency in what I’m doing.
On the other hand, to have a simple method to find the access frequency of files over a given period of time, say 24 hours or so, within a specified subtree like /var, might clarify the demand analytically, where to take hands on. I’m not very skilled on command-line tools, so if you got an idee to get it that way, I’d bee grateful for an advice. Another problem seems to me, in redirecting e.c. /var/log to another place like /ssd/var/log, especially in a running system and what happens, if the ssd crashes? Is it possible to have something like a fall-back to the original (physical) path /var/log?
Ok. A lot of details that make a simple think look complicated. May bee just for me. So thanks in advance for your patience.
Meanwhile installing 11.1 completely on the ssd, as proposed in the post of Axeia.
Some strange effects occur though but the facts first :
The ssd is /dev/sdb 32 GB with two partitions,
sdb1 is mounted as / and is holding the complete installation.
sdb2 is the mounted swap partition.
The SATA hd 1 TB is recognized as /dev/sda with only one partition,
sda1 which is mounted as /data.
Additionally I installed spindown.rpm adapted for 11.1, to improve the flexibility for power saving of the hard disks. Here is the configuration file:
/etc/spindown.conf
[General]
cycle-time = 60 # Number of seconds between cycles.
idle-time = 3600 # The time in seconds a disk has to be idle before it is sp
# Do not use a value that is to small (less than 3600) unle
# know what you are doing.
syslog = 0 # Turn this on if you want spindown to log messages using s
# You will not want to turn this on if you are trying to sp
# the disk with the root filesystem on.
[Disk data]
id = scsi-SATA_SAMSUNG_HD103UJS13PJ9BQ900693
spindown = 1
idle-time = 600
[Disk ssd]
id = scsi-SATA_TS32GSSD25S-M_0023050800B7
spindown = 0
Though no additional links point somewhere into /data and no manual access is made, the hd spins up and down from time to time. That brings me back to the question in my previous post, not answered yet: How to monitor file access and its frequency on a given subtree?