2

Is there a way to keep disk caches when hibernating Linux? I understand and agree that dropping them is the best option all-around. But my system is unresponsive after resume and I would like to give a try to keeping that data.

Of course, both hibernating and resuming will be slower, since disk cache data will have to be written to swap and restored. And, of course, I will need more swap space than I do now. But still, I've been looking all over the web and I haven't found any way to set things up like that.

I use pm-utils with uswsusp as backend, but I'm open to suggestions.

ateijelo
  • 151
  • 4
  • What do you mean by "unresponsive"? If I understand correctly, loading files from a contiguous cache on disk is only going to improve performance (and probably only slightly at that) if most of the files cached are going to be used immediately after you resume. How much RAM and disk swap space do you have, and what's unique about you usage patterns (i.e. why don't standard memory and I/O optimization algorithms apply to the way you use your computer?) – depquid Mar 27 '13 at 13:17
  • By unresponsive I mean that switching between already opened apps becomes slow and a lot of disk activity happens. After a while, system responsiveness becomes normal. My (I think not so crazy) guess is that, since all disk caches were dropped for hibernation, the system needs to hit the disk again when it needs the data. I may be wrong. I have 8Gb of RAM and 4Gb of swap. In my last hibernation, s2both image size was a bit over 1.2Gb. – ateijelo Mar 27 '13 at 14:18
  • `hibernate-ram` keeps the caches around, but it requires power to keep alive. – frostschutz Mar 27 '13 at 16:09
  • @ateijelo Your guess is correct. Another approach could be to force-read some processes's mapped memory into RAM with a tool like [tct's `pcat`](http://www.porcupine.org/forensics/tct.html) or [my Python script](http://unix.stackexchange.com/questions/6267/how-to-unswap-my-desktop/6271#6271) (but it doesn't just load what was in RAM before, it loads everything, which can be inconveniently large). – Gilles 'SO- stop being evil' Mar 27 '13 at 23:03

2 Answers2

2

I know your pain, after hibernation my computer is much slower.

But there is another way. It's called TuxOnIce.

It is much faster than the in-kernel hibernation and it is able to save ALL the memory in use, including the disk caches. It makes a massive difference and actually makes hibernation in Linux usable on a day to day basis.

Beware - it does involve building your own kernel. If you're a Debian or Ubuntu user, this is fairly easy using kernel-package (make-kpkg). Good luck!

John-o
  • 21
  • 1
  • Thanks, I'll give it a try. I do use Debian, so I think it won't be hard. I'll post whether it made a difference or not. – ateijelo Mar 28 '13 at 12:39
0

You should check your swappiness value. Usually it is left at 60. It should be 10 or 15:

cat /proc/sys/vm/swappiness

You can change it on the fly with

sudo bash -c "echo -e 10 > /proc/sys/vm/swappiness"

and make it stick with

sudo bash -c "echo 'vm.swappiness = 15' >> /etc/sysctl.conf"

Reinaert Albrecht
  • 11
  • 1