[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Query about existence of way to free up unnecessary RAM usage



On Wed, 10 Sep 2014 16:04:46 +0800
Bret Busby <bret.busby@gmail.com> wrote:

> On 10/09/2014, Steve Litt <slitt@troubleshooters.com> wrote:
> > On Tue, 9 Sep 2014 21:51:35 +0200
> > Bzzzz <lazyvirus@gmx.com> wrote:
> >
> >> On Wed, 10 Sep 2014 03:30:40 +0800
> >> Bret Busby <bret.busby@gmail.com> wrote:
> >>
> >> > Alright, then; it is doing token swapping - with 99% of 16GB
> >> > memory usage, and, swapping only 4% of (about) 40GB swap
> >> > capacity, you can't seriously tell me that the swapping is
> >> > working as it should be.
> >>
> >> Anyway, a swap of 40GB is too much for a RAM of 16GB (should be
> >> around 16-20GB), unless you perform operations that generates
> >> a lot of intermediary results).
> >
> > The reason I agree with you is that if you need to swap anywhere
> > near 40GB of swap, your computer will be crawling to the point
> > where you might as well abort the guilty program (if you can
> > navigate to do so), or reboot the compute.
> >
> > That said, I just found out my swap partition is 44GB for a 16GB
> > semiconductor RAM.
> >
> 
> So, does swapping work on your system?
> 
> Does your RAM usage go above 50%, without swapping, and, above 90%,
> with less than 5% of the swap capacity being used?

Hi Bret,

I can't answer your question in English, because I've never really
understood Linux memory usage. There's an algorithm to figure out how
much RAM to use in order to cache disk accesses, and another algorithm
to figure out how much swap partition to use to substitute for RAM, and
it just twirls my head.

So the best thing I can do is show the results of a couple commands:

slitt@mydesq2:~$ vmstat -SM
procs -----------memory---------- ---swap-- -----io---- 
 r  b   swpd   free   buff  cache   si   so    bi    bo 
 0  0      0   7113    682   5172    0    0    14    20 
slitt@mydesq2:~$ free -h
       total       used   free  shared buffers  cached
Mem:     14G       8.0G   6.9G      0B    682M    5.1G
-/+ buffers/cache: 2.2G    12G
Swap:    44G         0B    44G
slitt@mydesq2:~$ ps avx k -%mem --cols 70 | less
  PID TTY      STAT   TIME  MAJFL   TRS   DRS   RSS %MEM COMMAND
 5206 tty1     Sl    85:29    199    90 1469065 537184  3.4 iceweasel
 3684 tty7     Ss+  164:46     53     0 292516 185980  1.1 /usr/bin/X 
10843 tty1     Sl     0:24      2  3208 620275 152020  0.9 /scratch/ex
15930 tty1     SLl  194:54      1    75 570556 100924  0.6 /usr/lib/ic
14880 tty1     Sl     5:48    113   606 561065 68632  0.4 bluefish
 2384 ?        Ssl    0:13      3     0 172692 37200  0.2 /sbin/zfs-fu
10303 tty1     SL    12:00      0  4001 329142 36720  0.2 /usr/bin/mpl
30977 tty1     Sl     5:39      0  2392 246343 31864  0.2 smplayer
12107 tty1     Sl     3:17     33   627 374508 28320  0.1 xchat
 3733 tty1     Sl     0:00     45  2392 230091 26132  0.1 /usr/bin/pyt
 3731 tty1     S      0:00     44  2392 183983 22568  0.1 /usr/bin/pyt
12124 tty1     Sl     0:06     15   169 436406 19080  0.1 xfterm
10416 ?        Ss     0:00      3  2255 449748 18680  0.1 gvim
20935 pts/14   SL+    3:42      2  4001 358522 18392  0.1 mplayer -loo
 3717 tty1     S      0:23     66   333 156834 14088  0.0 /usr/bin/ope
22826 tty1     Sl     0:01      0    49 300502 12172  0.0 lxt
 3734 tty1     Sl     0:00     28    12 369919 11604  0.0 /usr/lib/x86
 3246 ?        Sl     0:00     96     0 362952 11032  0.0 /usr/lib/x86

Bret, I just remembered, I use the xxxterm browser a lot, and sometimes
when I use it a lot, all of a sudden the RAM consumed by X goes up to
about 99% and the system slows to a crawl, and I have to start killing
xxxterms. As a matter of fact, that's why I'm starting to use more
iceweasel and less xxxterm. xxxterm is a touch typist's dream browser,
but it gets out of control sometimes.

By the way, I'm using Openbox for a windowmanager/desktop. The commands
I gave you represent a typical usage for me, except I have less than
normal xxxterms running.

Anyway, I hope this helps.

SteveT

Steve Litt                *  http://www.troubleshooters.com/
Troubleshooting Training  *  Human Performance


Reply to: