How get faster app loads--ext2 partition? Other tricks?
-
- Posts: 47
- Joined: Fri 01 Sep 2006, 02:40
- Location: Southwestern U.S.
How get faster app loads--ext2 partition? Other tricks?
Does anyone knows any tricks to get apps to load faster?
I like the idea of using old hardware w good apps, hence Puppy.
A Linux newbie, I went from Puppy 2.10 on a CD to on my Win 98 hard disk partition via Pup4Dos, but speed went up only 10%.
I've noticed the second load of an app is faster (2 times as fast--caching?) but the first load is what a user remembers.
The apps load slowly in comparison to similar apps on Win 98
(load time is until screen has complete screen displayed):
Example:...First load of Word 97 in Win 98.................9 seconds
..................First load of AbiWord in Puppy 2.10........20 seconds
Example:...First load of Excel 97 in Win 98.................4 seconds
..................First load of Gnumeric in Puppy 2.10.........9 seconds
Example:...First load of Netscape 6.1 in Win 98........25 seconds
..................First load of Seamonkey in Puppy 2.10....45 seconds
Boot times for Win 98 and Puppy are about the same--2 minutes.
My next idea has been to put Puppy on an ext2 partition, but any other speed-up ideas are welcomed because I'm nervous about accidentally deleting the old Win 98 stuff and I don't have a good backup mechanism for the old Win 98 OS and software.
I've read an ext2 partition may not be any faster than Pup4DOS:
http://www.murga.org/~puppy/viewtopic.p ... peed+puppy
but I'm not sure; would like comments this vs. other speedups.
Here's the current laptop I am using:
Dell Inspiron 3200 laptop (NeoMagic 128XD display chip via Xorg)
-233 MHZ Pentium II
-144 MB RAM
-3.8 GB hard disk (4,000 rpm; defragged; 1.4 GB free)
-high-speed ethernet via PCMCIA card (Puppy firewall installed)
-pup_save.3fs 805 MB (w OpenOffice & Opera but icons disappeared
-pupswap.swp 105 MB
Thanks for any help you can provide.
I like the idea of using old hardware w good apps, hence Puppy.
A Linux newbie, I went from Puppy 2.10 on a CD to on my Win 98 hard disk partition via Pup4Dos, but speed went up only 10%.
I've noticed the second load of an app is faster (2 times as fast--caching?) but the first load is what a user remembers.
The apps load slowly in comparison to similar apps on Win 98
(load time is until screen has complete screen displayed):
Example:...First load of Word 97 in Win 98.................9 seconds
..................First load of AbiWord in Puppy 2.10........20 seconds
Example:...First load of Excel 97 in Win 98.................4 seconds
..................First load of Gnumeric in Puppy 2.10.........9 seconds
Example:...First load of Netscape 6.1 in Win 98........25 seconds
..................First load of Seamonkey in Puppy 2.10....45 seconds
Boot times for Win 98 and Puppy are about the same--2 minutes.
My next idea has been to put Puppy on an ext2 partition, but any other speed-up ideas are welcomed because I'm nervous about accidentally deleting the old Win 98 stuff and I don't have a good backup mechanism for the old Win 98 OS and software.
I've read an ext2 partition may not be any faster than Pup4DOS:
http://www.murga.org/~puppy/viewtopic.p ... peed+puppy
but I'm not sure; would like comments this vs. other speedups.
Here's the current laptop I am using:
Dell Inspiron 3200 laptop (NeoMagic 128XD display chip via Xorg)
-233 MHZ Pentium II
-144 MB RAM
-3.8 GB hard disk (4,000 rpm; defragged; 1.4 GB free)
-high-speed ethernet via PCMCIA card (Puppy firewall installed)
-pup_save.3fs 805 MB (w OpenOffice & Opera but icons disappeared
-pupswap.swp 105 MB
Thanks for any help you can provide.
- BarryK
- Puppy Master
- Posts: 9392
- Joined: Mon 09 May 2005, 09:23
- Location: Perth, Western Australia
- Contact:
I would like to see your startup figures using puppy 2.02.
Puppy 2.10 introduces LZMA compression of the pup_210.sfs file (which has
all the apps, such as Gnumeric and Abiword). Pup 2.02 just used GZIP
compression.
If it turns out that there's a big slowdown on older hardware, I may have to
release Puppy 2.11 in two flavours, one with LZMA and one without (and maybe
less apps/libraries to keep its size down).
Puppy 2.10 introduces LZMA compression of the pup_210.sfs file (which has
all the apps, such as Gnumeric and Abiword). Pup 2.02 just used GZIP
compression.
If it turns out that there's a big slowdown on older hardware, I may have to
release Puppy 2.11 in two flavours, one with LZMA and one without (and maybe
less apps/libraries to keep its size down).
C3 and hybrid booting
Barry, the speed difference is noticeable even in the C3-800 Mhz using CF card as IDEHD. Booting the big PuppyOffice2.02 was a bit faster than booting the standard 2.10.
Hope that helps.
As to your question, FuturePerfect, a liveCD boot puts all files in RAM (if you have sufficient supply of 128 MB or more), and this makes loading of programs fast. The frugal install has a similar property, although the files are read from hard disk.
You can do a hybrid of these by copying pup_*.sfs to the same partition where you have the save file, pup_save.3fs. You then boot with the liveCD but enjoy the speed of loading pup*.sfs from the hard disk.
Hope that helps.
As to your question, FuturePerfect, a liveCD boot puts all files in RAM (if you have sufficient supply of 128 MB or more), and this makes loading of programs fast. The frugal install has a similar property, although the files are read from hard disk.
You can do a hybrid of these by copying pup_*.sfs to the same partition where you have the save file, pup_save.3fs. You then boot with the liveCD but enjoy the speed of loading pup*.sfs from the hard disk.
Puppy user since Oct 2004. Want FreeOffice? [url=http://puppylinux.info/topic/freeoffice-2012-sfs]Get the sfs (English only)[/url].
-
- Posts: 47
- Joined: Fri 01 Sep 2006, 02:40
- Location: Southwestern U.S.
Thanks to all for reading my post.
Barry: I renamed all my 2.10 files on my hard disk and substituted the properly named files from my 2.02 CD (I had never run 2.02 from hard disk before). Everything was the same as in 2.10, except I did not re-download Opera and Open Office 2 (but do the new releases of AbiWord and Gnumeric in 2.10 cause slowness?). I made the pup_save.3fs the same size (and the pupswap.swp file was automatically made the same size by Puppy 2.02). Let me know if I missed something.
What a difference 2.02 made:
..................First load of AbiWord in Puppy 2.02........8 seconds
..................First load of Gnumeric in Puppy 2.02.........5 seconds
..................First load of SeaMonkey in Puppy 2.02....20 seconds
Boot time for Puppy 2.02--1 min 25 secs
* I hope 2.10 and successor releases will be made available uncompressed so old hardware can still be fast loading applications.
2 quick related questions:
1) Does anyone know if using ext2 would make things even faster?
2) Does anyone know any way to get AbiWord and Gnumeric help to show from local files when Help is clicked?
Newbies can't always be online to see help; when I click help now these programs make me be online to get help from web sites.
AbiWord seemed to say there was a version you could download to get the help? Gnumeric help is just html and could maybe how be linked to be readable offline?
I think even old PCs like I use have enough disk space to include the help and it would make a much more useful package.
Thanks again.
Barry: I renamed all my 2.10 files on my hard disk and substituted the properly named files from my 2.02 CD (I had never run 2.02 from hard disk before). Everything was the same as in 2.10, except I did not re-download Opera and Open Office 2 (but do the new releases of AbiWord and Gnumeric in 2.10 cause slowness?). I made the pup_save.3fs the same size (and the pupswap.swp file was automatically made the same size by Puppy 2.02). Let me know if I missed something.
What a difference 2.02 made:
..................First load of AbiWord in Puppy 2.02........8 seconds
..................First load of Gnumeric in Puppy 2.02.........5 seconds
..................First load of SeaMonkey in Puppy 2.02....20 seconds
Boot time for Puppy 2.02--1 min 25 secs
* I hope 2.10 and successor releases will be made available uncompressed so old hardware can still be fast loading applications.
2 quick related questions:
1) Does anyone know if using ext2 would make things even faster?
2) Does anyone know any way to get AbiWord and Gnumeric help to show from local files when Help is clicked?
Newbies can't always be online to see help; when I click help now these programs make me be online to get help from web sites.
AbiWord seemed to say there was a version you could download to get the help? Gnumeric help is just html and could maybe how be linked to be readable offline?
I think even old PCs like I use have enough disk space to include the help and it would make a much more useful package.
Thanks again.
-
- Posts: 47
- Joined: Fri 01 Sep 2006, 02:40
- Location: Southwestern U.S.
i thought i would try to make some sort of benchmark test for Puppy, to test LZMA
in Puppy 108, /usr is:
Total: 453 M (25985 files, 3223 directories)
i rebooted Puppy, to clear any caching, and i typed:
# time tar -cf /dev/null /usr
tar: Removing leading `/' from member names
real 0m3.064s
user 0m0.260s
sys 0m1.540s
i did the same thing in Puppy 210, with these results:
Total: 357 Mb (21612 files, 2770 directories)
# time tar -cf /dev/null /usr
tar: Removing leading `/' from member names
real 0m2.626s
user 0m0.420s
sys 0m2.160s
Puppy 210 doesn't have as large a file system as 108, so i multiplied the 210 time by 453/357
# dc 2.626 453 \* 357 / p
3.33215
so these are the times to read about 450 megs of squashfs'ed files:
Puppy 108: 3.064s
Puppy 210: 3.332s
this is a ratio of
# dc 3.33215 3.064 / p
1.08752
that is, the time it takes to read 450 megs of squashfs files in Puppy 108 is about the same as the time it takes to read 450 megs of squashfs files in Puppy 210
that is, the difference is almost negligible
i tried copying the file in /usr/bin and /usr/local/bin to another partition, 10 times, about 800 megs:
Puppy 108: 33.049s
Puppy 210: 35.392s
again, not a lot of difference
i tried copying the files to /tmp:
Puppy 108: 7.668s
Puppy 210: 3m5.235s = 185.235s
that's a huge difference ... about 24 times slower
in Puppy 108, /tmp is ram ... in Puppy 210, /tmp is written to the save file
i wonder if unionfs is more of a factor than LZMA?
i checked the boot times too (to the Icewm desktop):
Puppy 108:
bootup: 38 secs
mozstart: 5 secs
abiword: 2 secs
Puppy 2.10
bootup: 27 secs
mozstart: 7 secs
abiword: 3 secs
anyone see if i'm doing anything wrong? i don't think tar is being clever and not really reading the bytes from the .sfs files ... i also tried it this way, with similar results:
time tar -c --to-stdout /usr > /dev/null
as far as i can tell, the actual decompressing of the files in the .sfs files is negligible compared to the tar and cp processes
in Puppy 108, /usr is:
Total: 453 M (25985 files, 3223 directories)
i rebooted Puppy, to clear any caching, and i typed:
# time tar -cf /dev/null /usr
tar: Removing leading `/' from member names
real 0m3.064s
user 0m0.260s
sys 0m1.540s
i did the same thing in Puppy 210, with these results:
Total: 357 Mb (21612 files, 2770 directories)
# time tar -cf /dev/null /usr
tar: Removing leading `/' from member names
real 0m2.626s
user 0m0.420s
sys 0m2.160s
Puppy 210 doesn't have as large a file system as 108, so i multiplied the 210 time by 453/357
# dc 2.626 453 \* 357 / p
3.33215
so these are the times to read about 450 megs of squashfs'ed files:
Puppy 108: 3.064s
Puppy 210: 3.332s
this is a ratio of
# dc 3.33215 3.064 / p
1.08752
that is, the time it takes to read 450 megs of squashfs files in Puppy 108 is about the same as the time it takes to read 450 megs of squashfs files in Puppy 210
that is, the difference is almost negligible
i tried copying the file in /usr/bin and /usr/local/bin to another partition, 10 times, about 800 megs:
Puppy 108: 33.049s
Puppy 210: 35.392s
again, not a lot of difference
i tried copying the files to /tmp:
Puppy 108: 7.668s
Puppy 210: 3m5.235s = 185.235s
that's a huge difference ... about 24 times slower
in Puppy 108, /tmp is ram ... in Puppy 210, /tmp is written to the save file
i wonder if unionfs is more of a factor than LZMA?
i checked the boot times too (to the Icewm desktop):
Puppy 108:
bootup: 38 secs
mozstart: 5 secs
abiword: 2 secs
Puppy 2.10
bootup: 27 secs
mozstart: 7 secs
abiword: 3 secs
anyone see if i'm doing anything wrong? i don't think tar is being clever and not really reading the bytes from the .sfs files ... i also tried it this way, with similar results:
time tar -c --to-stdout /usr > /dev/null
as far as i can tell, the actual decompressing of the files in the .sfs files is negligible compared to the tar and cp processes
No HD
Just a bit of info: the slowness that I noticed was very pronounced in the hard disk-less C3-800 Mhz with 256 MB RAM (if no HD, is /tmp used awkwardly, resulting in slowness?).
This discussion about /tmp could be related.
Edited - removed the suggestion that 2.10 behaves faster in liveCD boot with HD, which is really not so (after closer inspection).
This discussion about /tmp could be related.
Edited - removed the suggestion that 2.10 behaves faster in liveCD boot with HD, which is really not so (after closer inspection).
Last edited by raffy on Fri 22 Sep 2006, 23:34, edited 1 time in total.
Puppy user since Oct 2004. Want FreeOffice? [url=http://puppylinux.info/topic/freeoffice-2012-sfs]Get the sfs (English only)[/url].
i would think that a "CF card as IDEHD" would work the same way as a hard drive would
whether you have an option 1 or option 2 install, when you write to 210's /tmp, you are writing to the CF card (or real hard drive) ... this is a lot slower than writing to ram
but Puppy 202 should work the same way
i don't know how much Puppy is writing to /tmp ... at least a few programs and utilities and pipes use /tmp
also, writing to /var would be writing to the CF card ... Puppy 1.x is writing to ram when it write to /var
maybe writing to /dev and /proc writes to the CF card too, i don't know ... "everything is a file" in Linux
whether you have an option 1 or option 2 install, when you write to 210's /tmp, you are writing to the CF card (or real hard drive) ... this is a lot slower than writing to ram
but Puppy 202 should work the same way
i don't know how much Puppy is writing to /tmp ... at least a few programs and utilities and pipes use /tmp
also, writing to /var would be writing to the CF card ... Puppy 1.x is writing to ram when it write to /var
maybe writing to /dev and /proc writes to the CF card too, i don't know ... "everything is a file" in Linux
2.10 is slower for me
I have an IBM Thinkpad 600 with a 366 PII & 288MB RAM running Puppy off the internal IDE harddrive. I'm only using the CD to boot to the hard drive install.
Puppy 2.02 is much quicker on my machine:
2.02
-----
Boot - 40 sec
Start Browser - 13 sec
Start Abiword - 5 sec
2.10
----
Boot - 69 sec
Start Browser - 25 sec
Start Abiword - 11 sec
All of the "starts" above were initial...not cached.
For my use I would rather have the old gz compression or no compression at all.
Puppy 2.02 is much quicker on my machine:
2.02
-----
Boot - 40 sec
Start Browser - 13 sec
Start Abiword - 5 sec
2.10
----
Boot - 69 sec
Start Browser - 25 sec
Start Abiword - 11 sec
All of the "starts" above were initial...not cached.
For my use I would rather have the old gz compression or no compression at all.
-
- Posts: 295
- Joined: Sat 03 Dec 2005, 11:30
I must agree about speed - though GuestToo's results do seem to present the other side. I have not timed my programs but OpenOffice on a Celeron with 128MB took about 1 min to open in 2.02 - and takes a good three to four minutes in 2.10! Since this is double LZMA at work (from pup_210.sfs and from the squashfs file of OpenOffice [using grafpup's OpenOffice2.00 version]) that might indicate an LZMA problem.
Further evidence in that direction comes from the fact that Opera - installed as a dotpup - loads considerably faster than Seamonkey, and seems to load fast regardless of the speed of the system (I use my pen drive Puppy on a Sempron 1800, a Celeron and a PIV).
Further evidence in that direction comes from the fact that Opera - installed as a dotpup - loads considerably faster than Seamonkey, and seems to load fast regardless of the speed of the system (I use my pen drive Puppy on a Sempron 1800, a Celeron and a PIV).
i have been doing all my tests on a 900 mhz Duron with 256 megs of ram ... i do not have a slower machineI use my pen drive Puppy on a Sempron 1800
no problemFor my use I would rather have the old gz compression or no compression at all
this should work for Puppy 2.10:
to save abiword as an uncompressed file, do this:
cp /usr/bin/abiword /tmp
cp /tmp/abiword /usr/bin
you can copy from this web page by selecting the text with the left mouse button
you can paste in an rxvt console window by clicking the middle mouse button
abiword is about 5 megs ... this will use up 5 megs in your save file ... it will require about 10 megs free space to do this (5 megs for the abiword file in /tmp and 5 megs for the abiword file in /usr/bin)
if you want the compressed abiword file back, just delete the abiword file in /initrd/pup_rw/usr/bin ... then immediately reboot, because unionfs does not like you deleting files behind it's back, and might stop working until you reboot
you can do the same thing with Seamonkey:
cp -a /usr/lib/seamonkey-1.0.4/ /tmp/
cp -a /tmp/seamonkey-1.0.4/ /usr/lib/
this will require about 36 megs of free space in your save file, plus another 36 megs free space in your save file for the copy in /tmp ... a total of 72 megs of free space to so this
to get back the compressed Seamonkey files, delete the seamonkey-1.0.4 folder in /initrd/pup_rw/usr/lib/
you can do a benchmark test to see how fast abiword starts like this:
type in an rxvt console window:
time abiword
press and hold ctrl+Q
Abiword should start and immediately close ... the time it took should be displayed in the console window
i did this, rebooting each time so that the cache would be empty and conditions would be the same ... here are my results:
compressed:
# time abiword
real 0m4.568s
user 0m1.920s
sys 0m2.280s
uncompressed:
# time abiword
real 0m3.852s
user 0m1.810s
sys 0m1.410s
a difference of 0.7s, about 16% faster
for Seamonkey, my results were:
compressed:
# time mozstart
real 0m7.946s
user 0m3.100s
sys 0m4.440s
uncompressed:
# time mozstart
real 0m6.144s
user 0m3.190s
sys 0m0.940s
a difference of 1.802s, about 23% faster
it might be useful to have results for slower machines than mine
in any case, if you have the space, you can easily uncompress any files you like
Hmmmm....
There's more to the 2.02 vs 2.10 speed issue than compression!
Here are my results using GuestToo's process:
System:
IBM Thinkpad 600 PII-366 288MB RAM 40GB IDE
Puppy loaded on drive using CD only as boot disc.
2.10 compressed
--------------------
Abiword
real - 13.986
user - 3.86
sys - 7.75
Seamonkey
real - 22.605
user - 5.630
sys - 12.060
2.10 uncompressed
----------------------
Abiword
real - 12.213
user - 3.85
sys - 4.46
Seamonkey
real - 16.595
user - 5.61
sys - 1.66
2.02 compressed
--------------------
Abiword
real - 7.459
user - 2.470
sys - 2.330
Seamonkey
real - 11.575
user - 4.42
sys - 3.43
As you can plainly see, Puppy 2.02 smokes 2.10 even WITH gz compression!
Another thing I notice...
The Xorg window system boots up MUCH faster in 2.02. When booting in 2.10 it takes forever from the time the x cursor appears until the desktop is fully rendered. This makes up about 1/3 of the total boot time on my system.
Here are my results using GuestToo's process:
System:
IBM Thinkpad 600 PII-366 288MB RAM 40GB IDE
Puppy loaded on drive using CD only as boot disc.
2.10 compressed
--------------------
Abiword
real - 13.986
user - 3.86
sys - 7.75
Seamonkey
real - 22.605
user - 5.630
sys - 12.060
2.10 uncompressed
----------------------
Abiword
real - 12.213
user - 3.85
sys - 4.46
Seamonkey
real - 16.595
user - 5.61
sys - 1.66
2.02 compressed
--------------------
Abiword
real - 7.459
user - 2.470
sys - 2.330
Seamonkey
real - 11.575
user - 4.42
sys - 3.43
As you can plainly see, Puppy 2.02 smokes 2.10 even WITH gz compression!
Another thing I notice...
The Xorg window system boots up MUCH faster in 2.02. When booting in 2.10 it takes forever from the time the x cursor appears until the desktop is fully rendered. This makes up about 1/3 of the total boot time on my system.
that's what i've been thinkingThere's more to the 2.02 vs 2.10 speed issue than compression!
yes, it's a little slower using compression, but not much slower
on my machine, Abiword starts less than 1 second faster if it's completely uncompressed ... Seamonkey is less than 2 seconds faster
i think maybe you guys are barking up the wrong tree
- BarryK
- Puppy Master
- Posts: 9392
- Joined: Mon 09 May 2005, 09:23
- Location: Perth, Western Australia
- Contact:
Guys, there's one way to find out if LZMA is the culprit. Download 2.10
-pre-pre-alpha -- it is 2.10 but without the LZMA. Get it from:
http://www.puppyos.net/test/
... iso is about 82M I think.
...appreciate if someone can compare it.
Oh, and bootup with "puppy pfix=ram" as don't want the LZMA-compressed
devx_210.sfs to load.
-pre-pre-alpha -- it is 2.10 but without the LZMA. Get it from:
http://www.puppyos.net/test/
... iso is about 82M I think.
...appreciate if someone can compare it.
Oh, and bootup with "puppy pfix=ram" as don't want the LZMA-compressed
devx_210.sfs to load.
Here ya go
Here are my results of the test...BarryK wrote:Guys, there's one way to find out if LZMA is the culprit. Download 2.10
-pre-pre-alpha -- it is 2.10 but without the LZMA. Get it from:
http://www.puppyos.net/test/
... iso is about 82M I think.
...appreciate if someone can compare it.
Oh, and bootup with "puppy pfix=ram" as don't want the LZMA-compressed
devx_210.sfs to load.
To make things simple I booted off the live cd and ran Puppy in RAM. That's why the test results here are different from my previous results.
System:
IBM Thinkpad 600 PII-366 288MB RAM
boot: puppy acpi=force pci=noacpi pfix=ram
2.10 "Gold"
--------------------
abiword
real - 9.858
user - 3.540
sys - 5.780
mozstart
real - 26.982
user - 7.970
sys - 18.020
2.10 "pre-pre-alpha "
--------------------
abiword
real - 4.972
user - 3.530
sys - 1.000
mozstart
real - 10.443
user - 6.660
sys - 3.260
2.02
--------------------
abiword
real - 3.594
user - 2.180
sys - 0.910
mozstart
real - 10.279
user - 6.670
sys - 3.020
----------------------------
As you can see, 2.02 is still the champ but not by much. 2.10 PPA is almost the same.
Once again, the Xorg boot time for 2.10 Gold was MUCH slower than the other two.
Last edited by wpwood3 on Thu 21 Sep 2006, 16:36, edited 1 time in total.
running on a 900 mhz Duron, xorg, Jwm, frugal install, on a clean save file, apps run after a reboot
Boot time:
20s - 210 aa
25s - 210 gold
Abiword:
2s - 210 aa
3.4s - 210 gold
mozstart:
3.6s - 210 aa
7s - 210 gold
210 aa: puppy-2.10pre-pre-alpha.mozilla.iso
if the only significant difference between 210aa and 210 gold is LZMA, then LZMA causes these apps to start twice as slowly
booting is a little slower, but not much
in my earlier tests, running from the uncompressed files on the hard drive was almost the same as from LZMA compressed files
Boot time:
20s - 210 aa
25s - 210 gold
Abiword:
2s - 210 aa
3.4s - 210 gold
mozstart:
3.6s - 210 aa
7s - 210 gold
210 aa: puppy-2.10pre-pre-alpha.mozilla.iso
if the only significant difference between 210aa and 210 gold is LZMA, then LZMA causes these apps to start twice as slowly
booting is a little slower, but not much
in my earlier tests, running from the uncompressed files on the hard drive was almost the same as from LZMA compressed files