Dummies guide for putting Linux in a box

[FindersKeepers]

  TD ... will use that on an older intel box tomorrow to try it out.
Am committed to RHEL for the xeon server box, hopefully get it done tomorrow also.
Cheers, thanks for replying.

[tazzduke]

Hey FindersKeepers, all good.

Wow a server box, oh my, thats out of my league lol

Have fun

Regards

[Dingo]

The only Linux version that I successfully installed Nvidea drivers on was Centos because they have an excellent group that fix the packages for Centos.   Have a look at http://elrepo.org/tiki/tiki-index.php and especially http://elrepo.org/tiki/nvidia-detect

[FindersKeepers]

To be honest, I am overthinking everything coming from a windoze bloatware environment where you expect everything to go wrong with your permission. Linux is a paradigm shift in expectations. The RHEL install works and Linux is fully configurable miniware by comparison. It is not the full graphical experience I am used to from windoze but give it a bit of time. I am going to install different flavours just to see what works and what doesn't out of the box without configuration.

To be fair, windoze copied a lot of stuff from unix anyways so the Linux command line is like going back to the heady days of DOS.
A draft for comedic relief ... https://www.dropbox.com/s/jcod63qfh4zpvf0/IDIOTS%20GUIDE%20TO%20INSTALLING%20LINUX%20-%20draft%20revA.pdf

[FindersKeepers]

OK ... this is turning out to be a bit of fun. So I am going to explode my dummies document into a hobby project and examine the attached Linux distros (highlighted green). I will discover something useful or blow something up. The intent is to get BOINC running natively on Linux on each setup and compare the performance on 3 different pieces of hardware for each distro, 2 of which will include NVIDIA GFX. I use NVidia natively as that is what is supported by my cad projects.

Updated comedic relief of Linux experiences rev.B https://www.dropbox.com/s/it8kopkweuqgx3c/IDIOTS%20GUIDE%20TO%20INSTALLING%20LINUX%20-%20revB.pdf

Attached jpg for the 11 Linux distros I am going to try - picked randomly based on popularity in no particular order - plus RHEL v7 which I have started with.
note: boinc does not run stock standard on Linux without knowing which libraries to load, so I plan on trying to solve that for the various boinc/Linux versions without using WINE or a virtual box to run windoze.

[attachment deleted by admin]

[Sean]

This should be very interesting to see the results of.

Ubuntu is the main distro I use now, just runs easiest out-of-the-box. I have had one of the smaller distro running on a PII before (I think it was puppy?) but was super slow of course.

If you're feeling very brave and adventures you could also compile your own version of linux that is optimised for BOINC (minimum ram use etc) 

[FindersKeepers]

Yes ... looks like a bit of fun. Will also run some standard scripts to gauge speed and collect system data. I should add these will all be x64 bit distros - thinking the x32bit might be easier.
Does anyone have a copy of the Dotsch_UX Linux distro that includes Boinc? The site/iso is no longer online. The Boinc Berkeley makes mention of this version - perhaps this answers Seans query.
http://www.dotsch.de/Dotsch_UX

[edit1] vers 1.2 was the most recent Dotsch vers I think.
[edit2] don't suppose anyone has a compiled portable version of Boinc for Linux x64 using static libraries?? would be very handy about now ...
[edit3] the wayback machine gave me the filename, there appears to be a Linux x64 copy built for Linux Boinc here - [trimmed down] https://boinc.berkeley.edu/dl/dotsch_ux-12_x64.iso  <-- Sean, this might already have been done
[edit4] have attached the screenshot of the md5 checksum for anyone else interested
[edit5] the iso linked from wayback is the original with the correct md5 https://web.archive.org/web/20160316085550/http://aerospaceresearch.net/dotsch/dotsch_ux-12_x64.iso <-- looks like original ISO

Wayback links to the website -
https://web.archive.org/web/20160506021454/http://www.dotsch.de:80/boinc/Home.html
https://web.archive.org/web/20160316085550/http://www.dotsch.de/boinc/Dotsch_UX.html <-- how to links on bottom of page (read down)

To clarify .. this is the Lars Bausch ISO referred to by Berkeley Boinc https://boinc.berkeley.edu/trac/wiki/DownloadOther


[attachment deleted by admin]

[Dingo]

It is not too hard to compile BOINC for a specific machine.   I posted a thread a while ago about compiling on Centos here http://forum.boinc-australia.net/index.php?topic=1994.0

A couple of years ago a script by Daniel Carrion was created to Compile BOINC on Debian 64 bit.  Not sure if it still works, but it is a starting point and I  have listed it here:

Code Select Expand



#!/bin/bash
#
# Load up BOINC on 64-bit Debian
#
# Step 1 - Prepare BOINC for install
#

apt-get -y update
apt-get -y upgrade
apt-get -y install build-essential libcurl4-openssl-dev git automake libtool pkg-config libcurl4-openssl-dev libnotify-bin libnotify-dev
mkdir src
cd src
git clone https://github.com/BOINC/boinc boinc_repo
cd boinc_repo
./_autosetup
./configure --disable-server --enable-client CXXFLAGS="-O3 "
make -j 4
cd packages/generic/sea/
make -j 4
cp -r BOINC /tmp
#
# Step 2 - Install BOINC
#
useradd -d /opt/boinc -m boinc
su - boinc -c "mkdir bin"
su - boinc -c "cp /tmp/BOINC /boinc bin/"
su - boinc -c "touch gui_rpc_auth.cfg"
su - boinc -c "chmod 600 gui_rpc_auth.cfg"
touch /etc/init.d/boinc
echo "i" >> /etc/init.d/boinc
echo "# BOINC start/stop" >> /etc/init.d/boinc
echo "#" >> /eddtc/init.d/boinc
echo 'if [ $1 == start ];then' >> /etc/init.d/boinc
echo "   su - boinc -c 'boinc --daemon --allow_remote_gui_rpc'" >> /etc/init.d/boinc
echo "fi" >> /etc/init.d/boinc
echo 'if [ $1 == stop ];then' >> /etc/init.d/boinc
echo "   su - boinc -c 'pkill boinc'" >> /etc/init.d/boinc
echo "fi" >> /etc/init.d/boinc
chmod 755 /etc/init.d/boinc
update-rc.d boinc defaults
#
# Step 3 - Get some firewall action

#
apt-get -y install iptables-persistent
iptables -P INPUT DROP
iptables -P FORWARD DROP
iptables -P OUTPUT ACCEPT
iptables -A INPUT -i lo -j ACCEPT
iptables -A INPUT -m state --state RELATED,ESTABLISHED -j ACCEPT
iptables -A INPUT -p tcp --dport 22 -j ACCEPT
iptables -A INPUT -p tcp --dport 31416 -j ACCEPT
iptables -A INPUT -p icmp -j ACCEPT
iptables-save > /etc/iptables/rules
#
# Step 4 Ask user for BOINC password
#
echo ""
echo "Enter the password you want to connect to BOINC using:"
read BOINC_GUI_PWD
echo "$BOINC_GUI_PWD" > /opt/boinc/gui_rpc_auth.cfg
#
# Step 4 - Reboot the sucker and check!
#
shutdown -r now


[FindersKeepers]

Nice one Dingo, I'll be taking that for a spin for sure! 

Sean - that Dotsch UX version by Lars goes a long way to answering your question. It can run live from USB - what else do you need?

[kashi]

If you can find a way to update the old Linux version contained in Dotsch UX that would be very handy.

I used it for years but stopped using it a few years ago because it became incompatible with a number of projects. I have read that good practice is to "statically link" project applications to the libraries required but not all project developers do this. Because Dotsch UX has not been updated for so long the libraries it contains are not always new enough to support project applications that are not statically linked.

It was supposed to be possible to update the Linux version in Dotsch UX but I was unable to do this.

Don't know what statically link actually means or what libstdc++ libraries do, just remember that if I didn't have the right ones some projects wouldn't run.

[FindersKeepers]

First successful test run started !!! Boinc 7.6.33 running on Manjero 17.0 with Xfce front end using the pacman package installer. None of the instructions worked from the outset and had to upgrade the system and reinstall boinc twice. After the 3rd attempt found the boinc binaries to start the client without error and is now running via the manager window from terminal. Running on POGS to benchmark against windows 7 using exact same machine previously. Should have some results tomorrow. This is way more fun than I thought ....


If this proves the thesis, will expand to another machine and leave it running ... !(breath.held)

[Dingo]

You have to remember that POGS has different sized tasks and they take a different amount of time.  If you look under "View" Tasks then "Valid" and under "Tasks" you can see how many observations it has done.  Trying to compare a task that has done 15 observations and one 6 is not the way to go.   Just in case you had't noticed this. 

Code Select Expand

Stderr output

<core_client_version>7.4.42</core_client_version>
<![CDATA[
<stderr_txt>
10:11:25 (13048): wrapper (7.5.26014): starting
10:11:25 (13048): wrapper: running ./fit_sed (1 filters.dat observations.dat)
10:18:45 (13048): ./fit_sed exited; CPU time 436.406250
10:18:45 (13048): wrapper: running ./fit_sed (2 filters.dat observations.dat)
10:26:10 (13048): ./fit_sed exited; CPU time 436.359375
10:26:10 (13048): wrapper: running ./fit_sed (3 filters.dat observations.dat)
10:33:46 (13048): ./fit_sed exited; CPU time 446.437500
10:33:46 (13048): wrapper: running ./fit_sed (4 filters.dat observations.dat)
10:41:05 (13048): ./fit_sed exited; CPU time 436.296875
10:41:05 (13048): wrapper: running ./fit_sed (5 filters.dat observations.dat)
10:48:27 (13048): ./fit_sed exited; CPU time 436.406250
10:48:27 (13048): wrapper: running ./fit_sed (6 filters.dat observations.dat)
10:55:45 (13048): ./fit_sed exited; CPU time 436.312500
10:55:45 (13048): wrapper: running ./fit_sed (7 filters.dat observations.dat)
11:03:13 (13048): ./fit_sed exited; CPU time 446.015625
11:03:13 (13048): wrapper: running ./concat (7 output.fit)
11:03:14 (13048): ./concat exited; CPU time 0.000000
11:03:14 (13048): called boinc_finish(0)

</stderr_txt>
]]>


Code Select Expand

<core_client_version>7.4.42</core_client_version>
<![CDATA[
<stderr_txt>
09:01:50 (14048): wrapper (7.5.26014): starting
09:01:50 (14048): wrapper: running ./fit_sed (1 filters.dat observations.dat)
09:09:25 (14048): ./fit_sed exited; CPU time 446.468750
09:09:25 (14048): wrapper: running ./fit_sed (2 filters.dat observations.dat)
09:16:58 (14048): ./fit_sed exited; CPU time 446.500000
09:16:58 (14048): wrapper: running ./fit_sed (3 filters.dat observations.dat)
09:24:36 (14048): ./fit_sed exited; CPU time 456.765625
09:24:36 (14048): wrapper: running ./fit_sed (4 filters.dat observations.dat)
09:32:04 (14048): ./fit_sed exited; CPU time 446.562500
09:32:04 (14048): wrapper: running ./fit_sed (5 filters.dat observations.dat)
09:39:35 (14048): ./fit_sed exited; CPU time 446.578125
09:39:35 (14048): wrapper: running ./fit_sed (6 filters.dat observations.dat)
09:47:11 (14048): ./fit_sed exited; CPU time 446.515625
09:47:11 (14048): wrapper: running ./fit_sed (7 filters.dat observations.dat)
09:54:37 (14048): ./fit_sed exited; CPU time 436.234375
09:54:37 (14048): wrapper: running ./fit_sed (8 filters.dat observations.dat)
10:02:09 (14048): ./fit_sed exited; CPU time 446.437500
10:02:09 (14048): wrapper: running ./fit_sed (9 filters.dat observations.dat)
10:09:53 (14048): ./fit_sed exited; CPU time 456.843750
10:09:53 (14048): wrapper: running ./fit_sed (10 filters.dat observations.dat)
10:17:33 (14048): ./fit_sed exited; CPU time 456.656250
10:17:33 (14048): wrapper: running ./fit_sed (11 filters.dat observations.dat)
10:25:12 (14048): ./fit_sed exited; CPU time 456.734375
10:25:12 (14048): wrapper: running ./fit_sed (12 filters.dat observations.dat)
10:32:55 (14048): ./fit_sed exited; CPU time 456.515625
10:32:55 (14048): wrapper: running ./fit_sed (13 filters.dat observations.dat)
10:40:42 (14048): ./fit_sed exited; CPU time 456.531250
10:40:42 (14048): wrapper: running ./fit_sed (14 filters.dat observations.dat)
10:48:18 (14048): ./fit_sed exited; CPU time 446.375000
10:48:18 (14048): wrapper: running ./fit_sed (15 filters.dat observations.dat)
10:55:57 (14048): ./fit_sed exited; CPU time 456.484375
10:55:57 (14048): wrapper: running ./fit_sed (16 filters.dat observations.dat)
11:03:42 (14048): ./fit_sed exited; CPU time 456.515625
11:03:42 (14048): wrapper: running ./fit_sed (17 filters.dat observations.dat)
11:11:11 (14048): ./fit_sed exited; CPU time 446.406250
11:11:11 (14048): wrapper: running ./fit_sed (18 filters.dat observations.dat)
11:18:38 (14048): ./fit_sed exited; CPU time 446.078125
11:18:39 (14048): wrapper: running ./fit_sed (19 filters.dat observations.dat)
11:26:07 (14048): ./fit_sed exited; CPU time 446.343750
11:26:07 (14048): wrapper: running ./fit_sed (20 filters.dat observations.dat)
11:33:34 (14048): ./fit_sed exited; CPU time 436.031250
11:33:34 (14048): wrapper: running ./fit_sed (21 filters.dat observations.dat)
11:40:54 (14048): ./fit_sed exited; CPU time 435.984375
11:40:54 (14048): wrapper: running ./concat (21 output.fit)
11:40:55 (14048): ./concat exited; CPU time 0.000000
11:40:55 (14048): called boinc_finish(0)

</stderr_txt>
]]>


[FindersKeepers]

Yes Dingo, I did notice that early on. Is why I work on averages of the last 40 valid observations, whatever they happen to be. To measure max performance I average the same task credits per unit of time. For every machine I compared I used their last 40 valid results.

The initial observations look promising for Manjaro. Average of 9% increase in output from the identical machine running windoze, now running Linux. Will leave it run today and work on another machine running test no.2 next. For clarity, ave time per credit has reduced from 24.5 to 22.5 secs which increases daily yield.

[FindersKeepers]

Results confirm +8.5% increase in output from Win7 x64 to Linux x64 using i5-2400. This translates to an extra 1280 credits per day output for this machine running POGS.
Linux Manjaro v17.0, pacman updates, using pacman boinc package. Graphic user interface running boinc 7.6.33.

Side by side valid to compare ...
http://pogs.theskynet.org/pogs/results.php?hostid=798189&offset=0&show_names=0&state=4&appid= <= running win7 x64
http://pogs.theskynet.org/pogs/results.php?hostid=801398&offset=0&show_names=0&state=4&appid= <= running Linux x64

[FindersKeepers]

Just kicked off a dual xeon x5690 box on redhat running from command using Dingo's script above to compile the boinc service (worked a treat). I didn't get a chance to tidy up the network settings before it kicked off so have left it running. Still some things to sort out. No graphics card on this box so is all CPU only. Have 4 separate boot drives to load 4 different Linux images onto for a side by side boot/run/joyride comparison. Starting with redhat as you can tell ....

http://pogs.theskynet.org/pogs/hosts_user.php?userid=458685
In your pogs account login there is a link to insert an XML file into the boinc run directory, this loads the project automagically when the boinc service starts up.

Starting with no H/T for a baseline - 12 cores running are at 100%.

[EDIT] Discovered having a graphics card onboard with windoze penalises CPU crunching by about 5% on the 4 core i5's ... by virtue of some cpu overhead. Removing it improved output (staying with just the minimal m/b onchip GPU, no external GPU). As a matter of habit I plan on using only CPU machines for CPU projects (no plugin GFX). If you don't get the same results with a GFX card installed, try removing it if you are doing CPU only projects (rosetta/pogs). As kashi said previously, GPU crunching usually occupies one CPU thread per installed GPU card. I will be testing H/T and no H/T in this side project to measure differences and will tabulate everything for comedic relief. As I noted on a different thread, for dual xeon boxes I have seen a 20% increase in output turning H/T off. Something I want to test further  .... as time and patience allows. The numbers for the i5 machine are only getting better running Manjaro. Rock solid consistency in the numbers from the i5 so far. With RHEL I have noticed the gnome process consuming CPU time - I think I will switch Gnome off or pick simpler distro interface - the Xfce on Manjaro looks OK. Am setting workspaces to 1 on everything I do with Boinc.