[9fans] run and mount

[9fans] run and mount

[Sergey Reva]

Hello Fans,

Can someone suggest me way, how make this sample much better?

For now I should use network service to run 'slave' on CPU server,
but I wish use cpu command for that, how can I do that?

There is multihost decryptor for xor algoritm...
http://rs-rlab.narod.ru/files/slave.c
http://rs-rlab.narod.ru/files/master.c

Thanks in advance!
Sergey
-- 
http://rs-rlab.narod.ru                          mailto:rs_rlab@mail.ru

Re: [9fans] run and mount

[arisawa]

> Can someone suggest me way, how make this sample much better?
>
> For now I should use network service to run 'slave' on CPU server,
> but I wish use cpu command for that, how can I do that?
>
> There is multihost decryptor for xor algoritm...
> http://rs-rlab.narod.ru/files/slave.c
> http://rs-rlab.narod.ru/files/master.c
>

I couldn't open slave.c and master.c. That was blank.
Therefore I don't know what you really want to ask.

If you try something like grid computing, please look
	$home/src/taskfs
of my host "co.aichi-u.ac.jp"
You can login* this host using bell-labs's source account.
	* use cpu command to login

The rc script
	$home/src/taskfs/run
might be helpful for you to run "slave" on cpu server.

Someone else (for example YAMANASHI) have some other solutions.

Kenji Arisawa

Re: [9fans] run and mount

[Sergey Reva]

Hello Fans

Solution is there
http://www.tip9ug.jp/who/rlab/files/mtest.c

Sergey
-- 
http://rs-rlab.narod.ru                            mailto:rs_rlab@mail.ru

Re: [9fans] run and mount

[YAMANASHI Takeshi]

> > There is multihost decryptor for xor algoritm...
> > http://rs-rlab.narod.ru/files/slave.c
> > http://rs-rlab.narod.ru/files/master.c
 :
> Someone else (for example YAMANASHI) have some other solutions.

As you mentioned me...  I have a tirivial system which
accomplishes a SETI like computing frame work at:
	http://www.tip9ug.jp/who/nashi/9grid/taskbag.tgz

You write a program which reads data from fd 0 and writes
the result to fd 1 and prepare a set of input to feed to
the program, then you can distribute the program and data
using taskbag and hope 9fans' volunteers to crunch numbers
for your set of task.

As a matter of fact, I'm going to ask 9fans to volunteer
cpu seconds to compute 25 queen problem using this. :)
-- 

Re: [9fans] run and mount

[Sergey Reva]

Hello fans

Mirror of my public plan 9 files available at tip9ug.jp:
http://www.tip9ug.jp/who/rlab/files/slave.c
http://www.tip9ug.jp/who/rlab/files/master.c

http://www.tip9ug.jp/who/rlab/files/*
drawterm.zip - drawterm with Unicode for Windows
cap.c - screen capture
devi2c.c - nvram emulation with 24cXXX
life.c - "Life" for Plan 9
ln.c (line numerator)

> http://www.tip9ug.jp/who/nashi/9grid/taskbag.tgz
> taskfs
Thanks Kenji and Yamanashi, i'll now check this

Sergey
-- 
http://rs-rlab.narod.ru                            mailto:rs_rlab@mail.ru

Re: [9fans] run and mount

[arisawa]

> As a matter of fact, I'm going to ask 9fans to volunteer
> cpu seconds to compute 25 queen problem using this. :)

You can estimate for computing power to do this.

The task to get solution of N-queens problem can be divided
into small independent pieces of tasks.
We can get Open-MP style program qn24b-version1/omp
from http://www.yuba.is.uec.ac.jp/~kis/nq/index.htm

I transformed this program so that we can execute such like
         queen n i:j
this means n-queens problem of task(piece) range i to j
(You can get this program in co.aichi-u.ac.jp using cpu command
and look /usr/arisawa/src/taskfs/Q/src/)

I measured the time
	time queen n 0:1
0:1 means the first one piece.

The result measured on 2.4GHz Xeon
         n       time    # of pieces
         21      5s       41176
         22      33s     49480
         23      238s    64072
         24      1751s   75516
         25      13590s  95472
note that the time required to compute one piece increases rapidly

-bash$ python
Python 2.3 (#1, Sep 13 2003, 00:49:11)
[GCC 3.3 20030304 (Apple Computer, Inc. build 1495)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
 >>> 13590*95472
1297464480
 >>> 1297464480/(60*60*24*365)
41
 >>>

41 years on single computer!
500 volunteers who can offer one month?
I want rather 2000 9grid computers...

Kenji Arisawa

Re: [9fans] run and mount

[YAMANASHI Takeshi]

> You can estimate for computing power to do this.
 :
> The result measured on 2.4GHz Xeon
 :
> 41 years on single computer!

Hehe, I haven't seriously estimated the time needed
to complete the job.  41 years?  That's great.
I would see the possibitily of volunteer based
computing when it got as common as p2p file sharing. :)

> 500 volunteers who can offer one month?

Hope v9fs could be a help to distribute the task to
linux nodes.
-- 

Re: [9fans] run and mount

[arisawa]

>
>> 500 volunteers who can offer one month?
>
> Hope v9fs could be a help to distribute the task to
> linux nodes.

You are trying to achieve world record...
But impossible without help of *regular* unix users.
This might be a good exercise to have a simple interface
between Plan 9 and unix.

Kenji Arisawa

Re: [9fans] run and mount

[Ronald G. Minnich]

On Tue, 2 Aug 2005 arisawa@ar.aichi-u.ac.jp wrote:

> You are trying to achieve world record...
> But impossible without help of *regular* unix users.

It would be interesting to try to use xcpu to provide donor nodes. 

ron

Re: [9fans] run and mount

[arisawa]

> It would be interesting to try to use xcpu to provide donor nodes.
>

Thanks ron, but where is xcpu ?

Kenji Arisawa

Re: [9fans] run and mount

[Ronald G. Minnich]

On Tue, 2 Aug 2005 arisawa@ar.aichi-u.ac.jp wrote:

> Thanks ron, but where is xcpu ?

working on it ... although it may be in sources, not sure. I'll try to 
find out. Vic Zandy wrote it, I am changing it a bit.

ron

Re: [9fans] run and mount

[arisawa]

Hello,

I received private mail from Sergey Reva, but the replay was rejected:

> Your request ``mail net!mail.ru rs_rlab '' failed (code smtp 165353: 
> Permanent Failure).

therefore I post the mail to 9fans.
the contents is bellow.

> Sorry, but I can't login to your cpu server, can you put taskfs to
> sources or send it directly to me?
>

it seems my co.aichi-u.ac.jp in trouble...

I have put taskfs as grid.tgz into sources.
look /n/sources/contrib/arisawa

grid.tgz is a grid kit for Plan 9 that is under development
and should be tested using more examples.

Please give me your suggestion.

Kenji Arisawa