[ale] RE: [Fwd: [Fwd: EnFuzion 6.0 and apps]]

Kurt A. Brust cltkbrust at telocity.com
Fri Dec 1 16:07:05 EST 2000


Yes, this is one of the reason we can "failover", if a node in the cluster goes
down, the whole job DOES NOT need to restart, just the job on the certain
node

On Fri, 01 Dec 2000, Chris Fowler wrote:
> 
> Understood!  
> 
> Will all the apps in the cluster use a shared memory segment.  This will
> allow 'x' to constantly increment without starting from 0 at each node.
> 
> Chris
> 
> -----Original Message-----
> From: Kurt A. Brust [mailto:cltkbrust at telocity.com]
> Sent: Friday, December 01, 2000 1:16 PM
> To: Chris Fowler
> Cc: 'ale at ale.org'
> Subject: [ale] RE: [Fwd: [Fwd: EnFuzion 6.0 and apps]]
> 
> 
> Enfuzion is written to take advantage of any application, what I mean by
> that
> is, when you install enfuzion, it installs a 20-25k applet (size of file is
> determined by platform). You then need to install the .EXE (exacutable) on
> each
> node. With in the scripting of enfuzion, you determine what the paramters of
> the application you are running, for example, node 1 run with these
> paramters,
> node 2 run these, node 3 run these, etc..  In the case of a dual or quad
> processer box, the root server would run each of those processers in a
> single
> mode, as a cluster. for example, cpu 1 can be handling certain paramters,
> cpu 2
> code be handing different paramters, etc....
> 
> The "MAGIC" as you put it, happens from with in enfuzion, the way enfuzion
> handles the application.
> 
> Let me know if this helps you out.
> 
> 
> On Fri, 01 Dec 2000, Chris Fowler wrote:
> > 
> > This is the application:
> > 
> > 
> > #include <stdio.h>
> > 
> > 
> > void
> > main(void)
> > {
> > 	int X=0;
> > 
> > 	for (x=0;x=0;x++)  /* This will allways be true */
> > 	{
> > 		printf("X = %i",x);
> > 	}
> > 
> > 	return 0;
> > }
> > 
> > 
> > In this case the app will only run as fast as CPU1 will allow it to run.
> > There is no clone() call so it will not run on the 2nd CPU of an SMP based
> > system.
> > 
> > Now,  
> > 
> > The only way around this is for your software to implement some sort of
> > *MAGIC* that allows parallel execution on multiple nodes of this simple
> > code.  Many graphics programs are multi-threaded, this makes it possible
> for
> > them to benefit from a cluster.  But many simple apps that need more
> > horsepower that are single threaded will not benefit from SMP.  If that is
> > true, how can the same apps that do not use CPU2 on SMP use NODE1-NODE100
> on
> > EnFuzion?
> > 
> > Chris
> > 
> > -----Original Message-----
> > From: Kurt A. Brust [mailto:cltkbrust at telocity.com]
> > Sent: Friday, December 01, 2000 12:06 PM
> > To: Jacob Webb
> > Cc: cfowler at linuxiceberg.com
> > Subject: Re: [Fwd: [Fwd: EnFuzion 6.0 and apps]]
> > 
> > 
> > Chris,
> > 
> > I hope this helps.
> > 
> > > If I have a single threaded app that uses as much CPU as it can, will it
> > use
> > as much as it can on each nod of the cluster.  a 100 node can increase
> > excution
> > by 100x times?  
> > 
> > Enfuzion does not speed up the "application" , Enfuzion excelarates the
> > output
> > of the application. (Ex: If you have an application that's end result is a
> > text
> > file, enfuzion can break up and distribute parts of that text file to
> other
> > processer's, which in turn speeds up the processing of that file)
> > 
> > Yes, to answer your question it is linear, 10 nodes = 10 times as fast ...
> > 100
> > nodes = 100 times as fast.
> > 
> > 
> > 
> > 
> > 
> > n Fri, 01 Dec 2000, Jacob Webb wrote:
> > > 
> > > Hey Kurt,
> > > 
> > > I hope all is going well.  The lead below is asking some really
> > > technical questions that I can't answer.  Would you mind touching base
> > > with him.
> > > 
> > > Thanks, Jacob
> > > 
> > 
> > > X-Sieve: cmu-sieve 1.3
> > > Return-Path: <sward at turbolinux.com>
> > > Received: (from nobody at localhost)
> > > 	by mail.us.tlan (8.9.3/8.9.3) id IAA09123
> > > 	for <jwebb at munch-it.turbolinux.com>; Fri, 1 Dec 2000 08:47:03 -0800
> > > Received: from mail.turbolinux.com(38.170.88.25)
> > >  via SMTP by mail.us.tlan, id smtpdnqQIss; Fri Dec  1 08:46:54 2000
> > > Received: from mail.us.tlan (munch-it.turbolinux.com [38.170.88.129])
> > > 	by mail.turbolinux.com (8.9.3/8.9.3) with ESMTP id IAA25218
> > > 	for <jwebb at turbolinux.com>; Fri, 1 Dec 2000 08:46:53 -0800
> > > Received: (from nobody at localhost)
> > > 	by mail.us.tlan (8.9.3/8.9.3) id IAA09112
> > > 	for <jwebb at turbolinux.com>; Fri, 1 Dec 2000 08:46:53 -0800
> > > Received: from sward-lt.office.us.tlan(172.16.6.134), claiming to be
> > "turbolinux.com"
> > >  via SMTP by mail.us.tlan, id smtpdRyXlmi; Fri Dec  1 08:46:44 2000
> > > Sender: sward at turbolinux.com
> > > Message-ID: <3A27E3C4.8B4A0B81 at turbolinux.com>
> > > Date: Fri, 01 Dec 2000 09:45:40 -0800
> > > From: Steve Ward <sward at turbolinux.com>
> > > Organization: TurboLinux
> > > X-Mailer: Mozilla 4.72 [en] (X11; I; Linux 2.2.14-7 i686)
> > > X-Accept-Language: en
> > > MIME-Version: 1.0
> > > To: jwebb at turbolinux.com
> > > Subject: [Fwd: EnFuzion 6.0 and apps]
> > > Content-Type: multipart/mixed;
> > >  boundary="------------4104F3DC669D9A0560F0DC15"
> > > 
> > > This is a multi-part message in MIME format.
> > > --------------4104F3DC669D9A0560F0DC15
> > > Content-Type: text/plain; charset=us-ascii
> > > Content-Transfer-Encoding: 7bit
> > > 
> > > 
> > > 
> > > --------------4104F3DC669D9A0560F0DC15
> > > Content-Type: message/rfc822
> > > Content-Transfer-Encoding: 7bit
> > > Content-Disposition: inline
> > > 
> > > X-Sieve: cmu-sieve 1.3
> > > Return-Path: <nobody at www1.turbolinux.com>
> > > Received: (from nobody at localhost)
> > > 	by mail.us.tlan (8.9.3/8.9.3) id IAA08296
> > > 	for <sward at munch-it.turbolinux.com>; Fri, 1 Dec 2000 08:21:12 -0800
> > > Received: from mail.turbolinux.com(38.170.88.25)
> > >  via SMTP by mail.us.tlan, id smtpdTSTj7l; Fri Dec  1 08:21:11 2000
> > > Received: from www1.turbolinux.com (IDENT:root at www1.turbolinux.com
> > [38.170.88.20])
> > > 	by mail.turbolinux.com (8.9.3/8.9.3) with ESMTP id IAA25011
> > > 	for <sales at turbolinux.com>; Fri, 1 Dec 2000 08:21:10 -0800
> > > Received: (from nobody at localhost)
> > > 	by www1.turbolinux.com (8.9.3/8.9.3) id IAA30383;
> > > 	Fri, 1 Dec 2000 08:21:10 -0800
> > > Date: Fri, 1 Dec 2000 08:21:10 -0800
> > > Message-Id: <200012011621.IAA30383 at www1.turbolinux.com>
> > > To: sales at turbolinux.com
> > > Subject: EnFuzion 6.0 and apps
> > > From: "Chris D Fowler" <cfowler at linuxiceberg.com>
> > > 
> > >  IP: 160.77.2.161
> > > URL: http://www.turbolinux.com/forms/email.php3
> > > 
> > > Chris D Fowler
> > > 1060 Windward Ridge Parkway, Suite 100
> > > Alpharetta, GA 30005
> > > USA
> > > 
> > > Phone: 7706251627
> > > Fax: 0000000000
> > > Email: cfowler at linuxiceberg.com
> > > 
> > > Product: EnFuzion 6.0
> > > Version: 6.0
> > > Serial#: 
> > > 
> > > I\'m looking at the EnFuzion 6.0 product and it states that I do not
> need
> > to make any changes to my apps to get them to take advantage of the
> cluster.
> > 
> > > 
> > > My question:
> > > 
> > > If I have a single threaded app that uses as much CPU as it can, will it
> > use as much as it can on each nod of the cluster.  a 100 node can increase
> > excution by 100x times?  
> > > 
> > > Thanks,
> > > Chris
> > > 
> > > --------------4104F3DC669D9A0560F0DC15
> > > Content-Type: text/x-vcard; charset=us-ascii;
> > >  name="sward.vcf"
> > > Content-Transfer-Encoding: 7bit
> > > Content-Description: Card for Steve Ward
> > > Content-Disposition: attachment;
> > >  filename="sward.vcf"
> > > 
> > > begin:vcard 
> > > n:Ward;Steve
> > > tel;fax:650-228-5001
> > > tel;work:650-228-5079
> > > x-mozilla-html:FALSE
> > > adr:;;8000 Marina Blvd. 3rd Floor	;Brisbane;CA;94005;USA
> > > version:2.1
> > > email;internet:sward at turbolinux.com
> > > title:Account Manager
> > > x-mozilla-cpt:;-30400
> > > fn:Steve Ward
> > > end:vcard
> > > 
> > > --------------4104F3DC669D9A0560F0DC15--
> > > 
> > > 
> > 
> 
> ----------------------------------------
> Content-Type: text/html; name="unnamed"
> Content-Transfer-Encoding: quoted-printable
> Content-Description: 
> ----------------------------------------
> --
> To unsubscribe: mail majordomo at ale.org with "unsubscribe ale" in message
> body.
> 

----------------------------------------
Content-Type: text/html; name="unnamed"
Content-Transfer-Encoding: quoted-printable
Content-Description: 
----------------------------------------
--
To unsubscribe: mail majordomo at ale.org with "unsubscribe ale" in message body.





More information about the Ale mailing list