From marbles-isi@mailman.isi.edu  Sat Sep  1 01:19:45 2001
From: marbles-isi@mailman.isi.edu (Martin Frank)
Date: Fri, 31 Aug 2001 17:19:45 -0700
Subject: [Marbles-isi] Marbles2 problemgen directory
Message-ID: <5.1.0.14.0.20010831171607.00b063d8@tnt.isi.edu>

I checked in a marbles2/problemgen directory that is an empty shell for 
problem generator invocation. "make fresh" will get it, but something less 
drastic will likely also do it.

Cheers, yours, Martin

P.S.: There is a shell script for invoking it in that directory, like this:

$ sh runproblemgen.sh
starting the Marbles2 problem generator

M2PgParameters {
   numberOfTasks 5
   numberOfRequirements 2
   numberOfResources 12
}
0 tasks, 0 resources
<ProblemDefinition>
   <Nickname>Marbles2 problem generator - synthetic problem</Nickname>
   <Description>
     #tasks: 5, #requirements: 2, #resources: 12, generated on Fri Aug 31 
17:18:5
3 PDT 2001
   </Description>
   <SpResources>
   </SpResources>
   <SpTasks>
   </SpTasks>
</ProblemDefinition>
done with the Marbles2 problem generator


From marbles-isi@mailman.isi.edu  Wed Sep 19 22:57:02 2001
From: marbles-isi@mailman.isi.edu (Martin Frank)
Date: Wed, 19 Sep 2001 14:57:02 -0700
Subject: [Marbles-isi] we have a Marbles2 solver!
Message-ID: <5.1.0.14.0.20010919144501.00ae1ad0@tnt.isi.edu>

All -

Min has checked in the first Marbles2 solver (centralized and not real-time 
for now, but at least on par with SNAP's current algorithm). Say "make 
cvsupdate" in marbles2 to get it. Run the shell script 
marbles2/main/marbles2.sh. It solves the "miniyuma" problem I made up by 
hand fine, as follows:

   invoking serialcrawler on miniyuma (7 resources, 3 tasks) for 1 minute
   serialcrawler returned a valid solution of value 200 in 4 seconds

Seong-Bin is working on a problem generator which we can use to test Min's 
solver more thoroughly.

Alejandro may work on another type of solver which translates a Marbles 2 
into a Marbles 1 problem statement and back.

Min will work on a "problem verifier" package that you can give a problem 
definition and it will tell you if there are any tasks in it that could 
never be fulfilled, even if there were no competing tasks; it will also 
tell you if there are any useless resources that can never be used by any 
task . He'll also work on a "solution verifier" package that given a 
problem definition and a solution will tell you if the solution really is 
one (no double-booking of resources, no booking of resources outside their 
availability, and so on.)

I will intermittenty improve on my SNAP-file-to-marbles2-problem generator 
that currently produces only simulator tasks so we have realistic domain 
problems for Marbles2 testing.

Cheers, yours, Martin


From marbles-isi@mailman.isi.edu  Wed Sep 19 23:12:52 2001
From: marbles-isi@mailman.isi.edu (Bob Neches)
Date: Wed, 19 Sep 2001 15:12:52 -0700
Subject: [Marbles-isi] we have a Marbles2 solver!
Message-ID: <41C31855E83ED31199D300C04F57A0734C1A63@hermes.isi.edu>

Cool!

An ATTEND question: with the information in the representation that Min will
be analzying to identify unfullfillable tasks and useless resources, would
this support any of the partitioning work we've discussed in ATTEND?

-- Bob

-----Original Message-----
From: Martin Frank [mailto:frank@ISI.EDU]
Sent: Wednesday, September 19, 2001 2:57 PM
To: marbles-isi@ISI.EDU
Cc: mcai@ISI.EDU
Subject: [Marbles-isi] we have a Marbles2 solver!


All -

Min has checked in the first Marbles2 solver (centralized and not real-time 
for now, but at least on par with SNAP's current algorithm). Say "make 
cvsupdate" in marbles2 to get it. Run the shell script 
marbles2/main/marbles2.sh. It solves the "miniyuma" problem I made up by 
hand fine, as follows:

   invoking serialcrawler on miniyuma (7 resources, 3 tasks) for 1 minute
   serialcrawler returned a valid solution of value 200 in 4 seconds

Seong-Bin is working on a problem generator which we can use to test Min's 
solver more thoroughly.

Alejandro may work on another type of solver which translates a Marbles 2 
into a Marbles 1 problem statement and back.

Min will work on a "problem verifier" package that you can give a problem 
definition and it will tell you if there are any tasks in it that could 
never be fulfilled, even if there were no competing tasks; it will also 
tell you if there are any useless resources that can never be used by any 
task . He'll also work on a "solution verifier" package that given a 
problem definition and a solution will tell you if the solution really is 
one (no double-booking of resources, no booking of resources outside their 
availability, and so on.)

I will intermittenty improve on my SNAP-file-to-marbles2-problem generator 
that currently produces only simulator tasks so we have realistic domain 
problems for Marbles2 testing.

Cheers, yours, Martin

_______________________________________________
marbles-isi mailing list
marbles-isi@mailman.isi.edu
http://mailman.isi.edu/mailman/listinfo/marbles-isi

From marbles-isi@mailman.isi.edu  Wed Sep 19 23:14:24 2001
From: marbles-isi@mailman.isi.edu (Alejandro Bugacov)
Date: Wed, 19 Sep 2001 15:14:24 -0700
Subject: [Marbles-isi] we have a Marbles2 solver!
References: <5.1.0.14.0.20010919144501.00ae1ad0@tnt.isi.edu>
Message-ID: <3BA918C0.9635FC1F@isi.edu>

great! 
We should have Seongbin's problem generator by the end of this week 
and probably also a first pass of a random solver that we are working 
on with Donghan. 
alejandro.  

Martin Frank wrote:
> 
> All -
> 
> Min has checked in the first Marbles2 solver (centralized and not real-time
> for now, but at least on par with SNAP's current algorithm). Say "make
> cvsupdate" in marbles2 to get it. Run the shell script
> marbles2/main/marbles2.sh. It solves the "miniyuma" problem I made up by
> hand fine, as follows:
> 
>    invoking serialcrawler on miniyuma (7 resources, 3 tasks) for 1 minute
>    serialcrawler returned a valid solution of value 200 in 4 seconds
> 
> Seong-Bin is working on a problem generator which we can use to test Min's
> solver more thoroughly.
> 
> Alejandro may work on another type of solver which translates a Marbles 2
> into a Marbles 1 problem statement and back.
> 
> Min will work on a "problem verifier" package that you can give a problem
> definition and it will tell you if there are any tasks in it that could
> never be fulfilled, even if there were no competing tasks; it will also
> tell you if there are any useless resources that can never be used by any
> task . He'll also work on a "solution verifier" package that given a
> problem definition and a solution will tell you if the solution really is
> one (no double-booking of resources, no booking of resources outside their
> availability, and so on.)
> 
> I will intermittenty improve on my SNAP-file-to-marbles2-problem generator
> that currently produces only simulator tasks so we have realistic domain
> problems for Marbles2 testing.
> 
> Cheers, yours, Martin
> 
> _______________________________________________
> marbles-isi mailing list
> marbles-isi@mailman.isi.edu
> http://mailman.isi.edu/mailman/listinfo/marbles-isi

-- 
Alejandro Bugacov                                
USC Information Sciences Institute  
4676 Admiralty Way, Suite 1001                                
Marina del Rey, CA  90292               
Voice:   (310) 448-8269
Fax:     (310) 823-6714
http://www.isi.edu/~bugacov

From marbles-isi@mailman.isi.edu  Wed Sep 19 23:28:03 2001
From: marbles-isi@mailman.isi.edu (Bob Neches)
Date: Wed, 19 Sep 2001 15:28:03 -0700
Subject: [Marbles-isi] we have a Marbles2 solver!
Message-ID: <41C31855E83ED31199D300C04F57A0734C1A64@hermes.isi.edu>

sounds like a lot of things are starting to come together

-----Original Message-----
From: Alejandro Bugacov [mailto:bugacov@ISI.EDU]
Sent: Wednesday, September 19, 2001 3:14 PM
To: marbles-isi@mailman.isi.edu
Subject: Re: [Marbles-isi] we have a Marbles2 solver!


great! 
We should have Seongbin's problem generator by the end of this week 
and probably also a first pass of a random solver that we are working 
on with Donghan. 
alejandro.  

Martin Frank wrote:
> 
> All -
> 
> Min has checked in the first Marbles2 solver (centralized and not
real-time
> for now, but at least on par with SNAP's current algorithm). Say "make
> cvsupdate" in marbles2 to get it. Run the shell script
> marbles2/main/marbles2.sh. It solves the "miniyuma" problem I made up by
> hand fine, as follows:
> 
>    invoking serialcrawler on miniyuma (7 resources, 3 tasks) for 1 minute
>    serialcrawler returned a valid solution of value 200 in 4 seconds
> 
> Seong-Bin is working on a problem generator which we can use to test Min's
> solver more thoroughly.
> 
> Alejandro may work on another type of solver which translates a Marbles 2
> into a Marbles 1 problem statement and back.
> 
> Min will work on a "problem verifier" package that you can give a problem
> definition and it will tell you if there are any tasks in it that could
> never be fulfilled, even if there were no competing tasks; it will also
> tell you if there are any useless resources that can never be used by any
> task . He'll also work on a "solution verifier" package that given a
> problem definition and a solution will tell you if the solution really is
> one (no double-booking of resources, no booking of resources outside their
> availability, and so on.)
> 
> I will intermittenty improve on my SNAP-file-to-marbles2-problem generator
> that currently produces only simulator tasks so we have realistic domain
> problems for Marbles2 testing.
> 
> Cheers, yours, Martin
> 
> _______________________________________________
> marbles-isi mailing list
> marbles-isi@mailman.isi.edu
> http://mailman.isi.edu/mailman/listinfo/marbles-isi

-- 
Alejandro Bugacov                                
USC Information Sciences Institute  
4676 Admiralty Way, Suite 1001                                
Marina del Rey, CA  90292               
Voice:   (310) 448-8269
Fax:     (310) 823-6714
http://www.isi.edu/~bugacov
_______________________________________________
marbles-isi mailing list
marbles-isi@mailman.isi.edu
http://mailman.isi.edu/mailman/listinfo/marbles-isi

From marbles-isi@mailman.isi.edu  Wed Sep 19 23:30:13 2001
From: marbles-isi@mailman.isi.edu (Alejandro Bugacov)
Date: Wed, 19 Sep 2001 15:30:13 -0700
Subject: [Marbles-isi] we have a Marbles2 solver!
References: <41C31855E83ED31199D300C04F57A0734C1A63@hermes.isi.edu>
Message-ID: <3BA91C75.DD6A71DC@isi.edu>


Bob Neches wrote:
> 
> Cool!
> 
> An ATTEND question: with the information in the representation that Min will
> be analzying to identify unfullfillable tasks and useless resources, would
> this support any of the partitioning work we've discussed in ATTEND?
> 
that's a good point, i believe it will provide a filtering of tasks 
that can never be filled, which can be seem as an initial partinioning 
or grouping of tasks. 
alejandro. 


> -- Bob
> 
> -----Original Message-----
> From: Martin Frank [mailto:frank@ISI.EDU]
> Sent: Wednesday, September 19, 2001 2:57 PM
> To: marbles-isi@ISI.EDU
> Cc: mcai@ISI.EDU
> Subject: [Marbles-isi] we have a Marbles2 solver!
> 
> All -
> 
> Min has checked in the first Marbles2 solver (centralized and not real-time
> for now, but at least on par with SNAP's current algorithm). Say "make
> cvsupdate" in marbles2 to get it. Run the shell script
> marbles2/main/marbles2.sh. It solves the "miniyuma" problem I made up by
> hand fine, as follows:
> 
>    invoking serialcrawler on miniyuma (7 resources, 3 tasks) for 1 minute
>    serialcrawler returned a valid solution of value 200 in 4 seconds
> 
> Seong-Bin is working on a problem generator which we can use to test Min's
> solver more thoroughly.
> 
> Alejandro may work on another type of solver which translates a Marbles 2
> into a Marbles 1 problem statement and back.
> 
> Min will work on a "problem verifier" package that you can give a problem
> definition and it will tell you if there are any tasks in it that could
> never be fulfilled, even if there were no competing tasks; it will also
> tell you if there are any useless resources that can never be used by any
> task . He'll also work on a "solution verifier" package that given a
> problem definition and a solution will tell you if the solution really is
> one (no double-booking of resources, no booking of resources outside their
> availability, and so on.)
> 
> I will intermittenty improve on my SNAP-file-to-marbles2-problem generator
> that currently produces only simulator tasks so we have realistic domain
> problems for Marbles2 testing.
> 
> Cheers, yours, Martin
> 
> _______________________________________________
> marbles-isi mailing list
> marbles-isi@mailman.isi.edu
> http://mailman.isi.edu/mailman/listinfo/marbles-isi
> _______________________________________________
> marbles-isi mailing list
> marbles-isi@mailman.isi.edu
> http://mailman.isi.edu/mailman/listinfo/marbles-isi

-- 
Alejandro Bugacov                                
USC Information Sciences Institute  
4676 Admiralty Way, Suite 1001                                
Marina del Rey, CA  90292               
Voice:   (310) 448-8269
Fax:     (310) 823-6714
http://www.isi.edu/~bugacov

From marbles-isi@mailman.isi.edu  Wed Sep 19 23:40:19 2001
From: marbles-isi@mailman.isi.edu (Seongbin Park)
Date: Wed, 19 Sep 2001 15:40:19 -0700
Subject: [Marbles-isi] we have a Marbles2 solver!
In-Reply-To: Your message of "Wed, 19 Sep 2001 15:12:52 PDT."
 <41C31855E83ED31199D300C04F57A0734C1A63@hermes.isi.edu>
Message-ID: <200109192240.f8JMeJv12039@tnt.isi.edu>

> Cool!
> 
> An ATTEND question: with the information in the representation that Min will
> be analzying to identify unfullfillable tasks and useless resources, would
> this support any of the partitioning work we've discussed in ATTEND?
> 
> -- Bob
> 

I'm afraid that I don't know exactly how it can support the partitioning -- 
this is a high level description that needs to be tested, but my understanding 
is that "partitioning" has to do with "balancing" (i.e., how to keep the 
balance of the workload among participants, which may not be necessarily mean 
that all needs to be averaged over the set of participants, but rather as in 
the case of quicksort, certain constant factor that will help an algorithm 
yield a good performance behavior would suffice.).   Also, "combining cost" 
needs to be considered in such a way that the cost of solving each problem 
individually and combining all does not exceed that of directly solving the 
problem without partitioning.

As of yet, I don't know the relationship among the information inside the 
representation, balancing, and combining cost yet, but they look 
inter-related....

Seongbin



> -----Original Message-----
> From: Martin Frank [mailto:frank@ISI.EDU]
> Sent: Wednesday, September 19, 2001 2:57 PM
> To: marbles-isi@ISI.EDU
> Cc: mcai@ISI.EDU
> Subject: [Marbles-isi] we have a Marbles2 solver!
> 
> 
> All -
> 
> Min has checked in the first Marbles2 solver (centralized and not real-time 
> for now, but at least on par with SNAP's current algorithm). Say "make 
> cvsupdate" in marbles2 to get it. Run the shell script 
> marbles2/main/marbles2.sh. It solves the "miniyuma" problem I made up by 
> hand fine, as follows:
> 
>    invoking serialcrawler on miniyuma (7 resources, 3 tasks) for 1 minute
>    serialcrawler returned a valid solution of value 200 in 4 seconds
> 
> Seong-Bin is working on a problem generator which we can use to test Min's 
> solver more thoroughly.
> 
> Alejandro may work on another type of solver which translates a Marbles 2 
> into a Marbles 1 problem statement and back.
> 
> Min will work on a "problem verifier" package that you can give a problem 
> definition and it will tell you if there are any tasks in it that could 
> never be fulfilled, even if there were no competing tasks; it will also 
> tell you if there are any useless resources that can never be used by any 
> task . He'll also work on a "solution verifier" package that given a 
> problem definition and a solution will tell you if the solution really is 
> one (no double-booking of resources, no booking of resources outside their 
> availability, and so on.)
> 
> I will intermittenty improve on my SNAP-file-to-marbles2-problem generator 
> that currently produces only simulator tasks so we have realistic domain 
> problems for Marbles2 testing.
> 
> Cheers, yours, Martin
> 
> _______________________________________________
> marbles-isi mailing list
> marbles-isi@mailman.isi.edu
> http://mailman.isi.edu/mailman/listinfo/marbles-isi
> _______________________________________________
> marbles-isi mailing list
> marbles-isi@mailman.isi.edu
> http://mailman.isi.edu/mailman/listinfo/marbles-isi



From marbles-isi@mailman.isi.edu  Wed Sep 19 23:40:19 2001
From: marbles-isi@mailman.isi.edu (Seongbin Park)
Date: Wed, 19 Sep 2001 15:40:19 -0700
Subject: [Marbles-isi] we have a Marbles2 solver!
In-Reply-To: Your message of "Wed, 19 Sep 2001 15:12:52 PDT."
 <41C31855E83ED31199D300C04F57A0734C1A63@hermes.isi.edu>
Message-ID: <200109192240.f8JMeJv12039@tnt.isi.edu>

> Cool!
> 
> An ATTEND question: with the information in the representation that Min will
> be analzying to identify unfullfillable tasks and useless resources, would
> this support any of the partitioning work we've discussed in ATTEND?
> 
> -- Bob
> 

I'm afraid that I don't know exactly how it can support the partitioning -- 
this is a high level description that needs to be tested, but my understanding 
is that "partitioning" has to do with "balancing" (i.e., how to keep the 
balance of the workload among participants, which may not be necessarily mean 
that all needs to be averaged over the set of participants, but rather as in 
the case of quicksort, certain constant factor that will help an algorithm 
yield a good performance behavior would suffice.).   Also, "combining cost" 
needs to be considered in such a way that the cost of solving each problem 
individually and combining all does not exceed that of directly solving the 
problem without partitioning.

As of yet, I don't know the relationship among the information inside the 
representation, balancing, and combining cost yet, but they look 
inter-related....

Seongbin



> -----Original Message-----
> From: Martin Frank [mailto:frank@ISI.EDU]
> Sent: Wednesday, September 19, 2001 2:57 PM
> To: marbles-isi@ISI.EDU
> Cc: mcai@ISI.EDU
> Subject: [Marbles-isi] we have a Marbles2 solver!
> 
> 
> All -
> 
> Min has checked in the first Marbles2 solver (centralized and not real-time 
> for now, but at least on par with SNAP's current algorithm). Say "make 
> cvsupdate" in marbles2 to get it. Run the shell script 
> marbles2/main/marbles2.sh. It solves the "miniyuma" problem I made up by 
> hand fine, as follows:
> 
>    invoking serialcrawler on miniyuma (7 resources, 3 tasks) for 1 minute
>    serialcrawler returned a valid solution of value 200 in 4 seconds
> 
> Seong-Bin is working on a problem generator which we can use to test Min's 
> solver more thoroughly.
> 
> Alejandro may work on another type of solver which translates a Marbles 2 
> into a Marbles 1 problem statement and back.
> 
> Min will work on a "problem verifier" package that you can give a problem 
> definition and it will tell you if there are any tasks in it that could 
> never be fulfilled, even if there were no competing tasks; it will also 
> tell you if there are any useless resources that can never be used by any 
> task . He'll also work on a "solution verifier" package that given a 
> problem definition and a solution will tell you if the solution really is 
> one (no double-booking of resources, no booking of resources outside their 
> availability, and so on.)
> 
> I will intermittenty improve on my SNAP-file-to-marbles2-problem generator 
> that currently produces only simulator tasks so we have realistic domain 
> problems for Marbles2 testing.
> 
> Cheers, yours, Martin
> 
> _______________________________________________
> marbles-isi mailing list
> marbles-isi@mailman.isi.edu
> http://mailman.isi.edu/mailman/listinfo/marbles-isi
> _______________________________________________
> marbles-isi mailing list
> marbles-isi@mailman.isi.edu
> http://mailman.isi.edu/mailman/listinfo/marbles-isi



From marbles-isi@mailman.isi.edu  Thu Sep 20 00:06:02 2001
From: marbles-isi@mailman.isi.edu (Seongbin Park)
Date: Wed, 19 Sep 2001 16:06:02 -0700
Subject: (an afterthought) Re: [Marbles-isi] we have a Marbles2 solver!
In-Reply-To: Your message of "Wed, 19 Sep 2001 15:40:19 PDT."
 <200109192240.f8JMeJv12039@tnt.isi.edu>
Message-ID: <200109192306.f8JN62v23060@tnt.isi.edu>

Hi,

I sent paragraphs, but there were parts that I have had asummed something:

> I'm afraid that I don't know exactly how it can support the partitioning -- 
> this is a high level description that needs to be tested, but my understanding 
> is that "partitioning" has to do with "balancing" (i.e., how to keep the 
> balance of the workload among participants, which may not be necessarily mean 
> that all needs to be averaged over the set of participants, but rather as in 
> the case of quicksort, certain constant factor that will help an algorithm 
> yield a good performance behavior would suffice.).   

==> This "balancing" factor may not matter much if we partition the whole 
problem once and for all at some point.   On the other hand, if we are to 
"recursively" the partitoned problems again and again (i.e., recursing on each 
of the partitioned part), maintaining constant balance factor is important 
since that contributes to the height of the tree that represents
the number of steps by making it proportional to a logarithm factor.  (i.e., 
O(log n)), where n is the size of the input.

Seongbin


From marbles-isi@mailman.isi.edu  Thu Sep 20 00:06:02 2001
From: marbles-isi@mailman.isi.edu (Seongbin Park)
Date: Wed, 19 Sep 2001 16:06:02 -0700
Subject: (an afterthought) Re: [Marbles-isi] we have a Marbles2 solver!
In-Reply-To: Your message of "Wed, 19 Sep 2001 15:40:19 PDT."
 <200109192240.f8JMeJv12039@tnt.isi.edu>
Message-ID: <200109192306.f8JN62v23060@tnt.isi.edu>

Hi,

I sent paragraphs, but there were parts that I have had asummed something:

> I'm afraid that I don't know exactly how it can support the partitioning -- 
> this is a high level description that needs to be tested, but my understanding 
> is that "partitioning" has to do with "balancing" (i.e., how to keep the 
> balance of the workload among participants, which may not be necessarily mean 
> that all needs to be averaged over the set of participants, but rather as in 
> the case of quicksort, certain constant factor that will help an algorithm 
> yield a good performance behavior would suffice.).   

==> This "balancing" factor may not matter much if we partition the whole 
problem once and for all at some point.   On the other hand, if we are to 
"recursively" the partitoned problems again and again (i.e., recursing on each 
of the partitioned part), maintaining constant balance factor is important 
since that contributes to the height of the tree that represents
the number of steps by making it proportional to a logarithm factor.  (i.e., 
O(log n)), where n is the size of the input.

Seongbin


From marbles-isi@mailman.isi.edu  Fri Sep 21 00:03:24 2001
From: marbles-isi@mailman.isi.edu (Martin Frank)
Date: Thu, 20 Sep 2001 16:03:24 -0700
Subject: [Marbles-isi] marbles2 now compiles again
Message-ID: <5.1.0.14.0.20010920160244.049f0c00@tnt.isi.edu>

You may have to manually remove some files in marbles2/problemgen.

Cheers, yours, Martin


From marbles-isi@mailman.isi.edu  Wed Sep 26 20:13:46 2001
From: marbles-isi@mailman.isi.edu (Martin Frank)
Date: Wed, 26 Sep 2001 12:13:46 -0700
Subject: [Marbles-isi] checked in: TimeInterval(s) are history
Message-ID: <3BB228EA.9020408@isi.edu>

util/TimeInterval is replaced by util/time/MtInterval
util/TimeIntervals is replaced by util/time/MtMergingIntervals

There are also new abstractions now, such as MtOverlappingTimeIntervals 
and MtTouchingIntervals.

The new abstractions do now allow client code to directly access  and 
manipulate the underlying MtIntervals list without validation, which 
should dramatically reduce the amount of time we spend debugging 
time-related code.

SNAP: you can run Aug-27-31-hardy.snap, I converted all the XML files 
that it needs. There are converters checked in for all files that need 
them, under runs/*.sed. Use them like this to convert more files: sed -f 
snapfile-converter.sed <old.snap >new.snap (I'll also do more converting 
myself after lunch.)

ATTEND: I changed the (very few) places in which TimeInterval(s) was 
used in solvers/centralized/randomtime.

I recommend you say "make fresh" in your top project directory.

Cheers, yours, Martin