Is there a size issue within MSP that limits Sensitivity Analyseswith Risk+

[cyager]

I am working in a file with about 28K lines. The file size is roughly
80MB. I'm using a relatively newer computer with 4 GB's RAM, using
Vista, and using MSP 2007. I am attempting a test run of a sensitivity
analysis on using only 5 iterations within the test. Will it work, or
is the project just too big? Also, just to mention, I stripped out any
macros, modules, formula's, and lookups. The file is clean. Thanks for
your help in advance.

 

[Rob Schneider]

I am working in a file with about 28K lines. The file size is roughly
80MB. I'm using a relatively newer computer with 4 GB's RAM, using
Vista, and using MSP 2007. I am attempting a test run of a sensitivity
analysis on using only 5 iterations within the test. Will it work, or
is the project just too big? Also, just to mention, I stripped out any
macros, modules, formula's, and lookups. The file is clean. Thanks for
your help in advance.

Project can, by the specification (see the Help file) by my recollection
handle millions of tasks. I don't exactly remember the number as it's
so big I've never come close.

28,000 tasks is a lot and is probably more than a normal PM or planner
can handle.

I don't know what you mean by "iterations". Project doesn't really
interate on anything (to my knowledge). What are you iterating and,
assuming this is an automated process, *who* wrote the code to interate
*what*?

 

[cyager]

Project can, by the specification (see the Help file) by my recollection
handle millions of tasks.  I don't exactly remember the number as it's
so big I've never come close.

28,000 tasks is a lot and is probably more than a normal PM or planner
can handle.

I don't know what you mean by "iterations".  Project doesn't really
interate on anything (to my knowledge).  What are you iterating and,
assuming this is an automated process, *who* wrote the code to interate
*what*?

Thank you for your response. Let me be more clear then, sorry. First
of all, I agree the file is large. Secondly, I certainly understand
the limitations of MSP, at least I hope I do. My question for the
forum is more about Deltek's Risk+ tool and how it interacts with
Microsoft Project. I read back to my original post and realized, that
I wasn't too clear, sorry again. Does anyone have any concrete
knowledge on capacity issues when running the Risk+ Sensitivity
Analyses. I am currently running the Sensitivity Analysis on the
project I mentioned earlier. I started at noon on Friday and did a
very small sample size of 5 Iterations for the analysis. It is now 9
a.m. Saturday and it is just about halfway done. That is pretty crazy
for just five iterations, so if anyone out there has some good
experienced knowledge on running this Risk+ tool with MSP, could you
pass it along?

 

[Rob Schneider]

Thank you for your response. Let me be more clear then, sorry. First
of all, I agree the file is large. Secondly, I certainly understand
the limitations of MSP, at least I hope I do. My question for the
forum is more about Deltek's Risk+ tool and how it interacts with
Microsoft Project. I read back to my original post and realized, that
I wasn't too clear, sorry again. Does anyone have any concrete
knowledge on capacity issues when running the Risk+ Sensitivity
Analyses. I am currently running the Sensitivity Analysis on the
project I mentioned earlier. I started at noon on Friday and did a
very small sample size of 5 Iterations for the analysis. It is now 9
a.m. Saturday and it is just about halfway done. That is pretty crazy
for just five iterations, so if anyone out there has some good
experienced knowledge on running this Risk+ tool with MSP, could you
pass it along?

I've never seen this tool discussed here. You'd be better to check out
http://www.deltek.com/communities/

My hunch, and it's just a hunch, is that 28,000 tasks is too much for
Risk+. Nothing to do with Project. I have had great experience with a
couple other Monte Carlo simulation programs but not this one. My
experience suggests keeping the model very simple. Has to be
simple-enough for people to understand it. 28,000 seems too much even
for the best people. I've at most had a few hundred lines even for
models of huge projects. Keep it simple.

 

[cyager]

I've never seen this tool discussed here. You'd be better to check outhttp://www.deltek.com/communities/

My hunch, and it's just a hunch, is that 28,000 tasks is too much for
Risk+. Nothing to do with Project. I have had great experience with a
couple other Monte Carlo simulation programs but not this one.  My
experience suggests keeping the model very simple.  Has to be
simple-enough for people to understand it.  28,000 seems too much even
for the best people.  I've at most had a few hundred lines even for
models of huge projects.  Keep it simple.- Hide quoted text -

- Show quoted text -

Any tools that you could suggest that could handle a monte carlo with
this many tasks? Thanks for responding on a Saturday!

 

[Rob Schneider]

Any tools that you could suggest that could handle a monte carlo with
this many tasks? Thanks for responding on a Saturday!

Sorry ... don't have a clue. 28,000 seems like an awful lot for any of
the add-ins. You are orders of magnitude beyond anything I've ever
tried nor wanted to try.

I guess if I were you and the modelling required this much detail, I'd
get a developer to do a bespoke development of some sort.

 

[cyager]

Sorry ... don't have a clue. 28,000 seems like an awful lot for any of
the add-ins.  You are orders of magnitude beyond anything I've ever
tried nor wanted to try.

I guess if I were you and the modelling required this much detail, I'd
get a developer to do a bespoke development of some sort.- Hide quoted text -

- Show quoted text -

Thanks for the advice. Have a good weekend.

 

[Jim Aksel]

Sorry for jumping in so late, I use Risk+ frequently and have some suggestions.
First, the issue is clearly with Risk+ ... with 28,000 tasks and a large
number of iterations, sooner or later you might snag some type of divide by
zero or negative number that blows things up. However, I do find that if you
supress error notification in MS Project, it generally runs pretty well.

As for your file size, I've done master projects with 95 inserted files and
about 200-1000 lines in some of the files. It eventually runs. What I did
was fire it up on Firday night and come back Monday morning and it was done.
How long? Forever...

I talked to Deltek about this a few years ago. The drivers are (1) Your
hardware including processing power and RAM including network architecture if
the mpp file is on a network share (2) The Risk+ architecture itself. What
happens is the screen wants to refresh after every iteration to show you the
progress ... nice eye candy but it slows you down to a crawl. There is no
setting in Risk+ to avoid this.

If you are shopping for another product, try PrimaVera PERTMASTER . I just
tried the link but their site is down.
www.primavera.com/products/pertmaster

I've used this, it is a much stronger product (which is purely my opinion).
It does have some drawbacks however, namely you have to import Project files
into it, it is not a Project add-on. Additionally, you have to tweak the file
some becuase Project and Primavera have different understandings of
"remaining work/duration." We have a Pertmaster Risk guy on staff, and would
be happy to help if requested. I don't want to publish my corporate e-mail
here but jeaksel (at) yahoo (dot) com seems to work pretty well . I'm on
vacation but can get back to you if you have further questions on the product
after June 15th.

As for speed, well it is ligthning fast compared to Risk+. I ran a 2000
line file on Risk+ that took a few hours. On Pertmaster, it took about 3
minutes -- including the screen eye candy. If I turn off the eye candy, it
takes about 5 seconds. I have brought this to Deltek's attention and have
seen no response.

--
If this post was helpful, please consider rating it.

Jim Aksel, MVP

Check out my blog for more information:
http://www.msprojectblog.com

 

[Rob Schneider]

Sorry for jumping in so late, I use Risk+ frequently and have some suggestions.
First, the issue is clearly with Risk+ ... with 28,000 tasks and a large
number of iterations, sooner or later you might snag some type of divide by
zero or negative number that blows things up. However, I do find that if you
supress error notification in MS Project, it generally runs pretty well.

As for your file size, I've done master projects with 95 inserted files and
about 200-1000 lines in some of the files. It eventually runs. What I did
was fire it up on Firday night and come back Monday morning and it was done.
How long? Forever...

I talked to Deltek about this a few years ago. The drivers are (1) Your
hardware including processing power and RAM including network architecture if
the mpp file is on a network share (2) The Risk+ architecture itself. What
happens is the screen wants to refresh after every iteration to show you the
progress ... nice eye candy but it slows you down to a crawl. There is no
setting in Risk+ to avoid this.

If you are shopping for another product, try PrimaVera PERTMASTER . I just
tried the link but their site is down.
www.primavera.com/products/pertmaster

I've used this, it is a much stronger product (which is purely my opinion).
It does have some drawbacks however, namely you have to import Project files
into it, it is not a Project add-on. Additionally, you have to tweak the file
some becuase Project and Primavera have different understandings of
"remaining work/duration." We have a Pertmaster Risk guy on staff, and would
be happy to help if requested. I don't want to publish my corporate e-mail
here but jeaksel (at) yahoo (dot) com seems to work pretty well . I'm on
vacation but can get back to you if you have further questions on the product
after June 15th.

As for speed, well it is ligthning fast compared to Risk+. I ran a 2000
line file on Risk+ that took a few hours. On Pertmaster, it took about 3
minutes -- including the screen eye candy. If I turn off the eye candy, it
takes about 5 seconds. I have brought this to Deltek's attention and have
seen no response.

Jim,

Cool. I'm guessing PERTMASTER is so fast since it's not using Project
for computation engine--just for the model. Risk+ and others (I use
@Risk) use Project for the computations which has overhead which will be
very noticable for large models (hence my concern about 28,000 tasks).
Worse I can see how it would be difficult for the developer of the
add-in do much optimisation since they don't control the compute engine.
Or do they? Who knows.

(Full disclosure: I learned Monte Carlo simulation in the late 1980's on
a Cray with computations done in FORTRAN. So I'm a bit biased on how to
architect this sort of thing. It wasn't schedules that we were
simulating, but as I look back we could have. Probably schedule
simulation would have been worth more to the company that what we
actually doing!).

I'm going to look into PERTMASTER. Thanks for the tip. I may have an
upcoming need for modelling something which will have a lot of tasks
simply because we are aggregating the work of dozens of people. Not
close to 28,000, but a lot none the less. We are thinking about want to
take the data "as is" instead of doing much mods to it to enable a
daily/weekly probabilistic forecast. Others have done this, so I know
possible.

 

[cyager]

Jim,

Cool.  I'm guessing PERTMASTER is so fast since it's not using Project
for computation engine--just for the model.  Risk+ and others (I use
@Risk) use Project for the computations which has overhead which will be
very noticable for large models (hence my concern about 28,000 tasks).
Worse I can see how it would be difficult for the developer of the
add-in do much optimisation since they don't control the compute engine.
   Or do they? Who knows.

(Full disclosure: I learned Monte Carlo simulation in the late 1980's on
a Cray with computations done in FORTRAN.  So I'm a bit biased on how to
architect this sort of thing. It wasn't schedules that we were
simulating, but as I look back we could have.  Probably schedule
simulation would have been worth more to the company that what we
actually doing!).

I'm going to look into PERTMASTER.  Thanks for the tip. I may have an
upcoming need for modelling something which will have a lot of tasks
simply because we are aggregating the work of dozens of people.  Not
close to 28,000, but a lot none the less. We are thinking about want to
take the data "as is" instead of doing much mods to it to enable a
daily/weekly probabilistic forecast.  Others have done this, so I know
possible.- Hide quoted text -

- Show quoted text -

Jim:

It's like you were reading my mind. Pertmaster was my next logical
step. My biggest hurdle with Pertmaster is that this file used to be a
ton of subprojects with multiple calendars. I've had issues with
converting files with multiple calendars into Pertmaster. Great
advice, both of you. I'll let the group know how it goes.

And by the way, that 5 iteration Sensitivity Analsysis I talked about
earlier. It's still going.....