[Met_help] [rt.rap.ucar.edu #60620] History for MET computational cost

[John Halley Gotway via RT]

----------------------------------------------------------------
  Initial Request
----------------------------------------------------------------

Hello.
I am looking at MET to do some analysis. I was just wondering how computationally expensive each of the tools could be. I know this will depend on the data but I am hoping for some general idea based on your experience with different kinds of data.

Thanks,
-Hamid

Dr. Amidu (Hamid) Oloso, PMP
Science Systems and Applications, Inc.
Code 610.3, NASA Goddard Space Flight Center
Greenbelt, MD 20771

(301)286-7885 (301)286-1775(f) Amidu.O.Oloso at nasa.gov

----------------------------------------------------------------
  Complete Ticket History
----------------------------------------------------------------

Subject: Re: [rt.rap.ucar.edu #60620] MET computational cost
From: John Halley Gotway
Time: Tue Mar 12 14:28:37 2013

Hamid,

Generally speaking, the MET tools run much, much quicker than the
models the are used to verify.  The tools run on a single processor,
rather than multiple processor, because each individual
verification task is generally not the computationally intensive.

That being said, there are times when the MET tools can run slowly and
consume too much memory:
  - Computing bootstrap confidence intervals and rank correlation
coefficients over a large number of matched pairs can be costly, but
they can easily be disabled in the config files.
  - The Ensemble-Stat and Series-Analysis (new in upcoming release of
METv4.1) can consume too much memory, but we're working to address
that.

There is some spin-up time in learning how to use the MET tools and
writing scripts to configure and run them to solve your particular
verification problem.  But generally speaking, the computational
expense in not very high, especially when compare the NWP model output
that they verify.

Hope that helps clarify.

Thanks,
John Halley Gotway
met_help at ucar.edu

On 03/12/2013 02:21 PM, Oloso, Amidu O.[SCIENCE SYSTEMS AND
APPLICATIONS INC] via RT wrote:
>
> Tue Mar 12 14:21:33 2013: Request 60620 was acted upon.
> Transaction: Ticket created by amidu.o.oloso at nasa.gov
>         Queue: met_help
>       Subject: MET computational cost
>         Owner: Nobody
>    Requestors: amidu.o.oloso at nasa.gov
>        Status: new
>   Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=60620 >
>
>
> Hello.
> I am looking at MET to do some analysis. I was just wondering how
computationally expensive each of the tools could be. I know this will
depend on the data but I am hoping for some general idea based on your
experience with different kinds of data.
>
> Thanks,
> -Hamid
>
> Dr. Amidu (Hamid) Oloso, PMP
> Science Systems and Applications, Inc.
> Code 610.3, NASA Goddard Space Flight Center
> Greenbelt, MD 20771
>
> (301)286-7885 (301)286-1775(f) Amidu.O.Oloso at nasa.gov
>

------------------------------------------------
Subject: RE: [rt.rap.ucar.edu #60620] MET computational cost
From: Oloso, Amidu O.[SCIENCE SYSTEMS AND APPLICATIONS INC]
Time: Tue Mar 12 15:07:58 2013

John,
Thank you very much for the quick response. Yes this helps.

-Hamid

________________________________________
From: John Halley Gotway via RT [met_help at ucar.edu]
Sent: Tuesday, March 12, 2013 4:28 PM
To: Oloso, Amidu O. (GSFC-610.3)[SCIENCE SYSTEMS AND APPLICATIONS INC]
Subject: Re: [rt.rap.ucar.edu #60620] MET computational cost

Hamid,

Generally speaking, the MET tools run much, much quicker than the
models the are used to verify.  The tools run on a single processor,
rather than multiple processor, because each individual
verification task is generally not the computationally intensive.

That being said, there are times when the MET tools can run slowly and
consume too much memory:
  - Computing bootstrap confidence intervals and rank correlation
coefficients over a large number of matched pairs can be costly, but
they can easily be disabled in the config files.
  - The Ensemble-Stat and Series-Analysis (new in upcoming release of
METv4.1) can consume too much memory, but we're working to address
that.

There is some spin-up time in learning how to use the MET tools and
writing scripts to configure and run them to solve your particular
verification problem.  But generally speaking, the computational
expense in not very high, especially when compare the NWP model output
that they verify.

Hope that helps clarify.

Thanks,
John Halley Gotway
met_help at ucar.edu

On 03/12/2013 02:21 PM, Oloso, Amidu O.[SCIENCE SYSTEMS AND
APPLICATIONS INC] via RT wrote:
>
> Tue Mar 12 14:21:33 2013: Request 60620 was acted upon.
> Transaction: Ticket created by amidu.o.oloso at nasa.gov
>         Queue: met_help
>       Subject: MET computational cost
>         Owner: Nobody
>    Requestors: amidu.o.oloso at nasa.gov
>        Status: new
>   Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=60620 >
>
>
> Hello.
> I am looking at MET to do some analysis. I was just wondering how
computationally expensive each of the tools could be. I know this will
depend on the data but I am hoping for some general idea based on your
experience with different kinds of data.
>
> Thanks,
> -Hamid
>
> Dr. Amidu (Hamid) Oloso, PMP
> Science Systems and Applications, Inc.
> Code 610.3, NASA Goddard Space Flight Center
> Greenbelt, MD 20771
>
> (301)286-7885 (301)286-1775(f) Amidu.O.Oloso at nasa.gov
>


------------------------------------------------